[
{
"complete": "Generate a 3D model of a glass 20-sided die. Ensure the result is high quality. The generation speed should be medium speed. Consider that this is a high fidelity mesh. Save the output to ./outputs/a_glass_20-sided_die.ply.\nThen, take the mesh at the output path of the previous step and generate a point cloud with 15000 samples. Store the output at ./outputs/example_mesh_from_ply_15000pts.ply.",
"file_path": null,
"ground_truth": [
{
"task": "Text to 3D Generation",
"quality": "High quality",
"speed": "Medium speed",
"usage": "object",
"output_format": "ply",
"output_path": "./outputs/a_glass_20-sided_die.ply",
"extra_info": "High fidelity mesh"
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 15000,
"output_path": "./outputs/example_mesh_from_ply_15000pts.ply"
}
],
"gt_summary": " Text to 3D Generation, Post Processing \n Generate a 3D model of a glass 20-sided die. Ensure the result is high quality. The generation speed should be medium speed. Consider that this is a high fidelity mesh. Save the output to ./outputs/a_glass_20-sided_die.ply.\nThen, take the mesh at the output path of the previous step and generate a point cloud with 15000 samples. Store the output at ./outputs/example_mesh_from_ply_15000pts.ply. \n Input 1:None; Input 2:./outputs/a_glass_20-sided_die.ply; \n Step 1: { Constraints: [quality: High quality, speed: Medium speed, features: High fidelity mesh]; Output Path: ./outputs/a_glass_20-sided_die.ply }; Step 2: { number_of_points: 15000; Output Path: ./outputs/example_mesh_from_ply_15000pts.ply } ",
"code": "self.executor.text_to_3d_generation(prompt=\"Generate a 3D model of a glass 20-sided die\", model=\"trellis\", output_path=\"./outputs/a_glass_20-sided_die.ply\")\nself.executor.post_processing(input_path='./outputs/a_glass_20-sided_die.ply', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_ply_15000pts.ply', parameters={\"number_of_points\": 15000})",
"id": 0,
"ambiguous": "Generate an object.\nThen, Convert the file at the output of the previous step.",
"missing_info": "Generate a 3D model. The model quality should be specified. Expect appropriate processing speed. Export format: 3D model. Focus on certain features. Save the output to somewhere.\nThen, I need to sample a number of points uniformly from the 3D model at a path. Output the processed ply file to a location.",
"mistake": "Generate a 3D model of a glass 20-sided die. Ensure the result is high quality. The generation speed should be medium speed. Consider that this is a high fidelity mesh. Save the output to ./outputs/a_glass_20-sided_die.ply.\nThen, take the mesh at the output path of the previous step and generate a point cloud with 15000 samples. Store the output at ./outputs/example_mesh_from_ply_15000pts.ply. Ensure low. Ensure extremely high speed. Ensure low quality.",
"mistake_type": "quality_speed"
},
{
"complete": "Generate a 3D model of a beautiful rainbow fish. I require high quality output. Expect medium speed processing. I need a ply file. It should be a high fidelity mesh. Store the generated model at ./outputs/a_beautiful_rainbow_fish.ply.\nThen, take the mesh at the output path of the previous step and generate a point cloud with 35000 samples. Store the output at ./outputs/example_mesh_from_ply_35000pts.ply.",
"file_path": null,
"ground_truth": [
{
"task": "Text to 3D Generation",
"quality": "High quality",
"speed": "Medium speed",
"usage": "object",
"output_format": "ply",
"output_path": "./outputs/a_beautiful_rainbow_fish.ply",
"extra_info": "High fidelity mesh"
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 35000,
"output_path": "./outputs/example_mesh_from_ply_35000pts.ply"
}
],
"gt_summary": " Text to 3D Generation, Post Processing \n Generate a 3D model of a beautiful rainbow fish. I require high quality output. Expect medium speed processing. I need a ply file. It should be a high fidelity mesh. Store the generated model at ./outputs/a_beautiful_rainbow_fish.ply.\nThen, take the mesh at the output path of the previous step and generate a point cloud with 35000 samples. Store the output at ./outputs/example_mesh_from_ply_35000pts.ply. \n Input 1:None; Input 2:./outputs/a_beautiful_rainbow_fish.ply; \n Step 1: { Constraints: [quality: High quality, speed: Medium speed, output_format: ply, features: High fidelity mesh]; Output Path: ./outputs/a_beautiful_rainbow_fish.ply }; Step 2: { number_of_points: 35000; Output Path: ./outputs/example_mesh_from_ply_35000pts.ply } ",
"code": "self.executor.text_to_3d_generation(prompt=\"Generate a 3D model of a beautiful rainbow fish\", model=\"trellis\", output_path=\"./outputs/a_beautiful_rainbow_fish.ply\")\nself.executor.post_processing(input_path='./outputs/a_beautiful_rainbow_fish.ply', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_ply_35000pts.ply', parameters={\"number_of_points\": 35000})",
"id": 1,
"ambiguous": "Make a 3D thing.\nThen, Do something with the output of the previous step.",
"missing_info": "Generate a 3D model. Ensure the result quality is defined. Expect appropriate processing speed. I need a file. Focus on certain features. The result should be saved in a location.\nThen, Take the mesh at a location and generate a point cloud with a certain number of samples. Store the output at a path.",
"mistake": "Generate a 3D model of a beautiful rainbow fish. I require high quality output. Expect medium speed processing. I need a ply file. It should be a high fidelity mesh. Store the generated model at ./outputs/a_beautiful_rainbow_fish.ply.\nThen, take the mesh at the output path of the previous step and generate a point cloud with 35000 samples. Store the output at ./outputs/example_mesh_from_ply_35000pts.ply. Ensure low. Ensure extremely high speed. Ensure low quality.",
"mistake_type": "quality_speed"
},
{
"complete": "Generate a 3D model of a glass 20-sided die. Ensure the result is high quality. Save the output to ./outputs/a_glass_20-sided_die.ply.\nThen, i need to sample 45000 points uniformly from the 3D model at the output path of the previous step. Output the processed ply file to ./outputs/example_mesh_from_obj_45000pts.ply.",
"file_path": null,
"ground_truth": [
{
"task": "Text to 3D Generation",
"quality": "High quality",
"speed": null,
"usage": "object",
"output_format": "ply",
"output_path": "./outputs/a_glass_20-sided_die.ply",
"extra_info": null
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 45000,
"output_path": "./outputs/example_mesh_from_obj_45000pts.ply"
}
],
"gt_summary": " Text to 3D Generation, Post Processing \n Generate a 3D model of a glass 20-sided die. Ensure the result is high quality. Save the output to ./outputs/a_glass_20-sided_die.ply.\nThen, i need to sample 45000 points uniformly from the 3D model at the output path of the previous step. Output the processed ply file to ./outputs/example_mesh_from_obj_45000pts.ply. \n Input 1:None; Input 2:./outputs/a_glass_20-sided_die.ply; \n Step 1: { Constraints: [quality: High quality]; Output Path: ./outputs/a_glass_20-sided_die.ply }; Step 2: { number_of_points: 45000; Output Path: ./outputs/example_mesh_from_obj_45000pts.ply } ",
"code": "self.executor.text_to_3d_generation(prompt=\"Generate a 3D model of a glass 20-sided die\", model=\"trellis\", output_path=\"./outputs/a_glass_20-sided_die.ply\")\nself.executor.post_processing(input_path='./outputs/a_glass_20-sided_die.ply', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_obj_45000pts.ply', parameters={\"number_of_points\": 45000})",
"id": 2,
"ambiguous": "Create something.\nThen, Do something with the output of the previous step.",
"missing_info": "Generate a 3D model. The model quality should be specified. Save the output to a file. Save the output to somewhere.\nThen, I need to sample a number of points uniformly from the 3D model at a path. Output the processed ply file to a location.",
"mistake": "Generate a 3D model of a glass 20-sided die. Ensure the result is high quality. Save the output to ./outputs/a_glass_20-sided_die.ply.\nThen, i need to sample 45000 points uniformly from the 3D model at the output path of the previous step. Output the processed ply file to ./outputs/example_mesh_from_obj_45000pts.ply. Ensure low quality.",
"mistake_type": "quality_speed"
},
{
"complete": "Generate a 3D model of a bald eagle carved out of wood. The model quality should be high quality. I need a glb file. It should be a high fidelity mesh. Store the generated model at ./outputs/a_bald_eagle_carved_out_of_wood.glb.\nThen, perform mesh-to-pointcloud conversion on the mesh at the output path of the previous step. Set the target number of points to 30000 and save as ./outputs/example_mesh_from_obj_30000pts.ply.",
"file_path": null,
"ground_truth": [
{
"task": "Text to 3D Generation",
"quality": "High quality",
"speed": null,
"usage": "object",
"output_format": "glb",
"output_path": "./outputs/a_bald_eagle_carved_out_of_wood.glb",
"extra_info": "High fidelity mesh"
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 30000,
"output_path": "./outputs/example_mesh_from_obj_30000pts.ply"
}
],
"gt_summary": " Text to 3D Generation, Post Processing \n Generate a 3D model of a bald eagle carved out of wood. The model quality should be high quality. I need a glb file. It should be a high fidelity mesh. Store the generated model at ./outputs/a_bald_eagle_carved_out_of_wood.glb.\nThen, perform mesh-to-pointcloud conversion on the mesh at the output path of the previous step. Set the target number of points to 30000 and save as ./outputs/example_mesh_from_obj_30000pts.ply. \n Input 1:None; Input 2:./outputs/a_bald_eagle_carved_out_of_wood.glb; \n Step 1: { Constraints: [quality: High quality, output_format: glb, features: High fidelity mesh]; Output Path: ./outputs/a_bald_eagle_carved_out_of_wood.glb }; Step 2: { number_of_points: 30000; Output Path: ./outputs/example_mesh_from_obj_30000pts.ply } ",
"code": "self.executor.text_to_3d_generation(prompt=\"Generate a 3D model of a bald eagle carved out of wood\", model=\"trellis\", output_path=\"./outputs/a_bald_eagle_carved_out_of_wood.glb\")\nself.executor.post_processing(input_path='./outputs/a_bald_eagle_carved_out_of_wood.glb', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_obj_30000pts.ply', parameters={\"number_of_points\": 30000})",
"id": 3,
"ambiguous": "Generate an object.\nThen, Convert the file at the output of the previous step.",
"missing_info": "Generate a 3D model. Ensure the result quality is defined. Save the output to a file. Consider that this has particular characteristics. The result should be saved in a location.\nThen, Please process the mesh at a file to create a point cloud. Use a sampling rate of a number of points and write the file to somewhere.",
"mistake": "Generate a 3D model of a bald eagle carved out of wood. The model quality should be high quality. I need a glb file. It should be a high fidelity mesh. Store the generated model at ./outputs/a_bald_eagle_carved_out_of_wood.glb.\nThen, perform mesh-to-pointcloud conversion on the mesh at the output path of the previous step. Set the target number of points to 30000 and save as ./outputs/example_mesh_from_obj_30000pts.ply. Ensure low. Ensure low quality.",
"mistake_type": "quality_speed"
},
{
"complete": "Generate a 3D model of a banana peeling itself. Ensure the result is low quality. It should be a simplified model. Save the output to ./outputs/a_banana_peeling_itself.obj.\nThen, convert the mesh file at the output path of the previous step into a point cloud by sampling 30000 points. Save the result to ./outputs/example_mesh_from_obj_30000pts.ply.",
"file_path": null,
"ground_truth": [
{
"task": "Text to 3D Generation",
"quality": "Low quality",
"speed": null,
"usage": "object",
"output_format": "obj",
"output_path": "./outputs/a_banana_peeling_itself.obj",
"extra_info": "Simplified model"
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 30000,
"output_path": "./outputs/example_mesh_from_obj_30000pts.ply"
}
],
"gt_summary": " Text to 3D Generation, Post Processing \n Generate a 3D model of a banana peeling itself. Ensure the result is low quality. It should be a simplified model. Save the output to ./outputs/a_banana_peeling_itself.obj.\nThen, convert the mesh file at the output path of the previous step into a point cloud by sampling 30000 points. Save the result to ./outputs/example_mesh_from_obj_30000pts.ply. \n Input 1:None; Input 2:./outputs/a_banana_peeling_itself.obj; \n Step 1: { Constraints: [quality: Low quality, features: Simplified model]; Output Path: ./outputs/a_banana_peeling_itself.obj }; Step 2: { number_of_points: 30000; Output Path: ./outputs/example_mesh_from_obj_30000pts.ply } ",
"code": "self.executor.text_to_3d_generation(prompt=\"Generate a 3D model of a banana peeling itself\", model=\"triplaneturbo\", output_path=\"./outputs/a_banana_peeling_itself.obj\")\nself.executor.post_processing(input_path='./outputs/a_banana_peeling_itself.obj', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_obj_30000pts.ply', parameters={\"number_of_points\": 30000})",
"id": 4,
"ambiguous": "I want a 3D asset.\nThen, Convert the file at the output of the previous step.",
"missing_info": "Create an object. The model quality should be specified. Export format: 3D model. Consider that this has particular characteristics. Store the generated model.\nThen, Convert the mesh file at a location into a point cloud by sampling a certain number of points. Save the result to somewhere.",
"mistake": "Generate a 3D model of a banana peeling itself. Ensure the result is low quality. It should be a simplified model. Save the output to ./outputs/a_banana_peeling_itself.obj.\nThen, convert the mesh file at the output path of the previous step into a point cloud by sampling 30000 points. Save the result to ./outputs/example_mesh_from_obj_30000pts.ply. Ensure complex model. Ensure high quality.",
"mistake_type": "quality_speed"
},
{
"complete": "Generate a 3D model of a blue motorcycle. Save the output to a obj file. Focus on lightweight model. The result should be saved in ./outputs/a_blue_motorcycle.obj.\nThen, i need to sample 10000 points uniformly from the 3D model at the output path of the previous step. Output the processed ply file to ./outputs/example_mesh_from_glb_10000pts.ply.",
"file_path": null,
"ground_truth": [
{
"task": "Text to 3D Generation",
"quality": null,
"speed": null,
"usage": "object",
"output_format": "obj",
"output_path": "./outputs/a_blue_motorcycle.obj",
"extra_info": "Lightweight model"
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 10000,
"output_path": "./outputs/example_mesh_from_glb_10000pts.ply"
}
],
"gt_summary": " Text to 3D Generation, Post Processing \n Generate a 3D model of a blue motorcycle. Save the output to a obj file. Focus on lightweight model. The result should be saved in ./outputs/a_blue_motorcycle.obj.\nThen, i need to sample 10000 points uniformly from the 3D model at the output path of the previous step. Output the processed ply file to ./outputs/example_mesh_from_glb_10000pts.ply. \n Input 1:None; Input 2:./outputs/a_blue_motorcycle.obj; \n Step 1: { Constraints: [output_format: obj, features: Lightweight model]; Output Path: ./outputs/a_blue_motorcycle.obj }; Step 2: { number_of_points: 10000; Output Path: ./outputs/example_mesh_from_glb_10000pts.ply } ",
"code": "self.executor.text_to_3d_generation(prompt=\"Generate a 3D model of a blue motorcycle\", model=\"triplaneturbo\", output_path=\"./outputs/a_blue_motorcycle.obj\")\nself.executor.post_processing(input_path='./outputs/a_blue_motorcycle.obj', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_glb_10000pts.ply', parameters={\"number_of_points\": 10000})",
"id": 5,
"ambiguous": "Generate an object.\nThen, Convert the file at the output of the previous step.",
"missing_info": "Create a 3D asset. I need a file. It should have specific attributes. Store the generated model.\nThen, Perform mesh-to-pointcloud conversion on the mesh at a file. Set the target number of points to a number and save as somewhere.",
"mistake": "Generate a 3D model of a blue motorcycle. Save the output to a obj file. Focus on lightweight model. The result should be saved in ./outputs/a_blue_motorcycle.obj.\nThen, i need to sample 10000 points uniformly from the 3D model at the output path of the previous step. Output the processed ply file to ./outputs/example_mesh_from_glb_10000pts.ply. Ensure heavy model.",
"mistake_type": "quality_speed"
},
{
"complete": "Generate a 3D model of a beautiful dress made out of garbage bags, on a mannequin, studio lighting. Ensure the result is low quality. I need it generated at fast speed. It should be a lightweight model. Save the output to ./outputs/a_beautiful_dress_made_out_of_garbage_bags.obj.\nThen, take the mesh at the output path of the previous step and generate a point cloud with 25000 samples. Store the output at ./outputs/example_mesh_from_obj_25000pts.ply.",
"file_path": null,
"ground_truth": [
{
"task": "Text to 3D Generation",
"quality": "Low quality",
"speed": "Fast speed",
"usage": "object",
"output_format": "obj",
"output_path": "./outputs/a_beautiful_dress_made_out_of_garbage_bags.obj",
"extra_info": "Lightweight model"
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 25000,
"output_path": "./outputs/example_mesh_from_obj_25000pts.ply"
}
],
"gt_summary": " Text to 3D Generation, Post Processing \n Generate a 3D model of a beautiful dress made out of garbage bags, on a mannequin, studio lighting. Ensure the result is low quality. I need it generated at fast speed. It should be a lightweight model. Save the output to ./outputs/a_beautiful_dress_made_out_of_garbage_bags.obj.\nThen, take the mesh at the output path of the previous step and generate a point cloud with 25000 samples. Store the output at ./outputs/example_mesh_from_obj_25000pts.ply. \n Input 1:None; Input 2:./outputs/a_beautiful_dress_made_out_of_garbage_bags.obj; \n Step 1: { Constraints: [quality: Low quality, speed: Fast speed, features: Lightweight model]; Output Path: ./outputs/a_beautiful_dress_made_out_of_garbage_bags.obj }; Step 2: { number_of_points: 25000; Output Path: ./outputs/example_mesh_from_obj_25000pts.ply } ",
"code": "self.executor.text_to_3d_generation(prompt=\"Generate a 3D model of a beautiful dress made out of garbage bags, on a mannequin, studio lighting\", model=\"triplaneturbo\", output_path=\"./outputs/a_beautiful_dress_made_out_of_garbage_bags.obj\")\nself.executor.post_processing(input_path='./outputs/a_beautiful_dress_made_out_of_garbage_bags.obj', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_obj_25000pts.ply', parameters={\"number_of_points\": 25000})",
"id": 6,
"ambiguous": "Create something.\nThen, Do something with the output of the previous step.",
"missing_info": "Create a 3D asset. The model quality should be specified. I need it generated at a certain speed. Save the output to a file. Focus on certain features. Store the generated model.\nThen, Convert the mesh file at a location into a point cloud by sampling a certain number of points. Save the result to somewhere.",
"mistake": "Generate a 3D model of a beautiful dress made out of garbage bags, on a mannequin, studio lighting. Ensure the result is low quality. I need it generated at fast speed. It should be a lightweight model. Save the output to ./outputs/a_beautiful_dress_made_out_of_garbage_bags.obj.\nThen, take the mesh at the output path of the previous step and generate a point cloud with 25000 samples. Store the output at ./outputs/example_mesh_from_obj_25000pts.ply. Ensure heavy model. Ensure slow speed. Ensure high quality.",
"mistake_type": "quality_speed"
},
{
"complete": "Generate a 3D model of an English castle, aerial view. I need a ply file. Save the output to ./outputs/an_English_castle.ply.\nThen, take the mesh at the output path of the previous step and generate a point cloud with 35000 samples. Store the output at ./outputs/example_mesh_from_glb_35000pts.ply.",
"file_path": null,
"ground_truth": [
{
"task": "Text to 3D Generation",
"quality": null,
"speed": null,
"usage": "scene",
"output_format": "ply",
"output_path": "./outputs/an_English_castle.ply",
"extra_info": null
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 35000,
"output_path": "./outputs/example_mesh_from_glb_35000pts.ply"
}
],
"gt_summary": " Text to 3D Generation, Post Processing \n Generate a 3D model of an English castle, aerial view. I need a ply file. Save the output to ./outputs/an_English_castle.ply.\nThen, take the mesh at the output path of the previous step and generate a point cloud with 35000 samples. Store the output at ./outputs/example_mesh_from_glb_35000pts.ply. \n Input 1:None; Input 2:./outputs/an_English_castle.ply; \n Step 1: { Constraints: [output_format: ply]; Output Path: ./outputs/an_English_castle.ply }; Step 2: { number_of_points: 35000; Output Path: ./outputs/example_mesh_from_glb_35000pts.ply } ",
"code": "self.executor.text_to_3d_generation(prompt=\"Generate a 3D model of an English castle, aerial view\", model=\"worldgen\", output_path=\"./outputs/an_English_castle.ply\")\nself.executor.post_processing(input_path='./outputs/an_English_castle.ply', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_glb_35000pts.ply', parameters={\"number_of_points\": 35000})",
"id": 7,
"ambiguous": "Generate an object.\nThen, Do something with the output of the previous step.",
"missing_info": "Generate something. Export format: 3D model. The result should be saved in a location.\nThen, Perform mesh-to-pointcloud conversion on the mesh at a file. Set the target number of points to a number and save as somewhere.",
"mistake": "Generate a 3D model of an English castle, aerial view. I need a ply file. Save the output to ./outputs/an_English_castle.ply.\nThen, take the mesh at the output path of the previous step and generate a point cloud with 35000 samples. Store the output at ./outputs/example_mesh_from_glb_35000pts.ply. Ensure the output format is \".pdf\"",
"mistake_type": "output_format"
},
{
"complete": "Generate a 3D model of Mont Saint-Michel, France, aerial view, in the style of a DSLR photo. Ensure the result is medium quality. The generation speed should be slow speed. Save the output to a glb file. Save the output to ./outputs/Mont_Saint-Michel.glb.\nThen, perform mesh-to-pointcloud conversion on the mesh at the output path of the previous step. Set the target number of points to 20000 and save as ./outputs/example_mesh_from_glb_20000pts.ply.",
"file_path": null,
"ground_truth": [
{
"task": "Text to 3D Generation",
"quality": "Medium quality",
"speed": "Slow speed",
"usage": "scene",
"output_format": "glb",
"output_path": "./outputs/Mont_Saint-Michel.glb",
"extra_info": null
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 20000,
"output_path": "./outputs/example_mesh_from_glb_20000pts.ply"
}
],
"gt_summary": " Text to 3D Generation, Post Processing \n Generate a 3D model of Mont Saint-Michel, France, aerial view, in the style of a DSLR photo. Ensure the result is medium quality. The generation speed should be slow speed. Save the output to a glb file. Save the output to ./outputs/Mont_Saint-Michel.glb.\nThen, perform mesh-to-pointcloud conversion on the mesh at the output path of the previous step. Set the target number of points to 20000 and save as ./outputs/example_mesh_from_glb_20000pts.ply. \n Input 1:None; Input 2:./outputs/Mont_Saint-Michel.glb; \n Step 1: { Constraints: [quality: Medium quality, speed: Slow speed, output_format: glb]; Output Path: ./outputs/Mont_Saint-Michel.glb }; Step 2: { number_of_points: 20000; Output Path: ./outputs/example_mesh_from_glb_20000pts.ply } ",
"code": "self.executor.text_to_3d_generation(prompt=\"Generate a 3D model of Mont Saint-Michel, France, aerial view, in the style of a DSLR photo\", model=\"worldgen\", output_path=\"./outputs/Mont_Saint-Michel.glb\")\nself.executor.post_processing(input_path='./outputs/Mont_Saint-Michel.glb', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_glb_20000pts.ply', parameters={\"number_of_points\": 20000})",
"id": 8,
"ambiguous": "I want a 3D asset.\nThen, Do something with the output of the previous step.",
"missing_info": "Generate a 3D model. Ensure the result quality is defined. I need it generated at a certain speed. I need a file. Save the output to somewhere.\nThen, Take the mesh at a location and generate a point cloud with a certain number of samples. Store the output at a path.",
"mistake": "Generate a 3D model of Mont Saint-Michel, France, aerial view, in the style of a DSLR photo. Ensure the result is medium quality. The generation speed should be slow speed. Save the output to a glb file. Save the output to ./outputs/Mont_Saint-Michel.glb.\nThen, perform mesh-to-pointcloud conversion on the mesh at the output path of the previous step. Set the target number of points to 20000 and save as ./outputs/example_mesh_from_glb_20000pts.ply. Ensure fast speed. Ensure extremely high quality.",
"mistake_type": "quality_speed"
},
{
"complete": "Generate a 3D model of Mount Fuji, aerial view, in the style of a DSLR photo. I require medium quality output. The generation speed should be slow speed. Export format: ply. Save the output to ./outputs/Mount_Fuji.ply.\nThen, i need to sample 1000 points uniformly from the 3D model at the output path of the previous step. Output the processed ply file to ./outputs/example_mesh_from_obj_1000pts.ply.",
"file_path": null,
"ground_truth": [
{
"task": "Text to 3D Generation",
"quality": "Medium quality",
"speed": "Slow speed",
"usage": "scene",
"output_format": "ply",
"output_path": "./outputs/Mount_Fuji.ply",
"extra_info": null
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 1000,
"output_path": "./outputs/example_mesh_from_obj_1000pts.ply"
}
],
"gt_summary": " Text to 3D Generation, Post Processing \n Generate a 3D model of Mount Fuji, aerial view, in the style of a DSLR photo. I require medium quality output. The generation speed should be slow speed. Export format: ply. Save the output to ./outputs/Mount_Fuji.ply.\nThen, i need to sample 1000 points uniformly from the 3D model at the output path of the previous step. Output the processed ply file to ./outputs/example_mesh_from_obj_1000pts.ply. \n Input 1:None; Input 2:./outputs/Mount_Fuji.ply; \n Step 1: { Constraints: [quality: Medium quality, speed: Slow speed, output_format: ply]; Output Path: ./outputs/Mount_Fuji.ply }; Step 2: { number_of_points: 1000; Output Path: ./outputs/example_mesh_from_obj_1000pts.ply } ",
"code": "self.executor.text_to_3d_generation(prompt=\"Generate a 3D model of Mount Fuji, aerial view, in the style of a DSLR photo\", model=\"worldgen\", output_path=\"./outputs/Mount_Fuji.ply\")\nself.executor.post_processing(input_path='./outputs/Mount_Fuji.ply', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_obj_1000pts.ply', parameters={\"number_of_points\": 1000})",
"id": 9,
"ambiguous": "I want a 3D asset.\nThen, Process the output of the previous step.",
"missing_info": "Generate a 3D model. Ensure the result quality is defined. I need it generated at a certain speed. I need a file. The result should be saved in a location.\nThen, Perform mesh-to-pointcloud conversion on the mesh at a file. Set the target number of points to a number and save as somewhere.",
"mistake": "Generate a 3D model of Mount Fuji, aerial view, in the style of a DSLR photo. I require medium quality output. The generation speed should be slow speed. Export format: ply. Save the output to ./outputs/Mount_Fuji.ply.\nThen, i need to sample 1000 points uniformly from the 3D model at the output path of the previous step. Output the processed ply file to ./outputs/example_mesh_from_obj_1000pts.ply. Ensure fast speed. Ensure extremely high quality.",
"mistake_type": "quality_speed"
},
{
"complete": "Perform single-view 3D reconstruction using {input_file_path} as input. I require high quality output. I need a obj file. Ensure high-fidelity mesh. Save the output to ./outputs/scan4.obj.\nThen, i need to sample 10000 points uniformly from the 3D model at the output path of the previous step. Output the processed ply file to ./outputs/example_mesh_from_glb_10000pts.ply.",
"file_path": "./data/source/singleview/object/scan4.jpg",
"ground_truth": [
{
"task": "Single-View 3D Reconstruction",
"usage": "object",
"quality": "High quality",
"speed": null,
"conf_thres": null,
"output_format": "obj",
"output_path": "./outputs/scan4.obj",
"extra_info": "High-fidelity mesh"
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 10000,
"output_path": "./outputs/example_mesh_from_glb_10000pts.ply"
}
],
"gt_summary": " Single-View 3D Reconstruction, Post Processing \n Perform single-view 3D reconstruction using {input_file_path} as input. I require high quality output. I need a obj file. Ensure high-fidelity mesh. Save the output to ./outputs/scan4.obj.\nThen, i need to sample 10000 points uniformly from the 3D model at the output path of the previous step. Output the processed ply file to ./outputs/example_mesh_from_glb_10000pts.ply. \n Input 1:./data/source/singleview/object/scan4.jpg; Input 2:./outputs/scan4.obj; \n Step 1: { Constraints: [quality: High quality, output_format: obj, features: High-fidelity mesh]; Output Path: ./outputs/scan4.obj }; Step 2: { number_of_points: 10000; Output Path: ./outputs/example_mesh_from_glb_10000pts.ply } ",
"code": "self.executor.single_view_reconstruction(image_path='./data/source/singleview/object/scan4.jpg', output_path='./outputs/scan4.obj', model='trellis')\nself.executor.post_processing(input_path='./outputs/scan4.obj', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_glb_10000pts.ply', parameters={\"number_of_points\": 10000})",
"id": 10,
"ambiguous": "Do something with {input_file_path}.\nThen, Do something with the output of the previous step.",
"missing_info": "Process the single-view image to create a 3D representation. Ensure the result quality is defined. Focus on certain features. Save the output to a file. The result should be saved in a location.\nThen, I need to sample a number of points uniformly from the 3D model at a path. Output the processed ply file to a location.",
"mistake": "Perform single-view 3D reconstruction using ./data/source/singleview/object/scan4.obj as input. I require high quality output. I need a obj file. Ensure high-fidelity mesh. Save the output to ./outputs/scan4.obj.\nThen, i need to sample 10000 points uniformly from the 3D model at the output path of the previous step. Output the processed ply file to ./outputs/example_mesh_from_glb_10000pts.ply.",
"mistake_type": "input_format"
},
{
"complete": "Perform single-view 3D reconstruction using {input_file_path} as input. Expect medium speed processing. Export format: ply. Store the result at ./outputs/scan110.ply.\nThen, convert the mesh file at the output path of the previous step into a point cloud by sampling 10000 points. Save the result to ./outputs/example_mesh_from_obj_10000pts.ply.",
"file_path": "./data/source/singleview/object/scan110.jpg",
"ground_truth": [
{
"task": "Single-View 3D Reconstruction",
"usage": "object",
"quality": null,
"speed": "Medium speed",
"conf_thres": null,
"output_format": "ply",
"output_path": "./outputs/scan110.ply",
"extra_info": null
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 10000,
"output_path": "./outputs/example_mesh_from_obj_10000pts.ply"
}
],
"gt_summary": " Single-View 3D Reconstruction, Post Processing \n Perform single-view 3D reconstruction using {input_file_path} as input. Expect medium speed processing. Export format: ply. Store the result at ./outputs/scan110.ply.\nThen, convert the mesh file at the output path of the previous step into a point cloud by sampling 10000 points. Save the result to ./outputs/example_mesh_from_obj_10000pts.ply. \n Input 1:./data/source/singleview/object/scan110.jpg; Input 2:./outputs/scan110.ply; \n Step 1: { Constraints: [speed: Medium speed, output_format: ply]; Output Path: ./outputs/scan110.ply }; Step 2: { number_of_points: 10000; Output Path: ./outputs/example_mesh_from_obj_10000pts.ply } ",
"code": "self.executor.single_view_reconstruction(image_path='./data/source/singleview/object/scan110.jpg', output_path='./outputs/scan110.ply', model='trellis')\nself.executor.post_processing(input_path='./outputs/scan110.ply', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_obj_10000pts.ply', parameters={\"number_of_points\": 10000})",
"id": 11,
"ambiguous": "Process {input_file_path}.\nThen, Process the output of the previous step.",
"missing_info": "Perform single-view 3D reconstruction. The generation speed should be specified. Save the output to a file. Save the output to somewhere.\nThen, Take the mesh at a location and generate a point cloud with a certain number of samples. Store the output at a path.",
"mistake": "Perform single-view 3D reconstruction using ./data/source/singleview/object/scan110.stl as input. Expect medium speed processing. Export format: ply. Store the result at ./outputs/scan110.ply.\nThen, convert the mesh file at the output path of the previous step into a point cloud by sampling 10000 points. Save the result to ./outputs/example_mesh_from_obj_10000pts.ply.",
"mistake_type": "input_format"
},
{
"complete": "Reconstruct a 3D asset from the image at {input_file_path}. I need a ply file. Focus on high-fidelity mesh. Store the result at ./outputs/scan62.ply.\nThen, take the mesh at the output path of the previous step and generate a point cloud with 30000 samples. Store the output at ./outputs/example_mesh_from_glb_30000pts.ply.",
"file_path": "./data/source/singleview/object/scan62.jpg",
"ground_truth": [
{
"task": "Single-View 3D Reconstruction",
"usage": "object",
"quality": null,
"speed": null,
"conf_thres": null,
"output_format": "ply",
"output_path": "./outputs/scan62.ply",
"extra_info": "High-fidelity mesh"
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 30000,
"output_path": "./outputs/example_mesh_from_glb_30000pts.ply"
}
],
"gt_summary": " Single-View 3D Reconstruction, Post Processing \n Reconstruct a 3D asset from the image at {input_file_path}. I need a ply file. Focus on high-fidelity mesh. Store the result at ./outputs/scan62.ply.\nThen, take the mesh at the output path of the previous step and generate a point cloud with 30000 samples. Store the output at ./outputs/example_mesh_from_glb_30000pts.ply. \n Input 1:./data/source/singleview/object/scan62.jpg; Input 2:./outputs/scan62.ply; \n Step 1: { Constraints: [output_format: ply, features: High-fidelity mesh]; Output Path: ./outputs/scan62.ply }; Step 2: { number_of_points: 30000; Output Path: ./outputs/example_mesh_from_glb_30000pts.ply } ",
"code": "self.executor.single_view_reconstruction(image_path='./data/source/singleview/object/scan62.jpg', output_path='./outputs/scan62.ply', model='trellis')\nself.executor.post_processing(input_path='./outputs/scan62.ply', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_glb_30000pts.ply', parameters={\"number_of_points\": 30000})",
"id": 12,
"ambiguous": "Fix {input_file_path} to 3D.\nThen, Do something with the output of the previous step.",
"missing_info": "Perform single-view 3D reconstruction. It should have specific attributes. I need a file. Store the generated model.\nThen, I need to sample a number of points uniformly from the 3D model at a path. Output the processed ply file to a location.",
"mistake": "Reconstruct a 3D asset from the image at ./data/source/singleview/object/scan62.stl. I need a ply file. Focus on high-fidelity mesh. Store the result at ./outputs/scan62.ply.\nThen, take the mesh at the output path of the previous step and generate a point cloud with 30000 samples. Store the output at ./outputs/example_mesh_from_glb_30000pts.ply.",
"mistake_type": "input_format"
},
{
"complete": "Load image {input_file_path} and reconstruct it into 3D asset. The reconstruction speed should be medium speed. Save the output as a ply file. Save the output to ./outputs/scan10.ply.\nThen, i need to sample 30000 points uniformly from the 3D model at the output path of the previous step. Output the processed ply file to ./outputs/example_mesh_from_glb_30000pts.ply.",
"file_path": "./data/source/singleview/object/scan10.jpg",
"ground_truth": [
{
"task": "Single-View 3D Reconstruction",
"usage": "object",
"quality": null,
"speed": "Medium speed",
"conf_thres": null,
"output_format": "ply",
"output_path": "./outputs/scan10.ply",
"extra_info": null
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 30000,
"output_path": "./outputs/example_mesh_from_glb_30000pts.ply"
}
],
"gt_summary": " Single-View 3D Reconstruction, Post Processing \n Load image {input_file_path} and reconstruct it into 3D asset. The reconstruction speed should be medium speed. Save the output as a ply file. Save the output to ./outputs/scan10.ply.\nThen, i need to sample 30000 points uniformly from the 3D model at the output path of the previous step. Output the processed ply file to ./outputs/example_mesh_from_glb_30000pts.ply. \n Input 1:./data/source/singleview/object/scan10.jpg; Input 2:./outputs/scan10.ply; \n Step 1: { Constraints: [speed: Medium speed, output_format: ply]; Output Path: ./outputs/scan10.ply }; Step 2: { number_of_points: 30000; Output Path: ./outputs/example_mesh_from_glb_30000pts.ply } ",
"code": "self.executor.single_view_reconstruction(image_path='./data/source/singleview/object/scan10.jpg', output_path='./outputs/scan10.ply', model='trellis')\nself.executor.post_processing(input_path='./outputs/scan10.ply', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_glb_30000pts.ply', parameters={\"number_of_points\": 30000})",
"id": 13,
"ambiguous": "Make 3D from {input_file_path}.\nThen, Process the output of the previous step.",
"missing_info": "Perform single-view 3D reconstruction. I need it generated at a certain speed. Save the output to a file. The result should be saved in a location.\nThen, Perform mesh-to-pointcloud conversion on the mesh at a file. Set the target number of points to a number and save as somewhere.",
"mistake": "Load image ./data/source/singleview/object/scan10.stl and reconstruct it into 3D asset. The reconstruction speed should be medium speed. Save the output as a ply file. Save the output to ./outputs/scan10.ply.\nThen, i need to sample 30000 points uniformly from the 3D model at the output path of the previous step. Output the processed ply file to ./outputs/example_mesh_from_glb_30000pts.ply.",
"mistake_type": "input_format"
},
{
"complete": "Process the single-view image located at {input_file_path} to create a 3D representation. The model quality should be high quality. The reconstruction speed should be medium speed. I need a glb file. The result should be saved in ./outputs/scan48.glb.\nThen, i need to sample 15000 points uniformly from the 3D model at the output path of the previous step. Output the processed ply file to ./outputs/example_mesh_from_glb_15000pts.ply.",
"file_path": "./data/source/singleview/object/scan48.jpg",
"ground_truth": [
{
"task": "Single-View 3D Reconstruction",
"usage": "object",
"quality": "High quality",
"speed": "Medium speed",
"conf_thres": null,
"output_format": "glb",
"output_path": "./outputs/scan48.glb",
"extra_info": null
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 15000,
"output_path": "./outputs/example_mesh_from_glb_15000pts.ply"
}
],
"gt_summary": " Single-View 3D Reconstruction, Post Processing \n Process the single-view image located at {input_file_path} to create a 3D representation. The model quality should be high quality. The reconstruction speed should be medium speed. I need a glb file. The result should be saved in ./outputs/scan48.glb.\nThen, i need to sample 15000 points uniformly from the 3D model at the output path of the previous step. Output the processed ply file to ./outputs/example_mesh_from_glb_15000pts.ply. \n Input 1:./data/source/singleview/object/scan48.jpg; Input 2:./outputs/scan48.glb; \n Step 1: { Constraints: [quality: High quality, speed: Medium speed, output_format: glb]; Output Path: ./outputs/scan48.glb }; Step 2: { number_of_points: 15000; Output Path: ./outputs/example_mesh_from_glb_15000pts.ply } ",
"code": "self.executor.single_view_reconstruction(image_path='./data/source/singleview/object/scan48.jpg', output_path='./outputs/scan48.glb', model='trellis')\nself.executor.post_processing(input_path='./outputs/scan48.glb', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_glb_15000pts.ply', parameters={\"number_of_points\": 15000})",
"id": 14,
"ambiguous": "Do something with {input_file_path}.\nThen, Convert the file at the output of the previous step.",
"missing_info": "Reconstruct a 3D asset from the image. Ensure the result quality is defined. Expect appropriate processing speed. Save the output to a file. The result should be saved in a location.\nThen, I need to sample a number of points uniformly from the 3D model at a path. Output the processed ply file to a location.",
"mistake": "Process the single-view image located at ./data/source/singleview/object/scan48.ply to create a 3D representation. The model quality should be high quality. The reconstruction speed should be medium speed. I need a glb file. The result should be saved in ./outputs/scan48.glb.\nThen, i need to sample 15000 points uniformly from the 3D model at the output path of the previous step. Output the processed ply file to ./outputs/example_mesh_from_glb_15000pts.ply.",
"mistake_type": "input_format"
},
{
"complete": "Process the single-view image located at {input_file_path} to create a 3D representation. I need it processed with slow speed. Export format: obj. Store the result at ./outputs/scan118.obj.\nThen, take the mesh at the output path of the previous step and generate a point cloud with 20000 samples. Store the output at ./outputs/example_mesh_from_obj_20000pts.ply.",
"file_path": "./data/source/singleview/object/scan118.jpg",
"ground_truth": [
{
"task": "Single-View 3D Reconstruction",
"usage": "object",
"quality": null,
"speed": "Slow speed",
"conf_thres": null,
"output_format": "obj",
"output_path": "./outputs/scan118.obj",
"extra_info": null
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 20000,
"output_path": "./outputs/example_mesh_from_obj_20000pts.ply"
}
],
"gt_summary": " Single-View 3D Reconstruction, Post Processing \n Process the single-view image located at {input_file_path} to create a 3D representation. I need it processed with slow speed. Export format: obj. Store the result at ./outputs/scan118.obj.\nThen, take the mesh at the output path of the previous step and generate a point cloud with 20000 samples. Store the output at ./outputs/example_mesh_from_obj_20000pts.ply. \n Input 1:./data/source/singleview/object/scan118.jpg; Input 2:./outputs/scan118.obj; \n Step 1: { Constraints: [speed: Slow speed, output_format: obj]; Output Path: ./outputs/scan118.obj }; Step 2: { number_of_points: 20000; Output Path: ./outputs/example_mesh_from_obj_20000pts.ply } ",
"code": "self.executor.single_view_reconstruction(image_path='./data/source/singleview/object/scan118.jpg', output_path='./outputs/scan118.obj', model='hunyuan')\nself.executor.post_processing(input_path='./outputs/scan118.obj', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_obj_20000pts.ply', parameters={\"number_of_points\": 20000})",
"id": 15,
"ambiguous": "Process {input_file_path}.\nThen, Convert the file at the output of the previous step.",
"missing_info": "Process the single-view image to create a 3D representation. I need it generated at a certain speed. Export format: 3D model. Store the generated model.\nThen, Please process the mesh at a file to create a point cloud. Use a sampling rate of a number of points and write the file to somewhere.",
"mistake": "Process the single-view image located at ./data/source/singleview/object/scan118.stl to create a 3D representation. I need it processed with slow speed. Export format: obj. Store the result at ./outputs/scan118.obj.\nThen, take the mesh at the output path of the previous step and generate a point cloud with 20000 samples. Store the output at ./outputs/example_mesh_from_obj_20000pts.ply.",
"mistake_type": "input_format"
},
{
"complete": "Perform single-view 3D reconstruction using {input_file_path} as input. I require ultra-high quality output. I need it processed with slow speed. Ensure mesh with detailed geometry. Save the output to ./outputs/scan13.glb.\nThen, take the mesh at the output path of the previous step and generate a point cloud with 15000 samples. Store the output at ./outputs/example_mesh_from_obj_15000pts.ply.",
"file_path": "./data/source/singleview/object/scan13.jpg",
"ground_truth": [
{
"task": "Single-View 3D Reconstruction",
"usage": "object",
"quality": "Ultra-High quality",
"speed": "Slow speed",
"conf_thres": null,
"output_format": null,
"output_path": "./outputs/scan13.glb",
"extra_info": "Mesh with detailed geometry"
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 15000,
"output_path": "./outputs/example_mesh_from_obj_15000pts.ply"
}
],
"gt_summary": " Single-View 3D Reconstruction, Post Processing \n Perform single-view 3D reconstruction using {input_file_path} as input. I require ultra-high quality output. I need it processed with slow speed. Ensure mesh with detailed geometry. Save the output to ./outputs/scan13.glb.\nThen, take the mesh at the output path of the previous step and generate a point cloud with 15000 samples. Store the output at ./outputs/example_mesh_from_obj_15000pts.ply. \n Input 1:./data/source/singleview/object/scan13.jpg; Input 2:./outputs/scan13.glb; \n Step 1: { Constraints: [quality: Ultra-High quality, speed: Slow speed, features: Mesh with detailed geometry]; Output Path: ./outputs/scan13.glb }; Step 2: { number_of_points: 15000; Output Path: ./outputs/example_mesh_from_obj_15000pts.ply } ",
"code": "self.executor.single_view_reconstruction(image_path='./data/source/singleview/object/scan13.jpg', output_path='./outputs/scan13.glb', model='hunyuan')\nself.executor.post_processing(input_path='./outputs/scan13.glb', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_obj_15000pts.ply', parameters={\"number_of_points\": 15000})",
"id": 16,
"ambiguous": "Fix {input_file_path} to 3D.\nThen, Process the output of the previous step.",
"missing_info": "Process the single-view image to create a 3D representation. Ensure the result quality is defined. I need it generated at a certain speed. Focus on certain features. Save the output to somewhere.\nThen, Convert the mesh file at a location into a point cloud by sampling a certain number of points. Save the result to somewhere.",
"mistake": "Perform single-view 3D reconstruction using ./data/source/singleview/object/scan13.ply as input. I require ultra-high quality output. I need it processed with slow speed. Ensure mesh with detailed geometry. Save the output to ./outputs/scan13.glb.\nThen, take the mesh at the output path of the previous step and generate a point cloud with 15000 samples. Store the output at ./outputs/example_mesh_from_obj_15000pts.ply.",
"mistake_type": "input_format"
},
{
"complete": "Process the single-view image located at {input_file_path} to create a 3D representation. Save the output as a obj file. It should be a mesh with detailed geometry. Save the output to ./outputs/scan11.obj.\nThen, convert the mesh file at the output path of the previous step into a point cloud by sampling 1000 points. Save the result to ./outputs/example_mesh_from_ply_1000pts.ply.",
"file_path": "./data/source/singleview/object/scan11.jpg",
"ground_truth": [
{
"task": "Single-View 3D Reconstruction",
"usage": "object",
"quality": null,
"speed": null,
"conf_thres": null,
"output_format": "obj",
"output_path": "./outputs/scan11.obj",
"extra_info": "Mesh with detailed geometry"
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 1000,
"output_path": "./outputs/example_mesh_from_ply_1000pts.ply"
}
],
"gt_summary": " Single-View 3D Reconstruction, Post Processing \n Process the single-view image located at {input_file_path} to create a 3D representation. Save the output as a obj file. It should be a mesh with detailed geometry. Save the output to ./outputs/scan11.obj.\nThen, convert the mesh file at the output path of the previous step into a point cloud by sampling 1000 points. Save the result to ./outputs/example_mesh_from_ply_1000pts.ply. \n Input 1:./data/source/singleview/object/scan11.jpg; Input 2:./outputs/scan11.obj; \n Step 1: { Constraints: [output_format: obj, features: Mesh with detailed geometry]; Output Path: ./outputs/scan11.obj }; Step 2: { number_of_points: 1000; Output Path: ./outputs/example_mesh_from_ply_1000pts.ply } ",
"code": "self.executor.single_view_reconstruction(image_path='./data/source/singleview/object/scan11.jpg', output_path='./outputs/scan11.obj', model='hunyuan')\nself.executor.post_processing(input_path='./outputs/scan11.obj', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_ply_1000pts.ply', parameters={\"number_of_points\": 1000})",
"id": 17,
"ambiguous": "Fix {input_file_path} to 3D.\nThen, Convert the file at the output of the previous step.",
"missing_info": "Perform single-view 3D reconstruction. It should have specific attributes. Export format: 3D model. Save the output to somewhere.\nThen, I need to sample a number of points uniformly from the 3D model at a path. Output the processed ply file to a location.",
"mistake": "Process the single-view image located at ./data/source/singleview/object/scan11.stl to create a 3D representation. Save the output as a obj file. It should be a mesh with detailed geometry. Save the output to ./outputs/scan11.obj.\nThen, convert the mesh file at the output path of the previous step into a point cloud by sampling 1000 points. Save the result to ./outputs/example_mesh_from_ply_1000pts.ply.",
"mistake_type": "input_format"
},
{
"complete": "Process the single-view image located at {input_file_path} to create a 3D representation. Ensure the mesh is ultra-high quality. It should be a mesh with detailed geometry. The result should be saved in ./outputs/scan75.glb.\nThen, convert the mesh file at the output path of the previous step into a point cloud by sampling 25000 points. Save the result to ./outputs/example_mesh_from_glb_25000pts.ply.",
"file_path": "./data/source/singleview/object/scan75.jpg",
"ground_truth": [
{
"task": "Single-View 3D Reconstruction",
"usage": "object",
"quality": "Ultra-High quality",
"speed": null,
"conf_thres": null,
"output_format": null,
"output_path": "./outputs/scan75.glb",
"extra_info": "Mesh with detailed geometry"
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 25000,
"output_path": "./outputs/example_mesh_from_glb_25000pts.ply"
}
],
"gt_summary": " Single-View 3D Reconstruction, Post Processing \n Process the single-view image located at {input_file_path} to create a 3D representation. Ensure the mesh is ultra-high quality. It should be a mesh with detailed geometry. The result should be saved in ./outputs/scan75.glb.\nThen, convert the mesh file at the output path of the previous step into a point cloud by sampling 25000 points. Save the result to ./outputs/example_mesh_from_glb_25000pts.ply. \n Input 1:./data/source/singleview/object/scan75.jpg; Input 2:./outputs/scan75.glb; \n Step 1: { Constraints: [quality: Ultra-High quality, features: Mesh with detailed geometry]; Output Path: ./outputs/scan75.glb }; Step 2: { number_of_points: 25000; Output Path: ./outputs/example_mesh_from_glb_25000pts.ply } ",
"code": "self.executor.single_view_reconstruction(image_path='./data/source/singleview/object/scan75.jpg', output_path='./outputs/scan75.glb', model='hunyuan')\nself.executor.post_processing(input_path='./outputs/scan75.glb', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_glb_25000pts.ply', parameters={\"number_of_points\": 25000})",
"id": 18,
"ambiguous": "Process {input_file_path}.\nThen, Convert the file at the output of the previous step.",
"missing_info": "Perform single-view 3D reconstruction. The model quality should be specified. Consider that this has particular characteristics. Store the generated model.\nThen, Take the mesh at a location and generate a point cloud with a certain number of samples. Store the output at a path.",
"mistake": "Process the single-view image located at ./data/source/singleview/object/scan75.ply to create a 3D representation. Ensure the mesh is ultra-high quality. It should be a mesh with detailed geometry. The result should be saved in ./outputs/scan75.glb.\nThen, convert the mesh file at the output path of the previous step into a point cloud by sampling 25000 points. Save the result to ./outputs/example_mesh_from_glb_25000pts.ply.",
"mistake_type": "input_format"
},
{
"complete": "Load image {input_file_path} and reconstruct it into 3D asset. It should be a mesh with detailed geometry. Save the output to ./outputs/scan118.glb.\nThen, please process the mesh at the output path of the previous step to create a point cloud. Use a sampling rate of 5000 points and write the file to ./outputs/example_mesh_from_obj_5000pts.ply.",
"file_path": "./data/source/singleview/object/scan118.jpg",
"ground_truth": [
{
"task": "Single-View 3D Reconstruction",
"usage": "object",
"quality": null,
"speed": null,
"conf_thres": null,
"output_format": null,
"output_path": "./outputs/scan118.glb",
"extra_info": "Mesh with detailed geometry"
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 5000,
"output_path": "./outputs/example_mesh_from_obj_5000pts.ply"
}
],
"gt_summary": " Single-View 3D Reconstruction, Post Processing \n Load image {input_file_path} and reconstruct it into 3D asset. It should be a mesh with detailed geometry. Save the output to ./outputs/scan118.glb.\nThen, please process the mesh at the output path of the previous step to create a point cloud. Use a sampling rate of 5000 points and write the file to ./outputs/example_mesh_from_obj_5000pts.ply. \n Input 1:./data/source/singleview/object/scan118.jpg; Input 2:./outputs/scan118.glb; \n Step 1: { Constraints: [features: Mesh with detailed geometry]; Output Path: ./outputs/scan118.glb }; Step 2: { number_of_points: 5000; Output Path: ./outputs/example_mesh_from_obj_5000pts.ply } ",
"code": "self.executor.single_view_reconstruction(image_path='./data/source/singleview/object/scan118.jpg', output_path='./outputs/scan118.glb', model='hunyuan')\nself.executor.post_processing(input_path='./outputs/scan118.glb', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_obj_5000pts.ply', parameters={\"number_of_points\": 5000})",
"id": 19,
"ambiguous": "Process {input_file_path}.\nThen, Convert the file at the output of the previous step.",
"missing_info": "Load image and reconstruct it into 3D asset. Focus on certain features. Store the generated model.\nThen, I need to sample a number of points uniformly from the 3D model at a path. Output the processed ply file to a location.",
"mistake": "Load image ./data/source/singleview/object/scan118.obj and reconstruct it into 3D asset. It should be a mesh with detailed geometry. Save the output to ./outputs/scan118.glb.\nThen, please process the mesh at the output path of the previous step to create a point cloud. Use a sampling rate of 5000 points and write the file to ./outputs/example_mesh_from_obj_5000pts.ply.",
"mistake_type": "input_format"
},
{
"complete": "Process the single-view image located at {input_file_path} to create a 3D representation. The reconstruction speed should be fast speed. Set the confidence threshold to 19. The result should be saved in ./outputs/counter.ply.\nThen, clean the point cloud file the output path of the previous step. I want to remove outliers with nb_neighbors set to 45 and standard deviation ratio set to 1.6. Output result to ./outputs/clean_nb45_std1.6.ply.",
"file_path": "./data/source/singleview/scene/counter.JPG",
"ground_truth": [
{
"task": "Single-View 3D Reconstruction",
"usage": "scene",
"quality": null,
"speed": "Fast speed",
"conf_thres": 19,
"output_format": null,
"output_path": "./outputs/counter.ply",
"extra_info": null
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 45,
"std_ratio": 1.6,
"output_path": "./outputs/clean_nb45_std1.6.ply"
}
],
"gt_summary": " Single-View 3D Reconstruction, Post Processing \n Process the single-view image located at {input_file_path} to create a 3D representation. The reconstruction speed should be fast speed. Set the confidence threshold to 19. The result should be saved in ./outputs/counter.ply.\nThen, clean the point cloud file the output path of the previous step. I want to remove outliers with nb_neighbors set to 45 and standard deviation ratio set to 1.6. Output result to ./outputs/clean_nb45_std1.6.ply. \n Input 1:./data/source/singleview/scene/counter.JPG; Input 2:./outputs/counter.ply; \n Step 1: { Constraints: [speed: Fast speed, conf_thres: 19]; Output Path: ./outputs/counter.ply }; Step 2: { nb_neighbors: 45, std_ratio: 1.6, Output Path: ./outputs/clean_nb45_std1.6.ply } ",
"code": "self.executor.single_view_reconstruction(image_path='./data/source/singleview/scene/counter.JPG', output_path='./outputs/counter.ply', model='vggt', conf_thres=19)\nself.executor.post_processing(input_path='./outputs/counter.ply', process_type='outliers_removal', output_path='./outputs/clean_nb45_std1.6.ply', parameters={\"nb_neighbors\": 45, \"std_ratio\": 1.6})",
"id": 20,
"ambiguous": "Fix {input_file_path} to 3D.\nThen, Do something with the output of the previous step.",
"missing_info": "Process the single-view image to create a 3D representation. Expect appropriate processing speed. Use a confidence cutoff of a value. The result should be saved in a location.\nThen, Perform post-processing on a file to remove noise. Use a neighbor count of a number and a std_ratio of a value, then save to somewhere.",
"mistake": "Process the single-view image located at {input_file_path} to create a 3D representation. The reconstruction speed should be fast speed. Set the confidence threshold to 1900. The result should be saved in ./outputs/counter.ply.\nThen, clean the point cloud file the output path of the previous step. I want to remove outliers with nb_neighbors set to 45 and standard deviation ratio set to 1.6. Output result to ./outputs/clean_nb45_std1.6.ply.",
"mistake_type": "parameter"
},
{
"complete": "Process the single-view image located at {input_file_path} to create a 3D representation. I need it processed with fast speed. Export format: glb. The result should be saved in ./outputs/room.glb.\nThen, help me filter the noise in the output path of the previous step. Use statistical removal with 35 neighbors and 1.3 std_ratio. The output should be saved at ./outputs/clean_nb35_std1.3.ply.",
"file_path": "./data/source/singleview/scene/room.JPG",
"ground_truth": [
{
"task": "Single-View 3D Reconstruction",
"usage": "scene",
"quality": null,
"speed": "Fast speed",
"conf_thres": null,
"output_format": "glb",
"output_path": "./outputs/room.glb",
"extra_info": null
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 35,
"std_ratio": 1.3,
"output_path": "./outputs/clean_nb35_std1.3.ply"
}
],
"gt_summary": " Single-View 3D Reconstruction, Post Processing \n Process the single-view image located at {input_file_path} to create a 3D representation. I need it processed with fast speed. Export format: glb. The result should be saved in ./outputs/room.glb.\nThen, help me filter the noise in the output path of the previous step. Use statistical removal with 35 neighbors and 1.3 std_ratio. The output should be saved at ./outputs/clean_nb35_std1.3.ply. \n Input 1:./data/source/singleview/scene/room.JPG; Input 2:./outputs/room.glb; \n Step 1: { Constraints: [speed: Fast speed, output_format: glb]; Output Path: ./outputs/room.glb }; Step 2: { nb_neighbors: 35, std_ratio: 1.3, Output Path: ./outputs/clean_nb35_std1.3.ply } ",
"code": "self.executor.single_view_reconstruction(image_path='./data/source/singleview/scene/room.JPG', output_path='./outputs/room.glb', model='vggt', conf_thres=20.0)\nself.executor.post_processing(input_path='./outputs/room.glb', process_type='outliers_removal', output_path='./outputs/clean_nb35_std1.3.ply', parameters={\"nb_neighbors\": 35, \"std_ratio\": 1.3})",
"id": 21,
"ambiguous": "Make 3D from {input_file_path}.\nThen, Clean the point cloud at the output of the previous step.",
"missing_info": "Load image and reconstruct it into 3D asset. Expect appropriate processing speed. Export format: 3D model. Store the generated model.\nThen, Clean the point cloud file at a path. I want to remove outliers with nb_neighbors set to a number and standard deviation ratio set to a value. Output result to a location.",
"mistake": "Process the single-view image located at {input_file_path} to create a 3D representation. I need it processed with fast speed. Export format: glb. The result should be saved in ./outputs/room.glb.\nThen, help me filter the noise in the output path of the previous step. Use statistical removal with 35 neighbors and 1.3 std_ratio. The output should be saved at ./outputs/clean_nb35_std1.3.ply. Ensure slow speed.",
"mistake_type": "quality_speed"
},
{
"complete": "Reconstruct a 3D asset from the image at {input_file_path}. I require medium quality output. Save the output as a glb file. Store the result at ./outputs/Courtroom.glb.\nThen, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 20 neighbors and a std_ratio of 1.7. Save the clean cloud to ./outputs/clean_nb20_std1.7.ply.",
"file_path": "./data/source/singleview/scene/Courtroom.jpg",
"ground_truth": [
{
"task": "Single-View 3D Reconstruction",
"usage": "scene",
"quality": "Medium quality",
"speed": null,
"conf_thres": null,
"output_format": "glb",
"output_path": "./outputs/Courtroom.glb",
"extra_info": null
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 20,
"std_ratio": 1.7,
"output_path": "./outputs/clean_nb20_std1.7.ply"
}
],
"gt_summary": " Single-View 3D Reconstruction, Post Processing \n Reconstruct a 3D asset from the image at {input_file_path}. I require medium quality output. Save the output as a glb file. Store the result at ./outputs/Courtroom.glb.\nThen, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 20 neighbors and a std_ratio of 1.7. Save the clean cloud to ./outputs/clean_nb20_std1.7.ply. \n Input 1:./data/source/singleview/scene/Courtroom.jpg; Input 2:./outputs/Courtroom.glb; \n Step 1: { Constraints: [quality: Medium quality, output_format: glb]; Output Path: ./outputs/Courtroom.glb }; Step 2: { nb_neighbors: 20, std_ratio: 1.7, Output Path: ./outputs/clean_nb20_std1.7.ply } ",
"code": "self.executor.single_view_reconstruction(image_path='./data/source/singleview/scene/Courtroom.jpg', output_path='./outputs/Courtroom.glb', model='vggt', conf_thres=20.0)\nself.executor.post_processing(input_path='./outputs/Courtroom.glb', process_type='outliers_removal', output_path='./outputs/clean_nb20_std1.7.ply', parameters={\"nb_neighbors\": 20, \"std_ratio\": 1.7})",
"id": 22,
"ambiguous": "Do something with {input_file_path}.\nThen, Fix the noise in the output of the previous step.",
"missing_info": "Process the single-view image to create a 3D representation. Ensure the result quality is defined. Export format: 3D model. Save the output to somewhere.\nThen, I have a noisy point cloud at a location. Please apply statistical outlier removal using a certain number of neighbors and a std_ratio of a value. Save the clean cloud to somewhere.",
"mistake": "Reconstruct a 3D asset from the image at {input_file_path}. I require medium quality output. Save the output as a glb file. Store the result at ./outputs/Courtroom.glb.\nThen, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 20 neighbors and a std_ratio of 1.7. Save the clean cloud to ./outputs/clean_nb20_std1.7.ply. Ensure extremely high quality.",
"mistake_type": "quality_speed"
},
{
"complete": "Perform single-view 3D reconstruction using {input_file_path} as input. I need it processed with fast speed. Export format: ply. The result should be saved in ./outputs/scan75.ply.\nThen, clean the point cloud file the output path of the previous step. I want to remove outliers with nb_neighbors set to 25 and standard deviation ratio set to 2.5. Output result to ./outputs/clean_nb25_std2.5.ply.",
"file_path": "./data/source/singleview/object/scan75.jpg",
"ground_truth": [
{
"task": "Single-View 3D Reconstruction",
"usage": "object",
"quality": null,
"speed": "Fast speed",
"conf_thres": null,
"output_format": "ply",
"output_path": "./outputs/scan75.ply",
"extra_info": null
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 25,
"std_ratio": 2.5,
"output_path": "./outputs/clean_nb25_std2.5.ply"
}
],
"gt_summary": " Single-View 3D Reconstruction, Post Processing \n Perform single-view 3D reconstruction using {input_file_path} as input. I need it processed with fast speed. Export format: ply. The result should be saved in ./outputs/scan75.ply.\nThen, clean the point cloud file the output path of the previous step. I want to remove outliers with nb_neighbors set to 25 and standard deviation ratio set to 2.5. Output result to ./outputs/clean_nb25_std2.5.ply. \n Input 1:./data/source/singleview/object/scan75.jpg; Input 2:./outputs/scan75.ply; \n Step 1: { Constraints: [speed: Fast speed, output_format: ply]; Output Path: ./outputs/scan75.ply }; Step 2: { nb_neighbors: 25, std_ratio: 2.5, Output Path: ./outputs/clean_nb25_std2.5.ply } ",
"code": "self.executor.single_view_reconstruction(image_path='./data/source/singleview/object/scan75.jpg', output_path='./outputs/scan75.ply', model='vggt', conf_thres=20.0)\nself.executor.post_processing(input_path='./outputs/scan75.ply', process_type='outliers_removal', output_path='./outputs/clean_nb25_std2.5.ply', parameters={\"nb_neighbors\": 25, \"std_ratio\": 2.5})",
"id": 23,
"ambiguous": "Fix {input_file_path} to 3D.\nThen, Clean the point cloud at the output of the previous step.",
"missing_info": "Reconstruct a 3D asset from the image. The generation speed should be specified. I need a file. Save the output to somewhere.\nThen, Help me filter the noise in a location. Use statistical removal with certain neighbors and a std_ratio. The output should be saved at a path.",
"mistake": "Perform single-view 3D reconstruction using ./data/source/singleview/object/scan75.obj as input. I need it processed with fast speed. Export format: ply. The result should be saved in ./outputs/scan75.ply.\nThen, clean the point cloud file the output path of the previous step. I want to remove outliers with nb_neighbors set to 25 and standard deviation ratio set to 2.5. Output result to ./outputs/clean_nb25_std2.5.ply.",
"mistake_type": "input_format"
},
{
"complete": "Perform single-view 3D reconstruction using {input_file_path} as input. The model quality should be medium quality. I need it processed with fast speed. Save the output to ./outputs/M60.ply.\nThen, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 20 and a std_ratio of 2.2, then save to ./outputs/clean_nb20_std2.2.ply.",
"file_path": "./data/source/singleview/scene/M60.jpg",
"ground_truth": [
{
"task": "Single-View 3D Reconstruction",
"usage": "scene",
"quality": "Medium quality",
"speed": "Fast speed",
"conf_thres": null,
"output_format": null,
"output_path": "./outputs/M60.ply",
"extra_info": null
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 20,
"std_ratio": 2.2,
"output_path": "./outputs/clean_nb20_std2.2.ply"
}
],
"gt_summary": " Single-View 3D Reconstruction, Post Processing \n Perform single-view 3D reconstruction using {input_file_path} as input. The model quality should be medium quality. I need it processed with fast speed. Save the output to ./outputs/M60.ply.\nThen, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 20 and a std_ratio of 2.2, then save to ./outputs/clean_nb20_std2.2.ply. \n Input 1:./data/source/singleview/scene/M60.jpg; Input 2:./outputs/M60.ply; \n Step 1: { Constraints: [quality: Medium quality, speed: Fast speed]; Output Path: ./outputs/M60.ply }; Step 2: { nb_neighbors: 20, std_ratio: 2.2, Output Path: ./outputs/clean_nb20_std2.2.ply } ",
"code": "self.executor.single_view_reconstruction(image_path='./data/source/singleview/scene/M60.jpg', output_path='./outputs/M60.ply', model='vggt', conf_thres=20.0)\nself.executor.post_processing(input_path='./outputs/M60.ply', process_type='outliers_removal', output_path='./outputs/clean_nb20_std2.2.ply', parameters={\"nb_neighbors\": 20, \"std_ratio\": 2.2})",
"id": 24,
"ambiguous": "Make 3D from {input_file_path}.\nThen, Clean the point cloud at the output of the previous step.",
"missing_info": "Reconstruct a 3D asset from the image. I require a certain quality. I need it generated at a certain speed. The result should be saved in a location.\nThen, Help me filter the noise in a location. Use statistical removal with certain neighbors and a std_ratio. The output should be saved at a path.",
"mistake": "Perform single-view 3D reconstruction using {input_file_path} as input. The model quality should be medium quality. I need it processed with fast speed. Save the output to ./outputs/M60.ply.\nThen, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 20 and a std_ratio of 2.2, then save to ./outputs/clean_nb20_std2.2.ply. Ensure slow speed. Ensure extremely high quality.",
"mistake_type": "quality_speed"
},
{
"complete": "Process the single-view image located at {input_file_path} to create a 3D representation. Ensure dense point cloud. The result should be saved in ./outputs/scan110.glb.\nThen, help me filter the noise in the output path of the previous step. Use statistical removal with 35 neighbors and 1.7 std_ratio. The output should be saved at ./outputs/clean_nb35_std1.7.ply.",
"file_path": "./data/source/singleview/object/scan110.jpg",
"ground_truth": [
{
"task": "Single-View 3D Reconstruction",
"usage": "object",
"quality": null,
"speed": null,
"conf_thres": null,
"output_format": null,
"output_path": "./outputs/scan110.glb",
"extra_info": "Dense point cloud"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 35,
"std_ratio": 1.7,
"output_path": "./outputs/clean_nb35_std1.7.ply"
}
],
"gt_summary": " Single-View 3D Reconstruction, Post Processing \n Process the single-view image located at {input_file_path} to create a 3D representation. Ensure dense point cloud. The result should be saved in ./outputs/scan110.glb.\nThen, help me filter the noise in the output path of the previous step. Use statistical removal with 35 neighbors and 1.7 std_ratio. The output should be saved at ./outputs/clean_nb35_std1.7.ply. \n Input 1:./data/source/singleview/object/scan110.jpg; Input 2:./outputs/scan110.glb; \n Step 1: { Constraints: [features: Dense point cloud]; Output Path: ./outputs/scan110.glb }; Step 2: { nb_neighbors: 35, std_ratio: 1.7, Output Path: ./outputs/clean_nb35_std1.7.ply } ",
"code": "self.executor.single_view_reconstruction(image_path='./data/source/singleview/object/scan110.jpg', output_path='./outputs/scan110.glb', model='vggt', conf_thres=20.0)\nself.executor.post_processing(input_path='./outputs/scan110.glb', process_type='outliers_removal', output_path='./outputs/clean_nb35_std1.7.ply', parameters={\"nb_neighbors\": 35, \"std_ratio\": 1.7})",
"id": 25,
"ambiguous": "Process {input_file_path}.\nThen, Do something with the output of the previous step.",
"missing_info": "Perform single-view 3D reconstruction. Consider that this has particular characteristics. Save the output to somewhere.\nThen, I have a noisy point cloud at a location. Please apply statistical outlier removal using a certain number of neighbors and a std_ratio of a value. Save the clean cloud to somewhere.",
"mistake": "Process the single-view image located at ./data/source/singleview/object/scan110.glb to create a 3D representation. Ensure dense point cloud. The result should be saved in ./outputs/scan110.glb.\nThen, help me filter the noise in the output path of the previous step. Use statistical removal with 35 neighbors and 1.7 std_ratio. The output should be saved at ./outputs/clean_nb35_std1.7.ply.",
"mistake_type": "input_format"
},
{
"complete": "Perform single-view 3D reconstruction using {input_file_path} as input. Save the output as a glb file. Focus on dense point cloud. Filter out points with confidence below 22. Save the output to ./outputs/kitchen.glb.\nThen, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 50 neighbors and a std_ratio of 1.8. Save the clean cloud to ./outputs/clean_nb50_std1.8.ply.",
"file_path": "./data/source/singleview/scene/kitchen.JPG",
"ground_truth": [
{
"task": "Single-View 3D Reconstruction",
"usage": "scene",
"quality": null,
"speed": null,
"conf_thres": 22,
"output_format": "glb",
"output_path": "./outputs/kitchen.glb",
"extra_info": "Dense point cloud"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 50,
"std_ratio": 1.8,
"output_path": "./outputs/clean_nb50_std1.8.ply"
}
],
"gt_summary": " Single-View 3D Reconstruction, Post Processing \n Perform single-view 3D reconstruction using {input_file_path} as input. Save the output as a glb file. Focus on dense point cloud. Filter out points with confidence below 22. Save the output to ./outputs/kitchen.glb.\nThen, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 50 neighbors and a std_ratio of 1.8. Save the clean cloud to ./outputs/clean_nb50_std1.8.ply. \n Input 1:./data/source/singleview/scene/kitchen.JPG; Input 2:./outputs/kitchen.glb; \n Step 1: { Constraints: [output_format: glb, features: Dense point cloud, conf_thres: 22]; Output Path: ./outputs/kitchen.glb }; Step 2: { nb_neighbors: 50, std_ratio: 1.8, Output Path: ./outputs/clean_nb50_std1.8.ply } ",
"code": "self.executor.single_view_reconstruction(image_path='./data/source/singleview/scene/kitchen.JPG', output_path='./outputs/kitchen.glb', model='vggt', conf_thres=22)\nself.executor.post_processing(input_path='./outputs/kitchen.glb', process_type='outliers_removal', output_path='./outputs/clean_nb50_std1.8.ply', parameters={\"nb_neighbors\": 50, \"std_ratio\": 1.8})",
"id": 26,
"ambiguous": "Process {input_file_path}.\nThen, Do something with the output of the previous step.",
"missing_info": "Perform single-view 3D reconstruction. Use a confidence cutoff of a value. It should have specific attributes. Save the output to a file. Store the generated model.\nThen, Help me filter the noise in a location. Use statistical removal with certain neighbors and a std_ratio. The output should be saved at a path.",
"mistake": "Perform single-view 3D reconstruction using {input_file_path} as input. Save the output as a glb file. Focus on dense point cloud. Filter out points with confidence below -22. Save the output to ./outputs/kitchen.glb.\nThen, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 50 neighbors and a std_ratio of 1.8. Save the clean cloud to ./outputs/clean_nb50_std1.8.ply.",
"mistake_type": "parameter"
},
{
"complete": "Reconstruct a 3D asset from the image at {input_file_path}. Ensure the mesh is medium quality. Export format: glb. The result should be saved in ./outputs/scan118.glb.\nThen, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 20 neighbors and a std_ratio of 0.6. Save the clean cloud to ./outputs/clean_nb20_std0.6.ply.",
"file_path": "./data/source/singleview/object/scan118.jpg",
"ground_truth": [
{
"task": "Single-View 3D Reconstruction",
"usage": "object",
"quality": "Medium quality",
"speed": null,
"conf_thres": null,
"output_format": "glb",
"output_path": "./outputs/scan118.glb",
"extra_info": null
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 20,
"std_ratio": 0.6,
"output_path": "./outputs/clean_nb20_std0.6.ply"
}
],
"gt_summary": " Single-View 3D Reconstruction, Post Processing \n Reconstruct a 3D asset from the image at {input_file_path}. Ensure the mesh is medium quality. Export format: glb. The result should be saved in ./outputs/scan118.glb.\nThen, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 20 neighbors and a std_ratio of 0.6. Save the clean cloud to ./outputs/clean_nb20_std0.6.ply. \n Input 1:./data/source/singleview/object/scan118.jpg; Input 2:./outputs/scan118.glb; \n Step 1: { Constraints: [quality: Medium quality, output_format: glb]; Output Path: ./outputs/scan118.glb }; Step 2: { nb_neighbors: 20, std_ratio: 0.6, Output Path: ./outputs/clean_nb20_std0.6.ply } ",
"code": "self.executor.single_view_reconstruction(image_path='./data/source/singleview/object/scan118.jpg', output_path='./outputs/scan118.glb', model='vggt', conf_thres=20.0)\nself.executor.post_processing(input_path='./outputs/scan118.glb', process_type='outliers_removal', output_path='./outputs/clean_nb20_std0.6.ply', parameters={\"nb_neighbors\": 20, \"std_ratio\": 0.6})",
"id": 27,
"ambiguous": "Make 3D from {input_file_path}.\nThen, Do something with the output of the previous step.",
"missing_info": "Process the single-view image to create a 3D representation. I require a certain quality. Save the output to a file. The result should be saved in a location.\nThen, Perform post-processing on a file to remove noise. Use a neighbor count of a number and a std_ratio of a value, then save to somewhere.",
"mistake": "Reconstruct a 3D asset from the image at {input_file_path}. Ensure the mesh is medium quality. Export format: glb. The result should be saved in ./outputs/scan118.glb.\nThen, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 20 neighbors and a std_ratio of 0.6. Save the clean cloud to ./outputs/clean_nb20_std0.6.ply. Ensure extremely high quality.",
"mistake_type": "quality_speed"
},
{
"complete": "Reconstruct a 3D asset from the image at {input_file_path}. The model quality should be medium quality. Save the output as a ply file. Save the output to ./outputs/scan11.ply.\nThen, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 40 neighbors and a std_ratio of 1.8. Save the clean cloud to ./outputs/clean_nb40_std1.8.ply.",
"file_path": "./data/source/singleview/object/scan11.jpg",
"ground_truth": [
{
"task": "Single-View 3D Reconstruction",
"usage": "object",
"quality": "Medium quality",
"speed": null,
"conf_thres": null,
"output_format": "ply",
"output_path": "./outputs/scan11.ply",
"extra_info": null
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 40,
"std_ratio": 1.8,
"output_path": "./outputs/clean_nb40_std1.8.ply"
}
],
"gt_summary": " Single-View 3D Reconstruction, Post Processing \n Reconstruct a 3D asset from the image at {input_file_path}. The model quality should be medium quality. Save the output as a ply file. Save the output to ./outputs/scan11.ply.\nThen, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 40 neighbors and a std_ratio of 1.8. Save the clean cloud to ./outputs/clean_nb40_std1.8.ply. \n Input 1:./data/source/singleview/object/scan11.jpg; Input 2:./outputs/scan11.ply; \n Step 1: { Constraints: [quality: Medium quality, output_format: ply]; Output Path: ./outputs/scan11.ply }; Step 2: { nb_neighbors: 40, std_ratio: 1.8, Output Path: ./outputs/clean_nb40_std1.8.ply } ",
"code": "self.executor.single_view_reconstruction(image_path='./data/source/singleview/object/scan11.jpg', output_path='./outputs/scan11.ply', model='vggt', conf_thres=20.0)\nself.executor.post_processing(input_path='./outputs/scan11.ply', process_type='outliers_removal', output_path='./outputs/clean_nb40_std1.8.ply', parameters={\"nb_neighbors\": 40, \"std_ratio\": 1.8})",
"id": 28,
"ambiguous": "Fix {input_file_path} to 3D.\nThen, Fix the noise in the output of the previous step.",
"missing_info": "Process the single-view image to create a 3D representation. Ensure the result quality is defined. Save the output to a file. The result should be saved in a location.\nThen, Perform post-processing on a file to remove noise. Use a neighbor count of a number and a std_ratio of a value, then save to somewhere.",
"mistake": "Reconstruct a 3D asset from the image at ./data/source/singleview/object/scan11.obj. The model quality should be medium quality. Save the output as a ply file. Save the output to ./outputs/scan11.ply.\nThen, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 40 neighbors and a std_ratio of 1.8. Save the clean cloud to ./outputs/clean_nb40_std1.8.ply.",
"mistake_type": "input_format"
},
{
"complete": "Load image {input_file_path} and reconstruct it into 3D asset. Export format: ply. Focus on dense point cloud. The result should be saved in ./outputs/scan62.ply.\nThen, help me filter the noise in the output path of the previous step. Use statistical removal with 45 neighbors and 2.3 std_ratio. The output should be saved at ./outputs/clean_nb45_std2.3.ply.",
"file_path": "./data/source/singleview/object/scan62.jpg",
"ground_truth": [
{
"task": "Single-View 3D Reconstruction",
"usage": "object",
"quality": null,
"speed": null,
"conf_thres": null,
"output_format": "ply",
"output_path": "./outputs/scan62.ply",
"extra_info": "Dense point cloud"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 45,
"std_ratio": 2.3,
"output_path": "./outputs/clean_nb45_std2.3.ply"
}
],
"gt_summary": " Single-View 3D Reconstruction, Post Processing \n Load image {input_file_path} and reconstruct it into 3D asset. Export format: ply. Focus on dense point cloud. The result should be saved in ./outputs/scan62.ply.\nThen, help me filter the noise in the output path of the previous step. Use statistical removal with 45 neighbors and 2.3 std_ratio. The output should be saved at ./outputs/clean_nb45_std2.3.ply. \n Input 1:./data/source/singleview/object/scan62.jpg; Input 2:./outputs/scan62.ply; \n Step 1: { Constraints: [output_format: ply, features: Dense point cloud]; Output Path: ./outputs/scan62.ply }; Step 2: { nb_neighbors: 45, std_ratio: 2.3, Output Path: ./outputs/clean_nb45_std2.3.ply } ",
"code": "self.executor.single_view_reconstruction(image_path='./data/source/singleview/object/scan62.jpg', output_path='./outputs/scan62.ply', model='vggt', conf_thres=20.0)\nself.executor.post_processing(input_path='./outputs/scan62.ply', process_type='outliers_removal', output_path='./outputs/clean_nb45_std2.3.ply', parameters={\"nb_neighbors\": 45, \"std_ratio\": 2.3})",
"id": 29,
"ambiguous": "Fix {input_file_path} to 3D.\nThen, Clean the point cloud at the output of the previous step.",
"missing_info": "Process the single-view image to create a 3D representation. Focus on certain features. Export format: 3D model. Save the output to somewhere.\nThen, I have a noisy point cloud at a location. Please apply statistical outlier removal using a certain number of neighbors and a std_ratio of a value. Save the clean cloud to somewhere.",
"mistake": "Load image ./data/source/singleview/object/scan62.stl and reconstruct it into 3D asset. Export format: ply. Focus on dense point cloud. The result should be saved in ./outputs/scan62.ply.\nThen, help me filter the noise in the output path of the previous step. Use statistical removal with 45 neighbors and 2.3 std_ratio. The output should be saved at ./outputs/clean_nb45_std2.3.ply.",
"mistake_type": "input_format"
},
{
"complete": "Load the video or image folder at {input_file_path} and reconstruct it into a 3D asset. Save the output as a obj file. Save the output to ./outputs/scan110.obj.\nThen, please process the mesh at the output path of the previous step to create a point cloud. Use a sampling rate of 10000 points and write the file to ./outputs/example_mesh_from_glb_10000pts.ply.",
"file_path": "./data/source/multiview/object/scan110",
"ground_truth": [
{
"task": "Multi-View 3D Reconstruction",
"usage": "object",
"quality": null,
"speed": null,
"conf_thres": null,
"output_format": "obj",
"output_path": "./outputs/scan110.obj",
"extra_info": null
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 10000,
"output_path": "./outputs/example_mesh_from_glb_10000pts.ply"
}
],
"gt_summary": " Multi-View 3D Reconstruction, Post Processing \n Load the video or image folder at {input_file_path} and reconstruct it into a 3D asset. Save the output as a obj file. Save the output to ./outputs/scan110.obj.\nThen, please process the mesh at the output path of the previous step to create a point cloud. Use a sampling rate of 10000 points and write the file to ./outputs/example_mesh_from_glb_10000pts.ply. \n Input 1:./data/source/multiview/object/scan110; Input 2:./outputs/scan110.obj; \n Step 1: { Constraints: [output_format: obj]; Output Path: ./outputs/scan110.obj }; Step 2: { number_of_points: 10000; Output Path: ./outputs/example_mesh_from_glb_10000pts.ply } ",
"code": "self.executor.multi_view_reconstruction(video_path='./data/source/multiview/object/scan110', output_path='./outputs/scan110.obj', model='trellis')\nself.executor.post_processing(input_path='./outputs/scan110.obj', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_glb_10000pts.ply', parameters={\"number_of_points\": 10000})",
"id": 30,
"ambiguous": "Make 3D from the views in {input_file_path}.\nThen, Do something with the output of the previous step.",
"missing_info": "Generate a 3D asset from the views provided in a location. Export format: 3D model. Store the generated model.\nThen, Convert the mesh file at a location into a point cloud by sampling a certain number of points. Save the result to somewhere.",
"mistake": "Load the video or image folder at {input_file_path} and reconstruct it into a 3D asset. Save the output as a obj file. Save the output to ./outputs/scan110.obj.\nThen, please process the mesh at the output path of the previous step to create a point cloud. Use a sampling rate of 10000 points and write the file to ./outputs/example_mesh_from_glb_10000pts.ply. Ensure the output format is \".pdf\"",
"mistake_type": "output_format"
},
{
"complete": "Reconstruct a 3D scene from the image sequence at {input_file_path}. I need a obj file. The result should be saved in ./outputs/scan62.obj.\nThen, please process the mesh at the output path of the previous step to create a point cloud. Use a sampling rate of 30000 points and write the file to ./outputs/example_mesh_from_obj_30000pts.ply.",
"file_path": "./data/source/multiview/object/scan62",
"ground_truth": [
{
"task": "Multi-View 3D Reconstruction",
"usage": "object",
"quality": null,
"speed": null,
"conf_thres": null,
"output_format": "obj",
"output_path": "./outputs/scan62.obj",
"extra_info": null
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 30000,
"output_path": "./outputs/example_mesh_from_obj_30000pts.ply"
}
],
"gt_summary": " Multi-View 3D Reconstruction, Post Processing \n Reconstruct a 3D scene from the image sequence at {input_file_path}. I need a obj file. The result should be saved in ./outputs/scan62.obj.\nThen, please process the mesh at the output path of the previous step to create a point cloud. Use a sampling rate of 30000 points and write the file to ./outputs/example_mesh_from_obj_30000pts.ply. \n Input 1:./data/source/multiview/object/scan62; Input 2:./outputs/scan62.obj; \n Step 1: { Constraints: [output_format: obj]; Output Path: ./outputs/scan62.obj }; Step 2: { number_of_points: 30000; Output Path: ./outputs/example_mesh_from_obj_30000pts.ply } ",
"code": "self.executor.multi_view_reconstruction(video_path='./data/source/multiview/object/scan62', output_path='./outputs/scan62.obj', model='trellis')\nself.executor.post_processing(input_path='./outputs/scan62.obj', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_obj_30000pts.ply', parameters={\"number_of_points\": 30000})",
"id": 31,
"ambiguous": "Make 3D from the views in {input_file_path}.\nThen, Transform the output of the previous step.",
"missing_info": "Reconstruct a 3D scene from the image sequence at a location. Export format: 3D model. Store the generated model.\nThen, I need to sample a number of points uniformly from the 3D model at a path. Output the processed ply file to a location.",
"mistake": "Reconstruct a 3D scene from the image sequence at {input_file_path}. I need a obj file. The result should be saved in ./outputs/scan62.obj.\nThen, please process the mesh at the output path of the previous step to create a point cloud. Use a sampling rate of 30000 points and write the file to ./outputs/example_mesh_from_obj_30000pts.ply. Ensure the output format is \".pdf\"",
"mistake_type": "output_format"
},
{
"complete": "Generate a 3D asset from the views provided in {input_file_path}. I need it processed with medium speed. Ensure high-fidelity mesh. Store the result at ./outputs/scan114.glb.\nThen, convert the mesh file at the output path of the previous step into a point cloud by sampling 50000 points. Save the result to ./outputs/example_mesh_from_glb_50000pts.ply.",
"file_path": "./data/source/multiview/object/scan114",
"ground_truth": [
{
"task": "Multi-View 3D Reconstruction",
"usage": "object",
"quality": null,
"speed": "Medium speed",
"conf_thres": null,
"output_format": null,
"output_path": "./outputs/scan114.glb",
"extra_info": "High-fidelity mesh"
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 50000,
"output_path": "./outputs/example_mesh_from_glb_50000pts.ply"
}
],
"gt_summary": " Multi-View 3D Reconstruction, Post Processing \n Generate a 3D asset from the views provided in {input_file_path}. I need it processed with medium speed. Ensure high-fidelity mesh. Store the result at ./outputs/scan114.glb.\nThen, convert the mesh file at the output path of the previous step into a point cloud by sampling 50000 points. Save the result to ./outputs/example_mesh_from_glb_50000pts.ply. \n Input 1:./data/source/multiview/object/scan114; Input 2:./outputs/scan114.glb; \n Step 1: { Constraints: [speed: Medium speed, features: High-fidelity mesh]; Output Path: ./outputs/scan114.glb }; Step 2: { number_of_points: 50000; Output Path: ./outputs/example_mesh_from_glb_50000pts.ply } ",
"code": "self.executor.multi_view_reconstruction(video_path='./data/source/multiview/object/scan114', output_path='./outputs/scan114.glb', model='trellis')\nself.executor.post_processing(input_path='./outputs/scan114.glb', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_glb_50000pts.ply', parameters={\"number_of_points\": 50000})",
"id": 32,
"ambiguous": "Generate 3D from {input_file_path}.\nThen, Process the output of the previous step.",
"missing_info": "Reconstruct a 3D scene from the image sequence at a location. The generation speed should be specified. Consider that this has particular characteristics. Save the output to somewhere.\nThen, Convert the mesh file at a location into a point cloud by sampling a certain number of points. Save the result to somewhere.",
"mistake": "Generate a 3D asset from the views provided in {input_file_path}. I need it processed with medium speed. Ensure high-fidelity mesh. Store the result at ./outputs/scan114.glb.\nThen, convert the mesh file at the output path of the previous step into a point cloud by sampling 50000 points. Save the result to ./outputs/example_mesh_from_glb_50000pts.ply. Ensure low. Ensure extremely high speed.",
"mistake_type": "quality_speed"
},
{
"complete": "Perform multi-view 3D reconstruction using {input_file_path} as the input source. Save the output as a obj file. Ensure high-fidelity mesh. Save the output to ./outputs/scan75.obj.\nThen, please process the mesh at the output path of the previous step to create a point cloud. Use a sampling rate of 5000 points and write the file to ./outputs/example_mesh_from_obj_5000pts.ply.",
"file_path": "./data/source/multiview/object/scan75",
"ground_truth": [
{
"task": "Multi-View 3D Reconstruction",
"usage": "object",
"quality": null,
"speed": null,
"conf_thres": null,
"output_format": "obj",
"output_path": "./outputs/scan75.obj",
"extra_info": "High-fidelity mesh"
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 5000,
"output_path": "./outputs/example_mesh_from_obj_5000pts.ply"
}
],
"gt_summary": " Multi-View 3D Reconstruction, Post Processing \n Perform multi-view 3D reconstruction using {input_file_path} as the input source. Save the output as a obj file. Ensure high-fidelity mesh. Save the output to ./outputs/scan75.obj.\nThen, please process the mesh at the output path of the previous step to create a point cloud. Use a sampling rate of 5000 points and write the file to ./outputs/example_mesh_from_obj_5000pts.ply. \n Input 1:./data/source/multiview/object/scan75; Input 2:./outputs/scan75.obj; \n Step 1: { Constraints: [output_format: obj, features: High-fidelity mesh]; Output Path: ./outputs/scan75.obj }; Step 2: { number_of_points: 5000; Output Path: ./outputs/example_mesh_from_obj_5000pts.ply } ",
"code": "self.executor.multi_view_reconstruction(video_path='./data/source/multiview/object/scan75', output_path='./outputs/scan75.obj', model='trellis')\nself.executor.post_processing(input_path='./outputs/scan75.obj', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_obj_5000pts.ply', parameters={\"number_of_points\": 5000})",
"id": 33,
"ambiguous": "Make 3D from the views in {input_file_path}.\nThen, Process the output of the previous step.",
"missing_info": "Load the video or image folder from somewhere and reconstruct it into a 3D asset. Focus on certain features. I need a file. Save the output to somewhere.\nThen, I need to sample a number of points uniformly from the 3D model at a path. Output the processed ply file to a location.",
"mistake": "Perform multi-view 3D reconstruction using ./data/source/multiview/object/scan75.glb as the input source. Save the output as a obj file. Ensure high-fidelity mesh. Save the output to ./outputs/scan75.obj.\nThen, please process the mesh at the output path of the previous step to create a point cloud. Use a sampling rate of 5000 points and write the file to ./outputs/example_mesh_from_obj_5000pts.ply.",
"mistake_type": "input_format"
},
{
"complete": "Perform multi-view 3D reconstruction using {input_file_path} as the input source. The reconstruction speed should be medium speed. Save the output as a obj file. Focus on high-fidelity mesh. The result should be saved in ./outputs/scan48.obj.\nThen, please process the mesh at the output path of the previous step to create a point cloud. Use a sampling rate of 15000 points and write the file to ./outputs/example_mesh_from_obj_15000pts.ply.",
"file_path": "./data/source/multiview/object/scan48",
"ground_truth": [
{
"task": "Multi-View 3D Reconstruction",
"usage": "object",
"quality": null,
"speed": "Medium speed",
"conf_thres": null,
"output_format": "obj",
"output_path": "./outputs/scan48.obj",
"extra_info": "High-fidelity mesh"
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 15000,
"output_path": "./outputs/example_mesh_from_obj_15000pts.ply"
}
],
"gt_summary": " Multi-View 3D Reconstruction, Post Processing \n Perform multi-view 3D reconstruction using {input_file_path} as the input source. The reconstruction speed should be medium speed. Save the output as a obj file. Focus on high-fidelity mesh. The result should be saved in ./outputs/scan48.obj.\nThen, please process the mesh at the output path of the previous step to create a point cloud. Use a sampling rate of 15000 points and write the file to ./outputs/example_mesh_from_obj_15000pts.ply. \n Input 1:./data/source/multiview/object/scan48; Input 2:./outputs/scan48.obj; \n Step 1: { Constraints: [speed: Medium speed, output_format: obj, features: High-fidelity mesh]; Output Path: ./outputs/scan48.obj }; Step 2: { number_of_points: 15000; Output Path: ./outputs/example_mesh_from_obj_15000pts.ply } ",
"code": "self.executor.multi_view_reconstruction(video_path='./data/source/multiview/object/scan48', output_path='./outputs/scan48.obj', model='trellis')\nself.executor.post_processing(input_path='./outputs/scan48.obj', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_obj_15000pts.ply', parameters={\"number_of_points\": 15000})",
"id": 34,
"ambiguous": "Do something with {input_file_path}.\nThen, Do something with the output of the previous step.",
"missing_info": "Load the video or image folder from somewhere and reconstruct it into a 3D asset. Expect appropriate processing speed. Focus on certain features. Export format: 3D model. The result should be saved in a location.\nThen, Please process the mesh at a file to create a point cloud. Use a sampling rate of a number of points and write the file to somewhere.",
"mistake": "Perform multi-view 3D reconstruction using {input_file_path} as the input source. The reconstruction speed should be medium speed. Save the output as a obj file. Focus on high-fidelity mesh. The result should be saved in ./outputs/scan48.obj.\nThen, please process the mesh at the output path of the previous step to create a point cloud. Use a sampling rate of 15000 points and write the file to ./outputs/example_mesh_from_obj_15000pts.ply. Ensure low. Ensure extremely high speed.",
"mistake_type": "quality_speed"
},
{
"complete": "Load the video or image folder at {input_file_path} and reconstruct it into a 3D asset. Save the output as a obj file. Ensure high-fidelity mesh. Store the result at ./outputs/scan4.obj.\nThen, i need to sample 35000 points uniformly from the 3D model at the output path of the previous step. Output the processed ply file to ./outputs/example_mesh_from_obj_35000pts.ply.",
"file_path": "./data/source/multiview/object/scan4",
"ground_truth": [
{
"task": "Multi-View 3D Reconstruction",
"usage": "object",
"quality": null,
"speed": null,
"conf_thres": null,
"output_format": "obj",
"output_path": "./outputs/scan4.obj",
"extra_info": "High-fidelity mesh"
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 35000,
"output_path": "./outputs/example_mesh_from_obj_35000pts.ply"
}
],
"gt_summary": " Multi-View 3D Reconstruction, Post Processing \n Load the video or image folder at {input_file_path} and reconstruct it into a 3D asset. Save the output as a obj file. Ensure high-fidelity mesh. Store the result at ./outputs/scan4.obj.\nThen, i need to sample 35000 points uniformly from the 3D model at the output path of the previous step. Output the processed ply file to ./outputs/example_mesh_from_obj_35000pts.ply. \n Input 1:./data/source/multiview/object/scan4; Input 2:./outputs/scan4.obj; \n Step 1: { Constraints: [output_format: obj, features: High-fidelity mesh]; Output Path: ./outputs/scan4.obj }; Step 2: { number_of_points: 35000; Output Path: ./outputs/example_mesh_from_obj_35000pts.ply } ",
"code": "self.executor.multi_view_reconstruction(video_path='./data/source/multiview/object/scan4', output_path='./outputs/scan4.obj', model='trellis')\nself.executor.post_processing(input_path='./outputs/scan4.obj', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_obj_35000pts.ply', parameters={\"number_of_points\": 35000})",
"id": 35,
"ambiguous": "Do something with {input_file_path}.\nThen, Process the output of the previous step.",
"missing_info": "Load the video or image folder from somewhere and reconstruct it into a 3D asset. Focus on certain features. I need a file. The result should be saved in a location.\nThen, I need to sample a number of points uniformly from the 3D model at a path. Output the processed ply file to a location.",
"mistake": "Load the video or image folder at {input_file_path} and reconstruct it into a 3D asset. Save the output as a obj file. Ensure high-fidelity mesh. Store the result at ./outputs/scan4.obj.\nThen, i need to sample 35000 points uniformly from the 3D model at the output path of the previous step. Output the processed ply file to ./outputs/example_mesh_from_obj_35000pts.ply. Ensure low.",
"mistake_type": "quality_speed"
},
{
"complete": "Generate a 3D asset from the views provided in {input_file_path}. I need it processed with medium speed. Export format: glb. The result should be saved in ./outputs/scan118.glb.\nThen, take the mesh at the output path of the previous step and generate a point cloud with 15000 samples. Store the output at ./outputs/example_mesh_from_ply_15000pts.ply.",
"file_path": "./data/source/multiview/object/scan118",
"ground_truth": [
{
"task": "Multi-View 3D Reconstruction",
"usage": "object",
"quality": null,
"speed": "Medium speed",
"conf_thres": null,
"output_format": "glb",
"output_path": "./outputs/scan118.glb",
"extra_info": null
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 15000,
"output_path": "./outputs/example_mesh_from_ply_15000pts.ply"
}
],
"gt_summary": " Multi-View 3D Reconstruction, Post Processing \n Generate a 3D asset from the views provided in {input_file_path}. I need it processed with medium speed. Export format: glb. The result should be saved in ./outputs/scan118.glb.\nThen, take the mesh at the output path of the previous step and generate a point cloud with 15000 samples. Store the output at ./outputs/example_mesh_from_ply_15000pts.ply. \n Input 1:./data/source/multiview/object/scan118; Input 2:./outputs/scan118.glb; \n Step 1: { Constraints: [speed: Medium speed, output_format: glb]; Output Path: ./outputs/scan118.glb }; Step 2: { number_of_points: 15000; Output Path: ./outputs/example_mesh_from_ply_15000pts.ply } ",
"code": "self.executor.multi_view_reconstruction(video_path='./data/source/multiview/object/scan118', output_path='./outputs/scan118.glb', model='trellis')\nself.executor.post_processing(input_path='./outputs/scan118.glb', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_ply_15000pts.ply', parameters={\"number_of_points\": 15000})",
"id": 36,
"ambiguous": "Generate 3D from {input_file_path}.\nThen, Process the output of the previous step.",
"missing_info": "Perform multi-view 3D reconstruction using a file path as the input source. I need it generated at a certain speed. I need a file. The result should be saved in a location.\nThen, Perform mesh-to-pointcloud conversion on the mesh at a file. Set the target number of points to a number and save as somewhere.",
"mistake": "Generate a 3D asset from the views provided in ./data/source/multiview/object/scan118.stl. I need it processed with medium speed. Export format: glb. The result should be saved in ./outputs/scan118.glb.\nThen, take the mesh at the output path of the previous step and generate a point cloud with 15000 samples. Store the output at ./outputs/example_mesh_from_ply_15000pts.ply.",
"mistake_type": "input_format"
},
{
"complete": "Perform multi-view 3D reconstruction using {input_file_path} as the input source. Focus on high-fidelity mesh. Store the result at ./outputs/scan114.ply.\nThen, convert the mesh file at the output path of the previous step into a point cloud by sampling 25000 points. Save the result to ./outputs/example_mesh_from_glb_25000pts.ply.",
"file_path": "./data/source/multiview/object/scan114",
"ground_truth": [
{
"task": "Multi-View 3D Reconstruction",
"usage": "object",
"quality": null,
"speed": null,
"conf_thres": null,
"output_format": null,
"output_path": "./outputs/scan114.ply",
"extra_info": "High-fidelity mesh"
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 25000,
"output_path": "./outputs/example_mesh_from_glb_25000pts.ply"
}
],
"gt_summary": " Multi-View 3D Reconstruction, Post Processing \n Perform multi-view 3D reconstruction using {input_file_path} as the input source. Focus on high-fidelity mesh. Store the result at ./outputs/scan114.ply.\nThen, convert the mesh file at the output path of the previous step into a point cloud by sampling 25000 points. Save the result to ./outputs/example_mesh_from_glb_25000pts.ply. \n Input 1:./data/source/multiview/object/scan114; Input 2:./outputs/scan114.ply; \n Step 1: { Constraints: [features: High-fidelity mesh]; Output Path: ./outputs/scan114.ply }; Step 2: { number_of_points: 25000; Output Path: ./outputs/example_mesh_from_glb_25000pts.ply } ",
"code": "self.executor.multi_view_reconstruction(video_path='./data/source/multiview/object/scan114', output_path='./outputs/scan114.ply', model='trellis')\nself.executor.post_processing(input_path='./outputs/scan114.ply', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_glb_25000pts.ply', parameters={\"number_of_points\": 25000})",
"id": 37,
"ambiguous": "Do something with {input_file_path}.\nThen, Do something with the output of the previous step.",
"missing_info": "Load the video or image folder from somewhere and reconstruct it into a 3D asset. Consider that this has particular characteristics. Store the generated model.\nThen, Convert the mesh file at a location into a point cloud by sampling a certain number of points. Save the result to somewhere.",
"mistake": "Perform multi-view 3D reconstruction using {input_file_path} as the input source. Focus on high-fidelity mesh. Store the result at ./outputs/scan114.ply.\nThen, convert the mesh file at the output path of the previous step into a point cloud by sampling 25000 points. Save the result to ./outputs/example_mesh_from_glb_25000pts.ply. Ensure low.",
"mistake_type": "quality_speed"
},
{
"complete": "Process the multi-view folder located at {input_file_path} to create a 3D representation. Ensure the reconstruction is high quality. Save the output as a ply file. Focus on high-fidelity mesh. Save the output to ./outputs/scan13.ply.\nThen, i need to sample 45000 points uniformly from the 3D model at the output path of the previous step. Output the processed ply file to ./outputs/example_mesh_from_obj_45000pts.ply.",
"file_path": "./data/source/multiview/object/scan13",
"ground_truth": [
{
"task": "Multi-View 3D Reconstruction",
"usage": "object",
"quality": "High quality",
"speed": null,
"conf_thres": null,
"output_format": "ply",
"output_path": "./outputs/scan13.ply",
"extra_info": "High-fidelity mesh"
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 45000,
"output_path": "./outputs/example_mesh_from_obj_45000pts.ply"
}
],
"gt_summary": " Multi-View 3D Reconstruction, Post Processing \n Process the multi-view folder located at {input_file_path} to create a 3D representation. Ensure the reconstruction is high quality. Save the output as a ply file. Focus on high-fidelity mesh. Save the output to ./outputs/scan13.ply.\nThen, i need to sample 45000 points uniformly from the 3D model at the output path of the previous step. Output the processed ply file to ./outputs/example_mesh_from_obj_45000pts.ply. \n Input 1:./data/source/multiview/object/scan13; Input 2:./outputs/scan13.ply; \n Step 1: { Constraints: [quality: High quality, output_format: ply, features: High-fidelity mesh]; Output Path: ./outputs/scan13.ply }; Step 2: { number_of_points: 45000; Output Path: ./outputs/example_mesh_from_obj_45000pts.ply } ",
"code": "self.executor.multi_view_reconstruction(video_path='./data/source/multiview/object/scan13', output_path='./outputs/scan13.ply', model='trellis')\nself.executor.post_processing(input_path='./outputs/scan13.ply', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_obj_45000pts.ply', parameters={\"number_of_points\": 45000})",
"id": 38,
"ambiguous": "Process {input_file_path} into 3D.\nThen, Do something with the output of the previous step.",
"missing_info": "Process the multi-view folder located somewhere to create a 3D representation. The model quality should be specified. Focus on certain features. Export format: 3D model. Save the output to somewhere.\nThen, Please process the mesh at a file to create a point cloud. Use a sampling rate of a number of points and write the file to somewhere.",
"mistake": "Process the multi-view folder located at ./data/source/multiview/object/scan13.stl to create a 3D representation. Ensure the reconstruction is high quality. Save the output as a ply file. Focus on high-fidelity mesh. Save the output to ./outputs/scan13.ply.\nThen, i need to sample 45000 points uniformly from the 3D model at the output path of the previous step. Output the processed ply file to ./outputs/example_mesh_from_obj_45000pts.ply.",
"mistake_type": "input_format"
},
{
"complete": "Generate a 3D asset from the views provided in {input_file_path}. Ensure the reconstruction is high quality. Ensure high-fidelity mesh. Store the result at ./outputs/scan4.ply.\nThen, take the mesh at the output path of the previous step and generate a point cloud with 15000 samples. Store the output at ./outputs/example_mesh_from_obj_15000pts.ply.",
"file_path": "./data/source/multiview/object/scan4",
"ground_truth": [
{
"task": "Multi-View 3D Reconstruction",
"usage": "object",
"quality": "High quality",
"speed": null,
"conf_thres": null,
"output_format": null,
"output_path": "./outputs/scan4.ply",
"extra_info": "High-fidelity mesh"
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 15000,
"output_path": "./outputs/example_mesh_from_obj_15000pts.ply"
}
],
"gt_summary": " Multi-View 3D Reconstruction, Post Processing \n Generate a 3D asset from the views provided in {input_file_path}. Ensure the reconstruction is high quality. Ensure high-fidelity mesh. Store the result at ./outputs/scan4.ply.\nThen, take the mesh at the output path of the previous step and generate a point cloud with 15000 samples. Store the output at ./outputs/example_mesh_from_obj_15000pts.ply. \n Input 1:./data/source/multiview/object/scan4; Input 2:./outputs/scan4.ply; \n Step 1: { Constraints: [quality: High quality, features: High-fidelity mesh]; Output Path: ./outputs/scan4.ply }; Step 2: { number_of_points: 15000; Output Path: ./outputs/example_mesh_from_obj_15000pts.ply } ",
"code": "self.executor.multi_view_reconstruction(video_path='./data/source/multiview/object/scan4', output_path='./outputs/scan4.ply', model='trellis')\nself.executor.post_processing(input_path='./outputs/scan4.ply', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_obj_15000pts.ply', parameters={\"number_of_points\": 15000})",
"id": 39,
"ambiguous": "Generate 3D from {input_file_path}.\nThen, Transform the output of the previous step.",
"missing_info": "Generate a 3D asset from the views provided in a location. The model quality should be specified. Consider that this has particular characteristics. Save the output to somewhere.\nThen, Convert the mesh file at a location into a point cloud by sampling a certain number of points. Save the result to somewhere.",
"mistake": "Generate a 3D asset from the views provided in {input_file_path}. Ensure the reconstruction is high quality. Ensure high-fidelity mesh. Store the result at ./outputs/scan4.ply.\nThen, take the mesh at the output path of the previous step and generate a point cloud with 15000 samples. Store the output at ./outputs/example_mesh_from_obj_15000pts.ply. Ensure low. Ensure low quality.",
"mistake_type": "quality_speed"
},
{
"complete": "Reconstruct a 3D scene from the image sequence at {input_file_path}. I require ultra-high quality output. Save the output as a ply file. It should be a dense point cloud. Store the result at ./outputs/scan62.ply.\nThen, clean the point cloud file the output path of the previous step. I want to remove outliers with nb_neighbors set to 45 and standard deviation ratio set to 2.2. Output result to ./outputs/clean_nb45_std2.2.ply.",
"file_path": "./data/source/multiview/object/scan62",
"ground_truth": [
{
"task": "Multi-View 3D Reconstruction",
"usage": "object",
"quality": "Ultra-High quality",
"speed": null,
"conf_thres": null,
"output_format": "ply",
"output_path": "./outputs/scan62.ply",
"extra_info": "Dense point cloud"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 45,
"std_ratio": 2.2,
"output_path": "./outputs/clean_nb45_std2.2.ply"
}
],
"gt_summary": " Multi-View 3D Reconstruction, Post Processing \n Reconstruct a 3D scene from the image sequence at {input_file_path}. I require ultra-high quality output. Save the output as a ply file. It should be a dense point cloud. Store the result at ./outputs/scan62.ply.\nThen, clean the point cloud file the output path of the previous step. I want to remove outliers with nb_neighbors set to 45 and standard deviation ratio set to 2.2. Output result to ./outputs/clean_nb45_std2.2.ply. \n Input 1:./data/source/multiview/object/scan62; Input 2:./outputs/scan62.ply; \n Step 1: { Constraints: [quality: Ultra-High quality, output_format: ply, features: Dense point cloud]; Output Path: ./outputs/scan62.ply }; Step 2: { nb_neighbors: 45, std_ratio: 2.2, Output Path: ./outputs/clean_nb45_std2.2.ply } ",
"code": "self.executor.multi_view_reconstruction(video_path='./data/source/multiview/object/scan62', output_path='./outputs/scan62.ply', model='vggt', conf_thres=20.0)\nself.executor.post_processing(input_path='./outputs/scan62.ply', process_type='outliers_removal', output_path='./outputs/clean_nb45_std2.2.ply', parameters={\"nb_neighbors\": 45, \"std_ratio\": 2.2})",
"id": 40,
"ambiguous": "Generate 3D from {input_file_path}.\nThen, Fix the noise in the output of the previous step.",
"missing_info": "Reconstruct a 3D scene from the image sequence at a location. Ensure the result quality is defined. Consider that this has particular characteristics. Save the output to a file. Save the output to somewhere.\nThen, I have a noisy point cloud at a location. Please apply statistical outlier removal using a certain number of neighbors and a std_ratio of a value. Save the clean cloud to somewhere.",
"mistake": "Reconstruct a 3D scene from the image sequence at ./data/source/multiview/object/scan62.stl. I require ultra-high quality output. Save the output as a ply file. It should be a dense point cloud. Store the result at ./outputs/scan62.ply.\nThen, clean the point cloud file the output path of the previous step. I want to remove outliers with nb_neighbors set to 45 and standard deviation ratio set to 2.2. Output result to ./outputs/clean_nb45_std2.2.ply.",
"mistake_type": "input_format"
},
{
"complete": "Perform multi-view 3D reconstruction using {input_file_path} as the input source. Save the output as a glb file. Use a confidence cutoff of 27. The result should be saved in ./outputs/scan110.glb.\nThen, clean the point cloud file the output path of the previous step. I want to remove outliers with nb_neighbors set to 50 and standard deviation ratio set to 0.5. Output result to ./outputs/clean_nb50_std0.5.ply.",
"file_path": "./data/source/multiview/object/scan110",
"ground_truth": [
{
"task": "Multi-View 3D Reconstruction",
"usage": "object",
"quality": null,
"speed": null,
"conf_thres": 27,
"output_format": "glb",
"output_path": "./outputs/scan110.glb",
"extra_info": null
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 50,
"std_ratio": 0.5,
"output_path": "./outputs/clean_nb50_std0.5.ply"
}
],
"gt_summary": " Multi-View 3D Reconstruction, Post Processing \n Perform multi-view 3D reconstruction using {input_file_path} as the input source. Save the output as a glb file. Use a confidence cutoff of 27. The result should be saved in ./outputs/scan110.glb.\nThen, clean the point cloud file the output path of the previous step. I want to remove outliers with nb_neighbors set to 50 and standard deviation ratio set to 0.5. Output result to ./outputs/clean_nb50_std0.5.ply. \n Input 1:./data/source/multiview/object/scan110; Input 2:./outputs/scan110.glb; \n Step 1: { Constraints: [output_format: glb, conf_thres: 27]; Output Path: ./outputs/scan110.glb }; Step 2: { nb_neighbors: 50, std_ratio: 0.5, Output Path: ./outputs/clean_nb50_std0.5.ply } ",
"code": "self.executor.multi_view_reconstruction(video_path='./data/source/multiview/object/scan110', output_path='./outputs/scan110.glb', model='vggt', conf_thres=27)\nself.executor.post_processing(input_path='./outputs/scan110.glb', process_type='outliers_removal', output_path='./outputs/clean_nb50_std0.5.ply', parameters={\"nb_neighbors\": 50, \"std_ratio\": 0.5})",
"id": 41,
"ambiguous": "Process {input_file_path} into 3D.\nThen, Fix the noise in the output of the previous step.",
"missing_info": "Reconstruct a 3D scene from the image sequence at a location. Set the confidence threshold to a value. Save the output to a file. The result should be saved in a location.\nThen, Clean the point cloud file at a path. I want to remove outliers with nb_neighbors set to a number and standard deviation ratio set to a value. Output result to a location.",
"mistake": "Perform multi-view 3D reconstruction using {input_file_path} as the input source. Save the output as a glb file. Use a confidence cutoff of 2700. The result should be saved in ./outputs/scan110.glb.\nThen, clean the point cloud file the output path of the previous step. I want to remove outliers with nb_neighbors set to 50 and standard deviation ratio set to 0.5. Output result to ./outputs/clean_nb50_std0.5.ply.",
"mistake_type": "parameter"
},
{
"complete": "Reconstruct a 3D scene from the image sequence at {input_file_path}. I need it processed with fast speed. Export format: ply. Focus on dense point cloud. Store the result at ./outputs/Panther.ply.\nThen, help me filter the noise in the output path of the previous step. Use statistical removal with 40 neighbors and 0.8 std_ratio. The output should be saved at ./outputs/clean_nb40_std0.8.ply.",
"file_path": "./data/source/multiview/scene/Panther",
"ground_truth": [
{
"task": "Multi-View 3D Reconstruction",
"usage": "scene",
"quality": null,
"speed": "Fast speed",
"conf_thres": null,
"output_format": "ply",
"output_path": "./outputs/Panther.ply",
"extra_info": "Dense point cloud"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 40,
"std_ratio": 0.8,
"output_path": "./outputs/clean_nb40_std0.8.ply"
}
],
"gt_summary": " Multi-View 3D Reconstruction, Post Processing \n Reconstruct a 3D scene from the image sequence at {input_file_path}. I need it processed with fast speed. Export format: ply. Focus on dense point cloud. Store the result at ./outputs/Panther.ply.\nThen, help me filter the noise in the output path of the previous step. Use statistical removal with 40 neighbors and 0.8 std_ratio. The output should be saved at ./outputs/clean_nb40_std0.8.ply. \n Input 1:./data/source/multiview/scene/Panther; Input 2:./outputs/Panther.ply; \n Step 1: { Constraints: [speed: Fast speed, output_format: ply, features: Dense point cloud]; Output Path: ./outputs/Panther.ply }; Step 2: { nb_neighbors: 40, std_ratio: 0.8, Output Path: ./outputs/clean_nb40_std0.8.ply } ",
"code": "self.executor.multi_view_reconstruction(video_path='./data/source/multiview/scene/Panther', output_path='./outputs/Panther.ply', model='vggt', conf_thres=20.0)\nself.executor.post_processing(input_path='./outputs/Panther.ply', process_type='outliers_removal', output_path='./outputs/clean_nb40_std0.8.ply', parameters={\"nb_neighbors\": 40, \"std_ratio\": 0.8})",
"id": 42,
"ambiguous": "Make 3D from the views in {input_file_path}.\nThen, Do something with the output of the previous step.",
"missing_info": "Perform multi-view 3D reconstruction using a file path as the input source. Expect appropriate processing speed. It should have specific attributes. Save the output to a file. The result should be saved in a location.\nThen, Clean the point cloud file at a path. I want to remove outliers with nb_neighbors set to a number and standard deviation ratio set to a value. Output result to a location.",
"mistake": "Reconstruct a 3D scene from the image sequence at {input_file_path}. I need it processed with fast speed. Export format: ply. Focus on dense point cloud. Store the result at ./outputs/Panther.ply.\nThen, help me filter the noise in the output path of the previous step. Use statistical removal with 40 neighbors and 0.8 std_ratio. The output should be saved at ./outputs/clean_nb40_std0.8.ply. Ensure sparse point cloud. Ensure slow speed.",
"mistake_type": "quality_speed"
},
{
"complete": "Generate a 3D asset from the views provided in {input_file_path}. I require ultra-high quality output. Save the output as a ply file. The result should be saved in ./outputs/M60.ply.\nThen, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 25 and a std_ratio of 1.9, then save to ./outputs/clean_nb25_std1.9.ply.",
"file_path": "./data/source/multiview/scene/M60",
"ground_truth": [
{
"task": "Multi-View 3D Reconstruction",
"usage": "scene",
"quality": "Ultra-High quality",
"speed": null,
"conf_thres": null,
"output_format": "ply",
"output_path": "./outputs/M60.ply",
"extra_info": null
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 25,
"std_ratio": 1.9,
"output_path": "./outputs/clean_nb25_std1.9.ply"
}
],
"gt_summary": " Multi-View 3D Reconstruction, Post Processing \n Generate a 3D asset from the views provided in {input_file_path}. I require ultra-high quality output. Save the output as a ply file. The result should be saved in ./outputs/M60.ply.\nThen, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 25 and a std_ratio of 1.9, then save to ./outputs/clean_nb25_std1.9.ply. \n Input 1:./data/source/multiview/scene/M60; Input 2:./outputs/M60.ply; \n Step 1: { Constraints: [quality: Ultra-High quality, output_format: ply]; Output Path: ./outputs/M60.ply }; Step 2: { nb_neighbors: 25, std_ratio: 1.9, Output Path: ./outputs/clean_nb25_std1.9.ply } ",
"code": "self.executor.multi_view_reconstruction(video_path='./data/source/multiview/scene/M60', output_path='./outputs/M60.ply', model='vggt', conf_thres=20.0)\nself.executor.post_processing(input_path='./outputs/M60.ply', process_type='outliers_removal', output_path='./outputs/clean_nb25_std1.9.ply', parameters={\"nb_neighbors\": 25, \"std_ratio\": 1.9})",
"id": 43,
"ambiguous": "Generate 3D from {input_file_path}.\nThen, Process the output of the previous step.",
"missing_info": "Generate a 3D asset from the views provided in a location. Ensure the result quality is defined. Export format: 3D model. Save the output to somewhere.\nThen, Help me filter the noise in a location. Use statistical removal with certain neighbors and a std_ratio. The output should be saved at a path.",
"mistake": "Generate a 3D asset from the views provided in ./data/source/multiview/scene/M60.ply. I require ultra-high quality output. Save the output as a ply file. The result should be saved in ./outputs/M60.ply.\nThen, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 25 and a std_ratio of 1.9, then save to ./outputs/clean_nb25_std1.9.ply.",
"mistake_type": "input_format"
},
{
"complete": "Generate a 3D asset from the views provided in {input_file_path}. The model quality should be ultra-high quality. Use a confidence cutoff of 23. The result should be saved in ./outputs/scan10.ply.\nThen, help me filter the noise in the output path of the previous step. Use statistical removal with 50 neighbors and 0.6 std_ratio. The output should be saved at ./outputs/clean_nb50_std0.6.ply.",
"file_path": "./data/source/multiview/object/scan10",
"ground_truth": [
{
"task": "Multi-View 3D Reconstruction",
"usage": "object",
"quality": "Ultra-High quality",
"speed": null,
"conf_thres": 23,
"output_format": null,
"output_path": "./outputs/scan10.ply",
"extra_info": null
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 50,
"std_ratio": 0.6,
"output_path": "./outputs/clean_nb50_std0.6.ply"
}
],
"gt_summary": " Multi-View 3D Reconstruction, Post Processing \n Generate a 3D asset from the views provided in {input_file_path}. The model quality should be ultra-high quality. Use a confidence cutoff of 23. The result should be saved in ./outputs/scan10.ply.\nThen, help me filter the noise in the output path of the previous step. Use statistical removal with 50 neighbors and 0.6 std_ratio. The output should be saved at ./outputs/clean_nb50_std0.6.ply. \n Input 1:./data/source/multiview/object/scan10; Input 2:./outputs/scan10.ply; \n Step 1: { Constraints: [quality: Ultra-High quality, conf_thres: 23]; Output Path: ./outputs/scan10.ply }; Step 2: { nb_neighbors: 50, std_ratio: 0.6, Output Path: ./outputs/clean_nb50_std0.6.ply } ",
"code": "self.executor.multi_view_reconstruction(video_path='./data/source/multiview/object/scan10', output_path='./outputs/scan10.ply', model='vggt', conf_thres=23)\nself.executor.post_processing(input_path='./outputs/scan10.ply', process_type='outliers_removal', output_path='./outputs/clean_nb50_std0.6.ply', parameters={\"nb_neighbors\": 50, \"std_ratio\": 0.6})",
"id": 44,
"ambiguous": "Process {input_file_path} into 3D.\nThen, Fix the noise in the output of the previous step.",
"missing_info": "Load the video or image folder from somewhere and reconstruct it into a 3D asset. The model quality should be specified. Filter out points with confidence below a value. Save the output to somewhere.\nThen, Help me filter the noise in a location. Use statistical removal with certain neighbors and a std_ratio. The output should be saved at a path.",
"mistake": "Generate a 3D asset from the views provided in {input_file_path}. The model quality should be ultra-high quality. Use a confidence cutoff of -23. The result should be saved in ./outputs/scan10.ply.\nThen, help me filter the noise in the output path of the previous step. Use statistical removal with 50 neighbors and 0.6 std_ratio. The output should be saved at ./outputs/clean_nb50_std0.6.ply.",
"mistake_type": "parameter"
},
{
"complete": "Process the multi-view folder located at {input_file_path} to create a 3D representation. I need it processed with extreme fast speed. It should be a low resource usage. The result should be saved in ./outputs/counter.ply.\nThen, help me filter the noise in the output path of the previous step. Use statistical removal with 35 neighbors and 2.5 std_ratio. The output should be saved at ./outputs/clean_nb35_std2.5.ply.",
"file_path": "./data/source/multiview/scene/counter",
"ground_truth": [
{
"task": "Multi-View 3D Reconstruction",
"usage": "scene",
"quality": null,
"speed": "Extreme Fast speed",
"conf_thres": null,
"output_format": null,
"output_path": "./outputs/counter.ply",
"extra_info": "Low resource usage"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 35,
"std_ratio": 2.5,
"output_path": "./outputs/clean_nb35_std2.5.ply"
}
],
"gt_summary": " Multi-View 3D Reconstruction, Post Processing \n Process the multi-view folder located at {input_file_path} to create a 3D representation. I need it processed with extreme fast speed. It should be a low resource usage. The result should be saved in ./outputs/counter.ply.\nThen, help me filter the noise in the output path of the previous step. Use statistical removal with 35 neighbors and 2.5 std_ratio. The output should be saved at ./outputs/clean_nb35_std2.5.ply. \n Input 1:./data/source/multiview/scene/counter; Input 2:./outputs/counter.ply; \n Step 1: { Constraints: [speed: Extreme Fast speed, features: Low resource usage]; Output Path: ./outputs/counter.ply }; Step 2: { nb_neighbors: 35, std_ratio: 2.5, Output Path: ./outputs/clean_nb35_std2.5.ply } ",
"code": "self.executor.multi_view_reconstruction(video_path='./data/source/multiview/scene/counter', output_path='./outputs/counter.ply', model='fast3r', conf_thres=20.0)\nself.executor.post_processing(input_path='./outputs/counter.ply', process_type='outliers_removal', output_path='./outputs/clean_nb35_std2.5.ply', parameters={\"nb_neighbors\": 35, \"std_ratio\": 2.5})",
"id": 45,
"ambiguous": "Make 3D from the views in {input_file_path}.\nThen, Process the output of the previous step.",
"missing_info": "Perform multi-view 3D reconstruction using a file path as the input source. I need it generated at a certain speed. It should have specific attributes. Store the generated model.\nThen, I have a noisy point cloud at a location. Please apply statistical outlier removal using a certain number of neighbors and a std_ratio of a value. Save the clean cloud to somewhere.",
"mistake": "Process the multi-view folder located at ./data/source/multiview/scene/counter.stl to create a 3D representation. I need it processed with extreme fast speed. It should be a low resource usage. The result should be saved in ./outputs/counter.ply.\nThen, help me filter the noise in the output path of the previous step. Use statistical removal with 35 neighbors and 2.5 std_ratio. The output should be saved at ./outputs/clean_nb35_std2.5.ply.",
"mistake_type": "input_format"
},
{
"complete": "Load the video or image folder at {input_file_path} and reconstruct it into a 3D asset. Ensure the reconstruction is medium quality. Set the confidence threshold to 25. Store the result at ./outputs/bonsai.ply.\nThen, help me filter the noise in the output path of the previous step. Use statistical removal with 25 neighbors and 2.5 std_ratio. The output should be saved at ./outputs/clean_nb25_std2.5.ply.",
"file_path": "./data/source/multiview/scene/bonsai",
"ground_truth": [
{
"task": "Multi-View 3D Reconstruction",
"usage": "scene",
"quality": "Medium quality",
"speed": null,
"conf_thres": 25,
"output_format": null,
"output_path": "./outputs/bonsai.ply",
"extra_info": null
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 25,
"std_ratio": 2.5,
"output_path": "./outputs/clean_nb25_std2.5.ply"
}
],
"gt_summary": " Multi-View 3D Reconstruction, Post Processing \n Load the video or image folder at {input_file_path} and reconstruct it into a 3D asset. Ensure the reconstruction is medium quality. Set the confidence threshold to 25. Store the result at ./outputs/bonsai.ply.\nThen, help me filter the noise in the output path of the previous step. Use statistical removal with 25 neighbors and 2.5 std_ratio. The output should be saved at ./outputs/clean_nb25_std2.5.ply. \n Input 1:./data/source/multiview/scene/bonsai; Input 2:./outputs/bonsai.ply; \n Step 1: { Constraints: [quality: Medium quality, conf_thres: 25]; Output Path: ./outputs/bonsai.ply }; Step 2: { nb_neighbors: 25, std_ratio: 2.5, Output Path: ./outputs/clean_nb25_std2.5.ply } ",
"code": "self.executor.multi_view_reconstruction(video_path='./data/source/multiview/scene/bonsai', output_path='./outputs/bonsai.ply', model='fast3r', conf_thres=25)\nself.executor.post_processing(input_path='./outputs/bonsai.ply', process_type='outliers_removal', output_path='./outputs/clean_nb25_std2.5.ply', parameters={\"nb_neighbors\": 25, \"std_ratio\": 2.5})",
"id": 46,
"ambiguous": "Do something with {input_file_path}.\nThen, Clean the point cloud at the output of the previous step.",
"missing_info": "Reconstruct a 3D scene from the image sequence at a location. The model quality should be specified. Set the confidence threshold to a value. The result should be saved in a location.\nThen, Perform post-processing on a file to remove noise. Use a neighbor count of a number and a std_ratio of a value, then save to somewhere.",
"mistake": "Load the video or image folder at {input_file_path} and reconstruct it into a 3D asset. Ensure the reconstruction is medium quality. Set the confidence threshold to -25. Store the result at ./outputs/bonsai.ply.\nThen, help me filter the noise in the output path of the previous step. Use statistical removal with -25 neighbors and 2.5 std_ratio. The output should be saved at ./outputs/clean_nb-25_std2.5.ply.",
"mistake_type": "parameter"
},
{
"complete": "Load the video or image folder at {input_file_path} and reconstruct it into a 3D asset. Focus on low resource usage. Filter out points with confidence below 20. The result should be saved in ./outputs/scan62.ply.\nThen, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 50 and a std_ratio of 2.5, then save to ./outputs/clean_nb50_std2.5.ply.",
"file_path": "./data/source/multiview/object/scan62",
"ground_truth": [
{
"task": "Multi-View 3D Reconstruction",
"usage": "object",
"quality": null,
"speed": null,
"conf_thres": 20,
"output_format": null,
"output_path": "./outputs/scan62.ply",
"extra_info": "Low resource usage"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 50,
"std_ratio": 2.5,
"output_path": "./outputs/clean_nb50_std2.5.ply"
}
],
"gt_summary": " Multi-View 3D Reconstruction, Post Processing \n Load the video or image folder at {input_file_path} and reconstruct it into a 3D asset. Focus on low resource usage. Filter out points with confidence below 20. The result should be saved in ./outputs/scan62.ply.\nThen, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 50 and a std_ratio of 2.5, then save to ./outputs/clean_nb50_std2.5.ply. \n Input 1:./data/source/multiview/object/scan62; Input 2:./outputs/scan62.ply; \n Step 1: { Constraints: [features: Low resource usage, conf_thres: 20]; Output Path: ./outputs/scan62.ply }; Step 2: { nb_neighbors: 50, std_ratio: 2.5, Output Path: ./outputs/clean_nb50_std2.5.ply } ",
"code": "self.executor.multi_view_reconstruction(video_path='./data/source/multiview/object/scan62', output_path='./outputs/scan62.ply', model='fast3r', conf_thres=20)\nself.executor.post_processing(input_path='./outputs/scan62.ply', process_type='outliers_removal', output_path='./outputs/clean_nb50_std2.5.ply', parameters={\"nb_neighbors\": 50, \"std_ratio\": 2.5})",
"id": 47,
"ambiguous": "Do something with {input_file_path}.\nThen, Fix the noise in the output of the previous step.",
"missing_info": "Reconstruct a 3D scene from the image sequence at a location. Filter out points with confidence below a value. Focus on certain features. Save the output to somewhere.\nThen, Help me filter the noise in a location. Use statistical removal with certain neighbors and a std_ratio. The output should be saved at a path.",
"mistake": "Load the video or image folder at {input_file_path} and reconstruct it into a 3D asset. Focus on low resource usage. Filter out points with confidence below -20. The result should be saved in ./outputs/scan62.ply.\nThen, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 50 and a std_ratio of 2.5, then save to ./outputs/clean_nb50_std2.5.ply.",
"mistake_type": "parameter"
},
{
"complete": "Process the multi-view folder located at {input_file_path} to create a 3D representation. Expect extreme fast speed processing. Ensure low resource usage. Save the output to ./outputs/scan75.ply.\nThen, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 25 and a std_ratio of 2.2, then save to ./outputs/clean_nb25_std2.2.ply.",
"file_path": "./data/source/multiview/object/scan75",
"ground_truth": [
{
"task": "Multi-View 3D Reconstruction",
"usage": "object",
"quality": null,
"speed": "Extreme Fast speed",
"conf_thres": null,
"output_format": null,
"output_path": "./outputs/scan75.ply",
"extra_info": "Low resource usage"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 25,
"std_ratio": 2.2,
"output_path": "./outputs/clean_nb25_std2.2.ply"
}
],
"gt_summary": " Multi-View 3D Reconstruction, Post Processing \n Process the multi-view folder located at {input_file_path} to create a 3D representation. Expect extreme fast speed processing. Ensure low resource usage. Save the output to ./outputs/scan75.ply.\nThen, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 25 and a std_ratio of 2.2, then save to ./outputs/clean_nb25_std2.2.ply. \n Input 1:./data/source/multiview/object/scan75; Input 2:./outputs/scan75.ply; \n Step 1: { Constraints: [speed: Extreme Fast speed, features: Low resource usage]; Output Path: ./outputs/scan75.ply }; Step 2: { nb_neighbors: 25, std_ratio: 2.2, Output Path: ./outputs/clean_nb25_std2.2.ply } ",
"code": "self.executor.multi_view_reconstruction(video_path='./data/source/multiview/object/scan75', output_path='./outputs/scan75.ply', model='fast3r', conf_thres=20.0)\nself.executor.post_processing(input_path='./outputs/scan75.ply', process_type='outliers_removal', output_path='./outputs/clean_nb25_std2.2.ply', parameters={\"nb_neighbors\": 25, \"std_ratio\": 2.2})",
"id": 48,
"ambiguous": "Make 3D from the views in {input_file_path}.\nThen, Process the output of the previous step.",
"missing_info": "Process the multi-view folder located somewhere to create a 3D representation. I need it generated at a certain speed. Focus on certain features. Store the generated model.\nThen, Help me filter the noise in a location. Use statistical removal with certain neighbors and a std_ratio. The output should be saved at a path.",
"mistake": "Process the multi-view folder located at {input_file_path} to create a 3D representation. Expect extreme fast speed processing. Ensure low resource usage. Save the output to ./outputs/scan75.ply.\nThen, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 25 and a std_ratio of 2.2, then save to ./outputs/clean_nb25_std2.2.ply. Ensure high. Ensure slow speed.",
"mistake_type": "quality_speed"
},
{
"complete": "Process the multi-view folder located at {input_file_path} to create a 3D representation. I need a ply file. It should be a efficient. Save the output to ./outputs/scan10.ply.\nThen, clean the point cloud file the output path of the previous step. I want to remove outliers with nb_neighbors set to 40 and standard deviation ratio set to 1.2. Output result to ./outputs/clean_nb40_std1.2.ply.",
"file_path": "./data/source/multiview/object/scan10",
"ground_truth": [
{
"task": "Multi-View 3D Reconstruction",
"usage": "object",
"quality": null,
"speed": null,
"conf_thres": null,
"output_format": "ply",
"output_path": "./outputs/scan10.ply",
"extra_info": "Efficient"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 40,
"std_ratio": 1.2,
"output_path": "./outputs/clean_nb40_std1.2.ply"
}
],
"gt_summary": " Multi-View 3D Reconstruction, Post Processing \n Process the multi-view folder located at {input_file_path} to create a 3D representation. I need a ply file. It should be a efficient. Save the output to ./outputs/scan10.ply.\nThen, clean the point cloud file the output path of the previous step. I want to remove outliers with nb_neighbors set to 40 and standard deviation ratio set to 1.2. Output result to ./outputs/clean_nb40_std1.2.ply. \n Input 1:./data/source/multiview/object/scan10; Input 2:./outputs/scan10.ply; \n Step 1: { Constraints: [output_format: ply, features: Efficient]; Output Path: ./outputs/scan10.ply }; Step 2: { nb_neighbors: 40, std_ratio: 1.2, Output Path: ./outputs/clean_nb40_std1.2.ply } ",
"code": "self.executor.multi_view_reconstruction(video_path='./data/source/multiview/object/scan10', output_path='./outputs/scan10.ply', model='fast3r', conf_thres=20.0)\nself.executor.post_processing(input_path='./outputs/scan10.ply', process_type='outliers_removal', output_path='./outputs/clean_nb40_std1.2.ply', parameters={\"nb_neighbors\": 40, \"std_ratio\": 1.2})",
"id": 49,
"ambiguous": "Do something with {input_file_path}.\nThen, Process the output of the previous step.",
"missing_info": "Process the multi-view folder located somewhere to create a 3D representation. Focus on certain features. Export format: 3D model. The result should be saved in a location.\nThen, I have a noisy point cloud at a location. Please apply statistical outlier removal using a certain number of neighbors and a std_ratio of a value. Save the clean cloud to somewhere.",
"mistake": "Process the multi-view folder located at ./data/source/multiview/object/scan10.ply to create a 3D representation. I need a ply file. It should be a efficient. Save the output to ./outputs/scan10.ply.\nThen, clean the point cloud file the output path of the previous step. I want to remove outliers with nb_neighbors set to 40 and standard deviation ratio set to 1.2. Output result to ./outputs/clean_nb40_std1.2.ply.",
"mistake_type": "input_format"
},
{
"complete": "Process the multi-view folder located at {input_file_path} to create a 3D representation. The model quality should be ultra-high quality. I need a ply file. Store the result at ./outputs/Courtroom.ply.\nThen, extract the wall from the point cloud at the output path of the previous step. Discard all other objects. Save the extracted objects to ./outputs/keep_wall.ply.",
"file_path": "./data/source/multiview/scene/Courtroom",
"ground_truth": [
{
"task": "Multi-View 3D Reconstruction",
"usage": "scene",
"quality": "Ultra-High quality",
"speed": null,
"conf_thres": null,
"output_format": "ply",
"output_path": "./outputs/Courtroom.ply",
"extra_info": null
},
{
"task": "Point Cloud Editing",
"targets": [
"wall"
],
"edit_type": "keep",
"output_format": "ply",
"output_path": "./outputs/keep_wall.ply"
}
],
"gt_summary": " Multi-View 3D Reconstruction, Point Cloud Editing \n Process the multi-view folder located at {input_file_path} to create a 3D representation. The model quality should be ultra-high quality. I need a ply file. Store the result at ./outputs/Courtroom.ply.\nThen, extract the wall from the point cloud at the output path of the previous step. Discard all other objects. Save the extracted objects to ./outputs/keep_wall.ply. \n Input 1:./data/source/multiview/scene/Courtroom; Input 2:./outputs/Courtroom.ply; \n Step 1: { Constraints: [quality: Ultra-High quality, output_format: ply]; Output Path: ./outputs/Courtroom.ply }; Step 2: { Targets: wall, Output Path: ./outputs/keep_wall.ply } ",
"code": "self.executor.multi_view_reconstruction(video_path='./data/source/multiview/scene/Courtroom', output_path='./outputs/Courtroom.ply', model='vggt', conf_thres=20.0)\nself.executor.pcd_edit(pointcloud_path='./outputs/Courtroom.ply', edit_parts=[\"wall\"], output_path='./outputs/keep_wall.ply', edit_type='keep')",
"id": 50,
"ambiguous": "Generate 3D from {input_file_path}.\nThen, Modify the file at the output of the previous step.",
"missing_info": "Reconstruct a 3D scene from the image sequence at a location. Ensure the result quality is defined. I need a file. Save the output to somewhere.\nThen, I have a point cloud file for a room, which is placed in a location. I want to keep certain objects in this point cloud and remove the rest. Please save the result to somewhere.",
"mistake": "Process the multi-view folder located at {input_file_path} to create a 3D representation. The model quality should be ultra-high quality. I need a ply file. Store the result at ./outputs/Courtroom.ply.\nThen, extract the wall from the point cloud at the output path of the previous step. Discard all other objects. Save the extracted objects to ./outputs/keep_wall.ply. Ensure low quality.",
"mistake_type": "quality_speed"
},
{
"complete": "Generate a 3D asset from the views provided in {input_file_path}. The reconstruction speed should be fast speed. Save the output as a ply file. Use a confidence cutoff of 21. Save the output to ./outputs/Courtroom.ply.\nThen, i have a point cloud file for a room, which is placed in the output path of the previous step. I want to keep the floor in this point cloud and remove the rest. Please save the result to ./outputs/keep_floor.ply.",
"file_path": "./data/source/multiview/scene/Courtroom",
"ground_truth": [
{
"task": "Multi-View 3D Reconstruction",
"usage": "scene",
"quality": null,
"speed": "Fast speed",
"conf_thres": 21,
"output_format": "ply",
"output_path": "./outputs/Courtroom.ply",
"extra_info": null
},
{
"task": "Point Cloud Editing",
"targets": [
"floor"
],
"edit_type": "keep",
"output_format": "ply",
"output_path": "./outputs/keep_floor.ply"
}
],
"gt_summary": " Multi-View 3D Reconstruction, Point Cloud Editing \n Generate a 3D asset from the views provided in {input_file_path}. The reconstruction speed should be fast speed. Save the output as a ply file. Use a confidence cutoff of 21. Save the output to ./outputs/Courtroom.ply.\nThen, i have a point cloud file for a room, which is placed in the output path of the previous step. I want to keep the floor in this point cloud and remove the rest. Please save the result to ./outputs/keep_floor.ply. \n Input 1:./data/source/multiview/scene/Courtroom; Input 2:./outputs/Courtroom.ply; \n Step 1: { Constraints: [speed: Fast speed, output_format: ply, conf_thres: 21]; Output Path: ./outputs/Courtroom.ply }; Step 2: { Targets: floor, Output Path: ./outputs/keep_floor.ply } ",
"code": "self.executor.multi_view_reconstruction(video_path='./data/source/multiview/scene/Courtroom', output_path='./outputs/Courtroom.ply', model='vggt', conf_thres=21)\nself.executor.pcd_edit(pointcloud_path='./outputs/Courtroom.ply', edit_parts=[\"floor\"], output_path='./outputs/keep_floor.ply', edit_type='keep')",
"id": 51,
"ambiguous": "Process {input_file_path} into 3D.\nThen, Edit the point cloud at the output of the previous step.",
"missing_info": "Process the multi-view folder located somewhere to create a 3D representation. The generation speed should be specified. Set the confidence threshold to a value. Export format: 3D model. Store the generated model.\nThen, I have a point cloud file for a room, which is placed in a location. I want to keep certain objects in this point cloud and remove the rest. Please save the result to somewhere.",
"mistake": "Generate a 3D asset from the views provided in {input_file_path}. The reconstruction speed should be fast speed. Save the output as a ply file. Use a confidence cutoff of -21. Save the output to ./outputs/Courtroom.ply.\nThen, i have a point cloud file for a room, which is placed in the output path of the previous step. I want to keep the floor in this point cloud and remove the rest. Please save the result to ./outputs/keep_floor.ply.",
"mistake_type": "parameter"
},
{
"complete": "Generate a 3D asset from the views provided in {input_file_path}. I require ultra-high quality output. I need it processed with fast speed. I need a ply file. Save the output to ./outputs/bonsai.ply.\nThen, i have a point cloud file for a room, which is placed in the output path of the previous step. I want to keep the floor in this point cloud and remove the rest. Please save the result to ./outputs/keep_floor.ply.",
"file_path": "./data/source/multiview/scene/bonsai",
"ground_truth": [
{
"task": "Multi-View 3D Reconstruction",
"usage": "scene",
"quality": "Ultra-High quality",
"speed": "Fast speed",
"conf_thres": null,
"output_format": "ply",
"output_path": "./outputs/bonsai.ply",
"extra_info": null
},
{
"task": "Point Cloud Editing",
"targets": [
"floor"
],
"edit_type": "keep",
"output_format": "ply",
"output_path": "./outputs/keep_floor.ply"
}
],
"gt_summary": " Multi-View 3D Reconstruction, Point Cloud Editing \n Generate a 3D asset from the views provided in {input_file_path}. I require ultra-high quality output. I need it processed with fast speed. I need a ply file. Save the output to ./outputs/bonsai.ply.\nThen, i have a point cloud file for a room, which is placed in the output path of the previous step. I want to keep the floor in this point cloud and remove the rest. Please save the result to ./outputs/keep_floor.ply. \n Input 1:./data/source/multiview/scene/bonsai; Input 2:./outputs/bonsai.ply; \n Step 1: { Constraints: [quality: Ultra-High quality, speed: Fast speed, output_format: ply]; Output Path: ./outputs/bonsai.ply }; Step 2: { Targets: floor, Output Path: ./outputs/keep_floor.ply } ",
"code": "self.executor.multi_view_reconstruction(video_path='./data/source/multiview/scene/bonsai', output_path='./outputs/bonsai.ply', model='vggt', conf_thres=20.0)\nself.executor.pcd_edit(pointcloud_path='./outputs/bonsai.ply', edit_parts=[\"floor\"], output_path='./outputs/keep_floor.ply', edit_type='keep')",
"id": 52,
"ambiguous": "Do something with {input_file_path}.\nThen, Process the output of the previous step.",
"missing_info": "Load the video or image folder from somewhere and reconstruct it into a 3D asset. The model quality should be specified. The generation speed should be specified. Export format: 3D model. Save the output to somewhere.\nThen, Extract the objects from the point cloud at a location. Discard all other objects. Save the extracted objects to a path.",
"mistake": "Generate a 3D asset from the views provided in {input_file_path}. I require ultra-high quality output. I need it processed with fast speed. I need a ply file. Save the output to ./outputs/bonsai.ply.\nThen, i have a point cloud file for a room, which is placed in the output path of the previous step. I want to keep the floor in this point cloud and remove the rest. Please save the result to ./outputs/keep_floor.ply. Ensure slow speed. Ensure low quality.",
"mistake_type": "quality_speed"
},
{
"complete": "Load the video or image folder at {input_file_path} and reconstruct it into a 3D asset. I require ultra-high quality output. Export format: ply. Use a confidence cutoff of 29. The result should be saved in ./outputs/room.ply.\nThen, i have a point cloud file for a room, which is placed in the output path of the previous step. I want to keep the floor in this point cloud and remove the rest. Please save the result to ./outputs/keep_floor.ply.",
"file_path": "./data/source/multiview/scene/room",
"ground_truth": [
{
"task": "Multi-View 3D Reconstruction",
"usage": "scene",
"quality": "Ultra-High quality",
"speed": null,
"conf_thres": 29,
"output_format": "ply",
"output_path": "./outputs/room.ply",
"extra_info": null
},
{
"task": "Point Cloud Editing",
"targets": [
"floor"
],
"edit_type": "keep",
"output_format": "ply",
"output_path": "./outputs/keep_floor.ply"
}
],
"gt_summary": " Multi-View 3D Reconstruction, Point Cloud Editing \n Load the video or image folder at {input_file_path} and reconstruct it into a 3D asset. I require ultra-high quality output. Export format: ply. Use a confidence cutoff of 29. The result should be saved in ./outputs/room.ply.\nThen, i have a point cloud file for a room, which is placed in the output path of the previous step. I want to keep the floor in this point cloud and remove the rest. Please save the result to ./outputs/keep_floor.ply. \n Input 1:./data/source/multiview/scene/room; Input 2:./outputs/room.ply; \n Step 1: { Constraints: [quality: Ultra-High quality, output_format: ply, conf_thres: 29]; Output Path: ./outputs/room.ply }; Step 2: { Targets: floor, Output Path: ./outputs/keep_floor.ply } ",
"code": "self.executor.multi_view_reconstruction(video_path='./data/source/multiview/scene/room', output_path='./outputs/room.ply', model='vggt', conf_thres=29)\nself.executor.pcd_edit(pointcloud_path='./outputs/room.ply', edit_parts=[\"floor\"], output_path='./outputs/keep_floor.ply', edit_type='keep')",
"id": 53,
"ambiguous": "Process {input_file_path} into 3D.\nThen, Edit the point cloud at the output of the previous step.",
"missing_info": "Perform multi-view 3D reconstruction using a file path as the input source. I require a certain quality. Set the confidence threshold to a value. Export format: 3D model. The result should be saved in a location.\nThen, Extract the objects from the point cloud at a location. Discard all other objects. Save the extracted objects to a path.",
"mistake": "Load the video or image folder at {input_file_path} and reconstruct it into a 3D asset. I require ultra-high quality output. Export format: ply. Use a confidence cutoff of -29. The result should be saved in ./outputs/room.ply.\nThen, i have a point cloud file for a room, which is placed in the output path of the previous step. I want to keep the floor in this point cloud and remove the rest. Please save the result to ./outputs/keep_floor.ply.",
"mistake_type": "parameter"
},
{
"complete": "Reconstruct a 3D scene from the image sequence at {input_file_path}. I require medium quality output. Expect extreme fast speed processing. Export format: ply. Store the result at ./outputs/room.ply.\nThen, extract the wall from the point cloud at the output path of the previous step. Discard all other objects. Save the extracted objects to ./outputs/keep_wall.ply.",
"file_path": "./data/source/multiview/scene/room",
"ground_truth": [
{
"task": "Multi-View 3D Reconstruction",
"usage": "scene",
"quality": "Medium quality",
"speed": "Extreme Fast speed",
"conf_thres": null,
"output_format": "ply",
"output_path": "./outputs/room.ply",
"extra_info": null
},
{
"task": "Point Cloud Editing",
"targets": [
"wall"
],
"edit_type": "keep",
"output_format": "ply",
"output_path": "./outputs/keep_wall.ply"
}
],
"gt_summary": " Multi-View 3D Reconstruction, Point Cloud Editing \n Reconstruct a 3D scene from the image sequence at {input_file_path}. I require medium quality output. Expect extreme fast speed processing. Export format: ply. Store the result at ./outputs/room.ply.\nThen, extract the wall from the point cloud at the output path of the previous step. Discard all other objects. Save the extracted objects to ./outputs/keep_wall.ply. \n Input 1:./data/source/multiview/scene/room; Input 2:./outputs/room.ply; \n Step 1: { Constraints: [quality: Medium quality, speed: Extreme Fast speed, output_format: ply]; Output Path: ./outputs/room.ply }; Step 2: { Targets: wall, Output Path: ./outputs/keep_wall.ply } ",
"code": "self.executor.multi_view_reconstruction(video_path='./data/source/multiview/scene/room', output_path='./outputs/room.ply', model='fast3r', conf_thres=20.0)\nself.executor.pcd_edit(pointcloud_path='./outputs/room.ply', edit_parts=[\"wall\"], output_path='./outputs/keep_wall.ply', edit_type='keep')",
"id": 54,
"ambiguous": "Process {input_file_path} into 3D.\nThen, Modify the file at the output of the previous step.",
"missing_info": "Process the multi-view folder located somewhere to create a 3D representation. The model quality should be specified. I need it generated at a certain speed. Save the output to a file. Save the output to somewhere.\nThen, I have a point cloud file for a room, which is placed in a location. I want to keep certain objects in this point cloud and remove the rest. Please save the result to somewhere.",
"mistake": "Reconstruct a 3D scene from the image sequence at ./data/source/multiview/scene/room.stl. I require medium quality output. Expect extreme fast speed processing. Export format: ply. Store the result at ./outputs/room.ply.\nThen, extract the wall from the point cloud at the output path of the previous step. Discard all other objects. Save the extracted objects to ./outputs/keep_wall.ply.",
"mistake_type": "input_format"
},
{
"complete": "Clean the point cloud file {input_file_path}. I want to remove outliers with nb_neighbors set to 45 and standard deviation ratio set to 2.2. Output result to ./outputs/clean_nb45_std2.2.ply.\nThen, i have a point cloud file for a room, which is placed in the output path of the previous step. I want to keep the floor in this point cloud and remove the rest. Please save the result to ./outputs/keep_floor.ply.",
"file_path": "./data/source/scannet_scene0000_01.ply",
"ground_truth": [
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 45,
"std_ratio": 2.2,
"output_path": "./outputs/clean_nb45_std2.2.ply"
},
{
"task": "Point Cloud Editing",
"targets": [
"floor"
],
"edit_type": "keep",
"output_format": "ply",
"output_path": "./outputs/keep_floor.ply"
}
],
"gt_summary": " Post Processing, Point Cloud Editing \n Clean the point cloud file {input_file_path}. I want to remove outliers with nb_neighbors set to 45 and standard deviation ratio set to 2.2. Output result to ./outputs/clean_nb45_std2.2.ply.\nThen, i have a point cloud file for a room, which is placed in the output path of the previous step. I want to keep the floor in this point cloud and remove the rest. Please save the result to ./outputs/keep_floor.ply. \n Input 1:./data/source/scannet_scene0000_01.ply; Input 2:./outputs/clean_nb45_std2.2.ply; \n Step 1: { nb_neighbors: 45, std_ratio: 2.2, Output Path: ./outputs/clean_nb45_std2.2.ply }; Step 2: { Targets: floor, Output Path: ./outputs/keep_floor.ply } ",
"code": "self.executor.post_processing(input_path='./data/source/scannet_scene0000_01.ply', process_type='outliers_removal', output_path='./outputs/clean_nb45_std2.2.ply', parameters={\"nb_neighbors\": 45, \"std_ratio\": 2.2})\nself.executor.pcd_edit(pointcloud_path='./outputs/clean_nb45_std2.2.ply', edit_parts=[\"floor\"], output_path='./outputs/keep_floor.ply', edit_type='keep')",
"id": 55,
"ambiguous": "Fix the noise in {input_file_path}.\nThen, Modify the file at the output of the previous step.",
"missing_info": "Perform post-processing on a file to remove noise. Use a neighbor count of a number and a std_ratio of a value, then save to somewhere.\nThen, I have a point cloud file for a room, which is placed in a location. I want to keep certain objects in this point cloud and remove the rest. Please save the result to somewhere.",
"mistake": "Clean the point cloud file {input_file_path}. I want to remove outliers with nb_neighbors set to 45 and standard deviation ratio set to -2.2. Output result to ./outputs/clean_nb45_std-2.2.ply.\nThen, i have a point cloud file for a room, which is placed in the output path of the previous step. I want to keep the floor in this point cloud and remove the rest. Please save the result to ./outputs/keep_floor.ply.",
"mistake_type": "parameter"
},
{
"complete": "Perform post-processing on {input_file_path} to remove noise. Use a neighbor count of 25 and a std_ratio of 2.3, then save to ./outputs/clean_nb25_std2.3.ply.\nThen, extract the wall from the point cloud at the output path of the previous step. Discard all other objects. Save the extracted objects to ./outputs/keep_wall.ply.",
"file_path": "./data/source/scannet_scene0000_01.ply",
"ground_truth": [
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 25,
"std_ratio": 2.3,
"output_path": "./outputs/clean_nb25_std2.3.ply"
},
{
"task": "Point Cloud Editing",
"targets": [
"wall"
],
"edit_type": "keep",
"output_format": "ply",
"output_path": "./outputs/keep_wall.ply"
}
],
"gt_summary": " Post Processing, Point Cloud Editing \n Perform post-processing on {input_file_path} to remove noise. Use a neighbor count of 25 and a std_ratio of 2.3, then save to ./outputs/clean_nb25_std2.3.ply.\nThen, extract the wall from the point cloud at the output path of the previous step. Discard all other objects. Save the extracted objects to ./outputs/keep_wall.ply. \n Input 1:./data/source/scannet_scene0000_01.ply; Input 2:./outputs/clean_nb25_std2.3.ply; \n Step 1: { nb_neighbors: 25, std_ratio: 2.3, Output Path: ./outputs/clean_nb25_std2.3.ply }; Step 2: { Targets: wall, Output Path: ./outputs/keep_wall.ply } ",
"code": "self.executor.post_processing(input_path='./data/source/scannet_scene0000_01.ply', process_type='outliers_removal', output_path='./outputs/clean_nb25_std2.3.ply', parameters={\"nb_neighbors\": 25, \"std_ratio\": 2.3})\nself.executor.pcd_edit(pointcloud_path='./outputs/clean_nb25_std2.3.ply', edit_parts=[\"wall\"], output_path='./outputs/keep_wall.ply', edit_type='keep')",
"id": 56,
"ambiguous": "Clean the point cloud at {input_file_path}.\nThen, Process the output of the previous step.",
"missing_info": "I have a noisy point cloud at a location. Please apply statistical outlier removal using a certain number of neighbors and a std_ratio of a value. Save the clean cloud to somewhere.\nThen, Extract the objects from the point cloud at a location. Discard all other objects. Save the extracted objects to a path.",
"mistake": "Perform post-processing on {input_file_path} to remove noise. Use a neighbor count of 25 and a std_ratio of -2.3, then save to ./outputs/clean_nb25_std-2.3.ply.\nThen, extract the wall from the point cloud at the output path of the previous step. Discard all other objects. Save the extracted objects to ./outputs/keep_wall.ply.",
"mistake_type": "parameter"
},
{
"complete": "Help me filter the noise in {input_file_path}. Use statistical removal with 40 neighbors and 0.8 std_ratio. The output should be saved at ./outputs/clean_nb40_std0.8.ply.\nThen, extract the wall from the point cloud at the output path of the previous step. Discard all other objects. Save the extracted objects to ./outputs/keep_wall.ply.",
"file_path": "./data/source/scannet_scene0000_01.ply",
"ground_truth": [
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 40,
"std_ratio": 0.8,
"output_path": "./outputs/clean_nb40_std0.8.ply"
},
{
"task": "Point Cloud Editing",
"targets": [
"wall"
],
"edit_type": "keep",
"output_format": "ply",
"output_path": "./outputs/keep_wall.ply"
}
],
"gt_summary": " Post Processing, Point Cloud Editing \n Help me filter the noise in {input_file_path}. Use statistical removal with 40 neighbors and 0.8 std_ratio. The output should be saved at ./outputs/clean_nb40_std0.8.ply.\nThen, extract the wall from the point cloud at the output path of the previous step. Discard all other objects. Save the extracted objects to ./outputs/keep_wall.ply. \n Input 1:./data/source/scannet_scene0000_01.ply; Input 2:./outputs/clean_nb40_std0.8.ply; \n Step 1: { nb_neighbors: 40, std_ratio: 0.8, Output Path: ./outputs/clean_nb40_std0.8.ply }; Step 2: { Targets: wall, Output Path: ./outputs/keep_wall.ply } ",
"code": "self.executor.post_processing(input_path='./data/source/scannet_scene0000_01.ply', process_type='outliers_removal', output_path='./outputs/clean_nb40_std0.8.ply', parameters={\"nb_neighbors\": 40, \"std_ratio\": 0.8})\nself.executor.pcd_edit(pointcloud_path='./outputs/clean_nb40_std0.8.ply', edit_parts=[\"wall\"], output_path='./outputs/keep_wall.ply', edit_type='keep')",
"id": 57,
"ambiguous": "Fix the noise in {input_file_path}.\nThen, Edit the point cloud at the output of the previous step.",
"missing_info": "Clean the point cloud file at a path. I want to remove outliers with nb_neighbors set to a number and standard deviation ratio set to a value. Output result to a location.\nThen, I have a point cloud file for a room, which is placed in a location. I want to keep certain objects in this point cloud and remove the rest. Please save the result to somewhere.",
"mistake": "Help me filter the noise in {input_file_path}. Use statistical removal with 40 neighbors and -0.8 std_ratio. The output should be saved at ./outputs/clean_nb40_std-0.8.ply.\nThen, extract the wall from the point cloud at the output path of the previous step. Discard all other objects. Save the extracted objects to ./outputs/keep_wall.ply.",
"mistake_type": "parameter"
},
{
"complete": "Help me filter the noise in {input_file_path}. Use statistical removal with 35 neighbors and 2.5 std_ratio. The output should be saved at ./outputs/clean_nb35_std2.5.ply.\nThen, i have a point cloud file for a room, which is placed in the output path of the previous step. I want to delete the chair and table in this point cloud. Please save the result to ./outputs/delete_chair_table.ply.",
"file_path": "./data/source/scannet_scene0000_01.ply",
"ground_truth": [
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 35,
"std_ratio": 2.5,
"output_path": "./outputs/clean_nb35_std2.5.ply"
},
{
"task": "Point Cloud Editing",
"targets": [
"chair",
"table"
],
"edit_type": "delete",
"output_format": "ply",
"output_path": "./outputs/delete_chair_table.ply"
}
],
"gt_summary": " Post Processing, Point Cloud Editing \n Help me filter the noise in {input_file_path}. Use statistical removal with 35 neighbors and 2.5 std_ratio. The output should be saved at ./outputs/clean_nb35_std2.5.ply.\nThen, i have a point cloud file for a room, which is placed in the output path of the previous step. I want to delete the chair and table in this point cloud. Please save the result to ./outputs/delete_chair_table.ply. \n Input 1:./data/source/scannet_scene0000_01.ply; Input 2:./outputs/clean_nb35_std2.5.ply; \n Step 1: { nb_neighbors: 35, std_ratio: 2.5, Output Path: ./outputs/clean_nb35_std2.5.ply }; Step 2: { Targets: chair, table, Output Path: ./outputs/delete_chair_table.ply } ",
"code": "self.executor.post_processing(input_path='./data/source/scannet_scene0000_01.ply', process_type='outliers_removal', output_path='./outputs/clean_nb35_std2.5.ply', parameters={\"nb_neighbors\": 35, \"std_ratio\": 2.5})\nself.executor.pcd_edit(pointcloud_path='./outputs/clean_nb35_std2.5.ply', edit_parts=[\"chair\", \"table\"], output_path='./outputs/delete_chair_table.ply', edit_type='delete')",
"id": 58,
"ambiguous": "Do something with {input_file_path}.\nThen, Do something with the output of the previous step.",
"missing_info": "I have a noisy point cloud at a location. Please apply statistical outlier removal using a certain number of neighbors and a std_ratio of a value. Save the clean cloud to somewhere.\nThen, I have a point cloud file for a room, which is placed in a location. I want to delete the objects in this point cloud. Please save the result to somewhere.",
"mistake": "Help me filter the noise in {input_file_path}. Use statistical removal with 35 neighbors and -2.5 std_ratio. The output should be saved at ./outputs/clean_nb35_std-2.5.ply.\nThen, i have a point cloud file for a room, which is placed in the output path of the previous step. I want to delete the chair and table in this point cloud. Please save the result to ./outputs/delete_chair_table.ply.",
"mistake_type": "parameter"
},
{
"complete": "I have a noisy point cloud at {input_file_path}. Please apply statistical outlier removal using 45 neighbors and a std_ratio of 0.8. Save the clean cloud to ./outputs/clean_nb45_std0.8.ply.\nThen, i have a point cloud file for a room, which is placed in the output path of the previous step. I want to delete the chair and table in this point cloud. Please save the result to ./outputs/delete_chair_table.ply.",
"file_path": "./data/source/scannet_scene0000_01.ply",
"ground_truth": [
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 45,
"std_ratio": 0.8,
"output_path": "./outputs/clean_nb45_std0.8.ply"
},
{
"task": "Point Cloud Editing",
"targets": [
"chair",
"table"
],
"edit_type": "delete",
"output_format": "ply",
"output_path": "./outputs/delete_chair_table.ply"
}
],
"gt_summary": " Post Processing, Point Cloud Editing \n I have a noisy point cloud at {input_file_path}. Please apply statistical outlier removal using 45 neighbors and a std_ratio of 0.8. Save the clean cloud to ./outputs/clean_nb45_std0.8.ply.\nThen, i have a point cloud file for a room, which is placed in the output path of the previous step. I want to delete the chair and table in this point cloud. Please save the result to ./outputs/delete_chair_table.ply. \n Input 1:./data/source/scannet_scene0000_01.ply; Input 2:./outputs/clean_nb45_std0.8.ply; \n Step 1: { nb_neighbors: 45, std_ratio: 0.8, Output Path: ./outputs/clean_nb45_std0.8.ply }; Step 2: { Targets: chair, table, Output Path: ./outputs/delete_chair_table.ply } ",
"code": "self.executor.post_processing(input_path='./data/source/scannet_scene0000_01.ply', process_type='outliers_removal', output_path='./outputs/clean_nb45_std0.8.ply', parameters={\"nb_neighbors\": 45, \"std_ratio\": 0.8})\nself.executor.pcd_edit(pointcloud_path='./outputs/clean_nb45_std0.8.ply', edit_parts=[\"chair\", \"table\"], output_path='./outputs/delete_chair_table.ply', edit_type='delete')",
"id": 59,
"ambiguous": "Fix the noise in {input_file_path}.\nThen, Process the output of the previous step.",
"missing_info": "I have a noisy point cloud at a location. Please apply statistical outlier removal using a certain number of neighbors and a std_ratio of a value. Save the clean cloud to somewhere.\nThen, I have a point cloud file for a room, which is placed in a location. I want to delete the objects in this point cloud. Please save the result to somewhere.",
"mistake": "I have a noisy point cloud at {input_file_path}. Please apply statistical outlier removal using 45 neighbors and a std_ratio of -0.8. Save the clean cloud to ./outputs/clean_nb45_std-0.8.ply.\nThen, i have a point cloud file for a room, which is placed in the output path of the previous step. I want to delete the chair and table in this point cloud. Please save the result to ./outputs/delete_chair_table.ply.",
"mistake_type": "parameter"
},
{
"complete": "I have a noisy point cloud at {input_file_path}. Please apply statistical outlier removal using 50 neighbors and a std_ratio of 1.8. Save the clean cloud to ./outputs/clean_nb50_std1.8.ply.\nThen, i have a point cloud file for a room, which is placed in the output path of the previous step. I want to keep the floor in this point cloud and remove the rest. Please save the result to ./outputs/keep_floor.ply.",
"file_path": "./data/source/scannet_scene0000_01.ply",
"ground_truth": [
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 50,
"std_ratio": 1.8,
"output_path": "./outputs/clean_nb50_std1.8.ply"
},
{
"task": "Point Cloud Editing",
"targets": [
"floor"
],
"edit_type": "keep",
"output_format": "ply",
"output_path": "./outputs/keep_floor.ply"
}
],
"gt_summary": " Post Processing, Point Cloud Editing \n I have a noisy point cloud at {input_file_path}. Please apply statistical outlier removal using 50 neighbors and a std_ratio of 1.8. Save the clean cloud to ./outputs/clean_nb50_std1.8.ply.\nThen, i have a point cloud file for a room, which is placed in the output path of the previous step. I want to keep the floor in this point cloud and remove the rest. Please save the result to ./outputs/keep_floor.ply. \n Input 1:./data/source/scannet_scene0000_01.ply; Input 2:./outputs/clean_nb50_std1.8.ply; \n Step 1: { nb_neighbors: 50, std_ratio: 1.8, Output Path: ./outputs/clean_nb50_std1.8.ply }; Step 2: { Targets: floor, Output Path: ./outputs/keep_floor.ply } ",
"code": "self.executor.post_processing(input_path='./data/source/scannet_scene0000_01.ply', process_type='outliers_removal', output_path='./outputs/clean_nb50_std1.8.ply', parameters={\"nb_neighbors\": 50, \"std_ratio\": 1.8})\nself.executor.pcd_edit(pointcloud_path='./outputs/clean_nb50_std1.8.ply', edit_parts=[\"floor\"], output_path='./outputs/keep_floor.ply', edit_type='keep')",
"id": 60,
"ambiguous": "Fix the noise in {input_file_path}.\nThen, Edit the point cloud at the output of the previous step.",
"missing_info": "Clean the point cloud file at a path. I want to remove outliers with nb_neighbors set to a number and standard deviation ratio set to a value. Output result to a location.\nThen, Extract the objects from the point cloud at a location. Discard all other objects. Save the extracted objects to a path.",
"mistake": "I have a noisy point cloud at {input_file_path}. Please apply statistical outlier removal using 50 neighbors and a std_ratio of -1.8. Save the clean cloud to ./outputs/clean_nb50_std-1.8.ply.\nThen, i have a point cloud file for a room, which is placed in the output path of the previous step. I want to keep the floor in this point cloud and remove the rest. Please save the result to ./outputs/keep_floor.ply.",
"mistake_type": "parameter"
},
{
"complete": "Help me filter the noise in {input_file_path}. Use statistical removal with 40 neighbors and 1.1 std_ratio. The output should be saved at ./outputs/clean_nb40_std1.1.ply.\nThen, extract the wall from the point cloud at the output path of the previous step. Discard all other objects. Save the extracted objects to ./outputs/keep_wall.ply.",
"file_path": "./data/source/scannet_scene0000_01.ply",
"ground_truth": [
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 40,
"std_ratio": 1.1,
"output_path": "./outputs/clean_nb40_std1.1.ply"
},
{
"task": "Point Cloud Editing",
"targets": [
"wall"
],
"edit_type": "keep",
"output_format": "ply",
"output_path": "./outputs/keep_wall.ply"
}
],
"gt_summary": " Post Processing, Point Cloud Editing \n Help me filter the noise in {input_file_path}. Use statistical removal with 40 neighbors and 1.1 std_ratio. The output should be saved at ./outputs/clean_nb40_std1.1.ply.\nThen, extract the wall from the point cloud at the output path of the previous step. Discard all other objects. Save the extracted objects to ./outputs/keep_wall.ply. \n Input 1:./data/source/scannet_scene0000_01.ply; Input 2:./outputs/clean_nb40_std1.1.ply; \n Step 1: { nb_neighbors: 40, std_ratio: 1.1, Output Path: ./outputs/clean_nb40_std1.1.ply }; Step 2: { Targets: wall, Output Path: ./outputs/keep_wall.ply } ",
"code": "self.executor.post_processing(input_path='./data/source/scannet_scene0000_01.ply', process_type='outliers_removal', output_path='./outputs/clean_nb40_std1.1.ply', parameters={\"nb_neighbors\": 40, \"std_ratio\": 1.1})\nself.executor.pcd_edit(pointcloud_path='./outputs/clean_nb40_std1.1.ply', edit_parts=[\"wall\"], output_path='./outputs/keep_wall.ply', edit_type='keep')",
"id": 61,
"ambiguous": "Do something with {input_file_path}.\nThen, Modify the file at the output of the previous step.",
"missing_info": "I have a noisy point cloud at a location. Please apply statistical outlier removal using a certain number of neighbors and a std_ratio of a value. Save the clean cloud to somewhere.\nThen, I have a point cloud file for a room, which is placed in a location. I want to keep certain objects in this point cloud and remove the rest. Please save the result to somewhere.",
"mistake": "Help me filter the noise in {input_file_path}. Use statistical removal with 40 neighbors and -1.1 std_ratio. The output should be saved at ./outputs/clean_nb40_std-1.1.ply.\nThen, extract the wall from the point cloud at the output path of the previous step. Discard all other objects. Save the extracted objects to ./outputs/keep_wall.ply.",
"mistake_type": "parameter"
},
{
"complete": "Help me filter the noise in {input_file_path}. Use statistical removal with 25 neighbors and 2.5 std_ratio. The output should be saved at ./outputs/clean_nb25_std2.5.ply.\nThen, i have a point cloud file for a room, which is placed in the output path of the previous step. I want to keep the floor in this point cloud and remove the rest. Please save the result to ./outputs/keep_floor.ply.",
"file_path": "./data/source/scannet_scene0000_01.ply",
"ground_truth": [
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 25,
"std_ratio": 2.5,
"output_path": "./outputs/clean_nb25_std2.5.ply"
},
{
"task": "Point Cloud Editing",
"targets": [
"floor"
],
"edit_type": "keep",
"output_format": "ply",
"output_path": "./outputs/keep_floor.ply"
}
],
"gt_summary": " Post Processing, Point Cloud Editing \n Help me filter the noise in {input_file_path}. Use statistical removal with 25 neighbors and 2.5 std_ratio. The output should be saved at ./outputs/clean_nb25_std2.5.ply.\nThen, i have a point cloud file for a room, which is placed in the output path of the previous step. I want to keep the floor in this point cloud and remove the rest. Please save the result to ./outputs/keep_floor.ply. \n Input 1:./data/source/scannet_scene0000_01.ply; Input 2:./outputs/clean_nb25_std2.5.ply; \n Step 1: { nb_neighbors: 25, std_ratio: 2.5, Output Path: ./outputs/clean_nb25_std2.5.ply }; Step 2: { Targets: floor, Output Path: ./outputs/keep_floor.ply } ",
"code": "self.executor.post_processing(input_path='./data/source/scannet_scene0000_01.ply', process_type='outliers_removal', output_path='./outputs/clean_nb25_std2.5.ply', parameters={\"nb_neighbors\": 25, \"std_ratio\": 2.5})\nself.executor.pcd_edit(pointcloud_path='./outputs/clean_nb25_std2.5.ply', edit_parts=[\"floor\"], output_path='./outputs/keep_floor.ply', edit_type='keep')",
"id": 62,
"ambiguous": "Clean the point cloud at {input_file_path}.\nThen, Edit the point cloud at the output of the previous step.",
"missing_info": "I have a noisy point cloud at a location. Please apply statistical outlier removal using a certain number of neighbors and a std_ratio of a value. Save the clean cloud to somewhere.\nThen, I have a point cloud file for a room, which is placed in a location. I want to keep certain objects in this point cloud and remove the rest. Please save the result to somewhere.",
"mistake": "Help me filter the noise in {input_file_path}. Use statistical removal with 25 neighbors and -2.5 std_ratio. The output should be saved at ./outputs/clean_nb25_std-2.5.ply.\nThen, i have a point cloud file for a room, which is placed in the output path of the previous step. I want to keep the floor in this point cloud and remove the rest. Please save the result to ./outputs/keep_floor.ply.",
"mistake_type": "parameter"
},
{
"complete": "Perform post-processing on {input_file_path} to remove noise. Use a neighbor count of 20 and a std_ratio of 2.2, then save to ./outputs/clean_nb20_std2.2.ply.\nThen, i have a point cloud file for a room, which is placed in the output path of the previous step. I want to delete the chair and table in this point cloud. Please save the result to ./outputs/delete_chair_table.ply.",
"file_path": "./data/source/scannet_scene0000_01.ply",
"ground_truth": [
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 20,
"std_ratio": 2.2,
"output_path": "./outputs/clean_nb20_std2.2.ply"
},
{
"task": "Point Cloud Editing",
"targets": [
"chair",
"table"
],
"edit_type": "delete",
"output_format": "ply",
"output_path": "./outputs/delete_chair_table.ply"
}
],
"gt_summary": " Post Processing, Point Cloud Editing \n Perform post-processing on {input_file_path} to remove noise. Use a neighbor count of 20 and a std_ratio of 2.2, then save to ./outputs/clean_nb20_std2.2.ply.\nThen, i have a point cloud file for a room, which is placed in the output path of the previous step. I want to delete the chair and table in this point cloud. Please save the result to ./outputs/delete_chair_table.ply. \n Input 1:./data/source/scannet_scene0000_01.ply; Input 2:./outputs/clean_nb20_std2.2.ply; \n Step 1: { nb_neighbors: 20, std_ratio: 2.2, Output Path: ./outputs/clean_nb20_std2.2.ply }; Step 2: { Targets: chair, table, Output Path: ./outputs/delete_chair_table.ply } ",
"code": "self.executor.post_processing(input_path='./data/source/scannet_scene0000_01.ply', process_type='outliers_removal', output_path='./outputs/clean_nb20_std2.2.ply', parameters={\"nb_neighbors\": 20, \"std_ratio\": 2.2})\nself.executor.pcd_edit(pointcloud_path='./outputs/clean_nb20_std2.2.ply', edit_parts=[\"chair\", \"table\"], output_path='./outputs/delete_chair_table.ply', edit_type='delete')",
"id": 63,
"ambiguous": "Do something with {input_file_path}.\nThen, Process the output of the previous step.",
"missing_info": "Help me filter the noise in a location. Use statistical removal with certain neighbors and a std_ratio. The output should be saved at a path.\nThen, Remove the objects from the point cloud file at a location. Save the remaining objects to a path.",
"mistake": "Perform post-processing on {input_file_path} to remove noise. Use a neighbor count of 20 and a std_ratio of -2.2, then save to ./outputs/clean_nb20_std-2.2.ply.\nThen, i have a point cloud file for a room, which is placed in the output path of the previous step. I want to delete the chair and table in this point cloud. Please save the result to ./outputs/delete_chair_table.ply.",
"mistake_type": "parameter"
},
{
"complete": "I have a noisy point cloud at {input_file_path}. Please apply statistical outlier removal using 20 neighbors and a std_ratio of 1.2. Save the clean cloud to ./outputs/clean_nb20_std1.2.ply.\nThen, i have a point cloud file for a room, which is placed in the output path of the previous step. I want to delete the chair and table in this point cloud. Please save the result to ./outputs/delete_chair_table.ply.",
"file_path": "./data/source/scannet_scene0000_01.ply",
"ground_truth": [
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 20,
"std_ratio": 1.2,
"output_path": "./outputs/clean_nb20_std1.2.ply"
},
{
"task": "Point Cloud Editing",
"targets": [
"chair",
"table"
],
"edit_type": "delete",
"output_format": "ply",
"output_path": "./outputs/delete_chair_table.ply"
}
],
"gt_summary": " Post Processing, Point Cloud Editing \n I have a noisy point cloud at {input_file_path}. Please apply statistical outlier removal using 20 neighbors and a std_ratio of 1.2. Save the clean cloud to ./outputs/clean_nb20_std1.2.ply.\nThen, i have a point cloud file for a room, which is placed in the output path of the previous step. I want to delete the chair and table in this point cloud. Please save the result to ./outputs/delete_chair_table.ply. \n Input 1:./data/source/scannet_scene0000_01.ply; Input 2:./outputs/clean_nb20_std1.2.ply; \n Step 1: { nb_neighbors: 20, std_ratio: 1.2, Output Path: ./outputs/clean_nb20_std1.2.ply }; Step 2: { Targets: chair, table, Output Path: ./outputs/delete_chair_table.ply } ",
"code": "self.executor.post_processing(input_path='./data/source/scannet_scene0000_01.ply', process_type='outliers_removal', output_path='./outputs/clean_nb20_std1.2.ply', parameters={\"nb_neighbors\": 20, \"std_ratio\": 1.2})\nself.executor.pcd_edit(pointcloud_path='./outputs/clean_nb20_std1.2.ply', edit_parts=[\"chair\", \"table\"], output_path='./outputs/delete_chair_table.ply', edit_type='delete')",
"id": 64,
"ambiguous": "Fix the noise in {input_file_path}.\nThen, Edit the point cloud at the output of the previous step.",
"missing_info": "Perform post-processing on a file to remove noise. Use a neighbor count of a number and a std_ratio of a value, then save to somewhere.\nThen, I have a point cloud file for a room, which is placed in a location. I want to delete the objects in this point cloud. Please save the result to somewhere.",
"mistake": "I have a noisy point cloud at {input_file_path}. Please apply statistical outlier removal using 20 neighbors and a std_ratio of -1.2. Save the clean cloud to ./outputs/clean_nb20_std-1.2.ply.\nThen, i have a point cloud file for a room, which is placed in the output path of the previous step. I want to delete the chair and table in this point cloud. Please save the result to ./outputs/delete_chair_table.ply.",
"mistake_type": "parameter"
},
{
"complete": "I have a point cloud file for a room, which is placed in {input_file_path}. I want to delete the bookshelf, door and sink in this point cloud. Please save the result to ./outputs/delete_bookshelf_door_sink.ply.\nThen, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 25 neighbors and a std_ratio of 1.9. Save the clean cloud to ./outputs/clean_nb25_std1.9.ply.",
"file_path": "./data/source/scannet_scene0000_01.ply",
"ground_truth": [
{
"task": "Point Cloud Editing",
"targets": [
"bookshelf",
"door",
"sink"
],
"edit_type": "delete",
"output_format": "ply",
"output_path": "./outputs/delete_bookshelf_door_sink.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 25,
"std_ratio": 1.9,
"output_path": "./outputs/clean_nb25_std1.9.ply"
}
],
"gt_summary": " Point Cloud Editing, Post Processing \n I have a point cloud file for a room, which is placed in {input_file_path}. I want to delete the bookshelf, door and sink in this point cloud. Please save the result to ./outputs/delete_bookshelf_door_sink.ply.\nThen, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 25 neighbors and a std_ratio of 1.9. Save the clean cloud to ./outputs/clean_nb25_std1.9.ply. \n Input 1:./data/source/scannet_scene0000_01.ply; Input 2:./outputs/delete_bookshelf_door_sink.ply; \n Step 1: { Targets: bookshelf, door, sink, Output Path: ./outputs/delete_bookshelf_door_sink.ply }; Step 2: { nb_neighbors: 25, std_ratio: 1.9, Output Path: ./outputs/clean_nb25_std1.9.ply } ",
"code": "self.executor.pcd_edit(pointcloud_path='./data/source/scannet_scene0000_01.ply', edit_parts=[\"bookshelf\", \"door\", \"sink\"], output_path='./outputs/delete_bookshelf_door_sink.ply', edit_type='delete')\nself.executor.post_processing(input_path='./outputs/delete_bookshelf_door_sink.ply', process_type='outliers_removal', output_path='./outputs/clean_nb25_std1.9.ply', parameters={\"nb_neighbors\": 25, \"std_ratio\": 1.9})",
"id": 65,
"ambiguous": "Do something with {input_file_path}.\nThen, Clean the point cloud at the output of the previous step.",
"missing_info": "Remove the objects from the point cloud file at a location. Save the remaining objects to a path.\nThen, Clean the point cloud file at a path. I want to remove outliers with nb_neighbors set to a number and standard deviation ratio set to a value. Output result to a location.",
"mistake": "I have a point cloud file for a room, which is placed in ./data/source/scannet_scene0000_01.jpg. I want to delete the bookshelf, door and sink in this point cloud. Please save the result to ./outputs/delete_bookshelf_door_sink.ply.\nThen, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 25 neighbors and a std_ratio of 1.9. Save the clean cloud to ./outputs/clean_nb25_std1.9.ply.",
"mistake_type": "input_format"
},
{
"complete": "Remove the bathtub and floor from the point cloud file at {input_file_path}. Save the remaining objects to ./outputs/delete_bathtub_floor.ply.\nThen, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 25 and a std_ratio of 2.3, then save to ./outputs/clean_nb25_std2.3.ply.",
"file_path": "./data/source/scannet_scene0000_01.ply",
"ground_truth": [
{
"task": "Point Cloud Editing",
"targets": [
"bathtub",
"floor"
],
"edit_type": "delete",
"output_format": "ply",
"output_path": "./outputs/delete_bathtub_floor.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 25,
"std_ratio": 2.3,
"output_path": "./outputs/clean_nb25_std2.3.ply"
}
],
"gt_summary": " Point Cloud Editing, Post Processing \n Remove the bathtub and floor from the point cloud file at {input_file_path}. Save the remaining objects to ./outputs/delete_bathtub_floor.ply.\nThen, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 25 and a std_ratio of 2.3, then save to ./outputs/clean_nb25_std2.3.ply. \n Input 1:./data/source/scannet_scene0000_01.ply; Input 2:./outputs/delete_bathtub_floor.ply; \n Step 1: { Targets: bathtub, floor, Output Path: ./outputs/delete_bathtub_floor.ply }; Step 2: { nb_neighbors: 25, std_ratio: 2.3, Output Path: ./outputs/clean_nb25_std2.3.ply } ",
"code": "self.executor.pcd_edit(pointcloud_path='./data/source/scannet_scene0000_01.ply', edit_parts=[\"bathtub\", \"floor\"], output_path='./outputs/delete_bathtub_floor.ply', edit_type='delete')\nself.executor.post_processing(input_path='./outputs/delete_bathtub_floor.ply', process_type='outliers_removal', output_path='./outputs/clean_nb25_std2.3.ply', parameters={\"nb_neighbors\": 25, \"std_ratio\": 2.3})",
"id": 66,
"ambiguous": "Edit the point cloud at {input_file_path}.\nThen, Process the output of the previous step.",
"missing_info": "Remove the objects from the point cloud file at a location. Save the remaining objects to a path.\nThen, Help me filter the noise in a location. Use statistical removal with certain neighbors and a std_ratio. The output should be saved at a path.",
"mistake": "Remove the bathtub and floor from the point cloud file at ./data/source/scannet_scene0000_01.jpg. Save the remaining objects to ./outputs/delete_bathtub_floor.ply.\nThen, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 25 and a std_ratio of 2.3, then save to ./outputs/clean_nb25_std2.3.ply.",
"mistake_type": "input_format"
},
{
"complete": "I have a point cloud file for a room, which is placed in {input_file_path}. I want to delete the sofa and window in this point cloud. Please save the result to ./outputs/delete_sofa_window.ply.\nThen, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 45 neighbors and a std_ratio of 0.8. Save the clean cloud to ./outputs/clean_nb45_std0.8.ply.",
"file_path": "./data/source/scannet_scene0000_01.ply",
"ground_truth": [
{
"task": "Point Cloud Editing",
"targets": [
"sofa",
"window"
],
"edit_type": "delete",
"output_format": "ply",
"output_path": "./outputs/delete_sofa_window.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 45,
"std_ratio": 0.8,
"output_path": "./outputs/clean_nb45_std0.8.ply"
}
],
"gt_summary": " Point Cloud Editing, Post Processing \n I have a point cloud file for a room, which is placed in {input_file_path}. I want to delete the sofa and window in this point cloud. Please save the result to ./outputs/delete_sofa_window.ply.\nThen, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 45 neighbors and a std_ratio of 0.8. Save the clean cloud to ./outputs/clean_nb45_std0.8.ply. \n Input 1:./data/source/scannet_scene0000_01.ply; Input 2:./outputs/delete_sofa_window.ply; \n Step 1: { Targets: sofa, window, Output Path: ./outputs/delete_sofa_window.ply }; Step 2: { nb_neighbors: 45, std_ratio: 0.8, Output Path: ./outputs/clean_nb45_std0.8.ply } ",
"code": "self.executor.pcd_edit(pointcloud_path='./data/source/scannet_scene0000_01.ply', edit_parts=[\"sofa\", \"window\"], output_path='./outputs/delete_sofa_window.ply', edit_type='delete')\nself.executor.post_processing(input_path='./outputs/delete_sofa_window.ply', process_type='outliers_removal', output_path='./outputs/clean_nb45_std0.8.ply', parameters={\"nb_neighbors\": 45, \"std_ratio\": 0.8})",
"id": 67,
"ambiguous": "Edit the point cloud at {input_file_path}.\nThen, Fix the noise in the output of the previous step.",
"missing_info": "I have a point cloud file for a room, which is placed in a location. I want to delete the objects in this point cloud. Please save the result to somewhere.\nThen, Perform post-processing on a file to remove noise. Use a neighbor count of a number and a std_ratio of a value, then save to somewhere.",
"mistake": "I have a point cloud file for a room, which is placed in ./data/source/scannet_scene0000_01.png. I want to delete the sofa and window in this point cloud. Please save the result to ./outputs/delete_sofa_window.ply.\nThen, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 45 neighbors and a std_ratio of 0.8. Save the clean cloud to ./outputs/clean_nb45_std0.8.ply.",
"mistake_type": "input_format"
},
{
"complete": "Remove the door from the point cloud file at {input_file_path}. Save the remaining objects to ./outputs/delete_door.ply.\nThen, help me filter the noise in the output path of the previous step. Use statistical removal with 45 neighbors and 2.3 std_ratio. The output should be saved at ./outputs/clean_nb45_std2.3.ply.",
"file_path": "./data/source/scannet_scene0000_01.ply",
"ground_truth": [
{
"task": "Point Cloud Editing",
"targets": [
"door"
],
"edit_type": "delete",
"output_format": "ply",
"output_path": "./outputs/delete_door.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 45,
"std_ratio": 2.3,
"output_path": "./outputs/clean_nb45_std2.3.ply"
}
],
"gt_summary": " Point Cloud Editing, Post Processing \n Remove the door from the point cloud file at {input_file_path}. Save the remaining objects to ./outputs/delete_door.ply.\nThen, help me filter the noise in the output path of the previous step. Use statistical removal with 45 neighbors and 2.3 std_ratio. The output should be saved at ./outputs/clean_nb45_std2.3.ply. \n Input 1:./data/source/scannet_scene0000_01.ply; Input 2:./outputs/delete_door.ply; \n Step 1: { Targets: door, Output Path: ./outputs/delete_door.ply }; Step 2: { nb_neighbors: 45, std_ratio: 2.3, Output Path: ./outputs/clean_nb45_std2.3.ply } ",
"code": "self.executor.pcd_edit(pointcloud_path='./data/source/scannet_scene0000_01.ply', edit_parts=[\"door\"], output_path='./outputs/delete_door.ply', edit_type='delete')\nself.executor.post_processing(input_path='./outputs/delete_door.ply', process_type='outliers_removal', output_path='./outputs/clean_nb45_std2.3.ply', parameters={\"nb_neighbors\": 45, \"std_ratio\": 2.3})",
"id": 68,
"ambiguous": "Do something with {input_file_path}.\nThen, Process the output of the previous step.",
"missing_info": "Remove the objects from the point cloud file at a location. Save the remaining objects to a path.\nThen, Perform post-processing on a file to remove noise. Use a neighbor count of a number and a std_ratio of a value, then save to somewhere.",
"mistake": "Remove the door from the point cloud file at ./data/source/scannet_scene0000_01.png. Save the remaining objects to ./outputs/delete_door.ply.\nThen, help me filter the noise in the output path of the previous step. Use statistical removal with 45 neighbors and 2.3 std_ratio. The output should be saved at ./outputs/clean_nb45_std2.3.ply.",
"mistake_type": "input_format"
},
{
"complete": "I have a point cloud file for a room, which is placed in {input_file_path}. I want to keep the desk, door, refrigerator and window in this point cloud and remove the rest. Please save the result to ./outputs/keep_desk_door_refrigerator_window.ply.\nThen, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 20 and a std_ratio of 1.2, then save to ./outputs/clean_nb20_std1.2.ply.",
"file_path": "./data/source/scannet_scene0000_01.ply",
"ground_truth": [
{
"task": "Point Cloud Editing",
"targets": [
"desk",
"door",
"refrigerator",
"window"
],
"edit_type": "keep",
"output_format": "ply",
"output_path": "./outputs/keep_desk_door_refrigerator_window.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 20,
"std_ratio": 1.2,
"output_path": "./outputs/clean_nb20_std1.2.ply"
}
],
"gt_summary": " Point Cloud Editing, Post Processing \n I have a point cloud file for a room, which is placed in {input_file_path}. I want to keep the desk, door, refrigerator and window in this point cloud and remove the rest. Please save the result to ./outputs/keep_desk_door_refrigerator_window.ply.\nThen, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 20 and a std_ratio of 1.2, then save to ./outputs/clean_nb20_std1.2.ply. \n Input 1:./data/source/scannet_scene0000_01.ply; Input 2:./outputs/keep_desk_door_refrigerator_window.ply; \n Step 1: { Targets: desk, door, refrigerator, window, Output Path: ./outputs/keep_desk_door_refrigerator_window.ply }; Step 2: { nb_neighbors: 20, std_ratio: 1.2, Output Path: ./outputs/clean_nb20_std1.2.ply } ",
"code": "self.executor.pcd_edit(pointcloud_path='./data/source/scannet_scene0000_01.ply', edit_parts=[\"desk\", \"door\", \"refrigerator\", \"window\"], output_path='./outputs/keep_desk_door_refrigerator_window.ply', edit_type='keep')\nself.executor.post_processing(input_path='./outputs/keep_desk_door_refrigerator_window.ply', process_type='outliers_removal', output_path='./outputs/clean_nb20_std1.2.ply', parameters={\"nb_neighbors\": 20, \"std_ratio\": 1.2})",
"id": 69,
"ambiguous": "Do something with {input_file_path}.\nThen, Do something with the output of the previous step.",
"missing_info": "I have a point cloud file for a room, which is placed in a location. I want to keep certain objects in this point cloud and remove the rest. Please save the result to somewhere.\nThen, I have a noisy point cloud at a location. Please apply statistical outlier removal using a certain number of neighbors and a std_ratio of a value. Save the clean cloud to somewhere.",
"mistake": "I have a point cloud file for a room, which is placed in ./data/source/scannet_scene0000_01.jpg. I want to keep the desk, door, refrigerator and window in this point cloud and remove the rest. Please save the result to ./outputs/keep_desk_door_refrigerator_window.ply.\nThen, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 20 and a std_ratio of 1.2, then save to ./outputs/clean_nb20_std1.2.ply.",
"mistake_type": "input_format"
},
{
"complete": "I have a point cloud file for a room, which is placed in {input_file_path}. I want to keep the cabinet, picture and table in this point cloud and remove the rest. Please save the result to ./outputs/keep_cabinet_picture_table.ply.\nThen, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 25 and a std_ratio of 0.6, then save to ./outputs/clean_nb25_std0.6.ply.",
"file_path": "./data/source/scannet_scene0000_01.ply",
"ground_truth": [
{
"task": "Point Cloud Editing",
"targets": [
"cabinet",
"picture",
"table"
],
"edit_type": "keep",
"output_format": "ply",
"output_path": "./outputs/keep_cabinet_picture_table.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 25,
"std_ratio": 0.6,
"output_path": "./outputs/clean_nb25_std0.6.ply"
}
],
"gt_summary": " Point Cloud Editing, Post Processing \n I have a point cloud file for a room, which is placed in {input_file_path}. I want to keep the cabinet, picture and table in this point cloud and remove the rest. Please save the result to ./outputs/keep_cabinet_picture_table.ply.\nThen, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 25 and a std_ratio of 0.6, then save to ./outputs/clean_nb25_std0.6.ply. \n Input 1:./data/source/scannet_scene0000_01.ply; Input 2:./outputs/keep_cabinet_picture_table.ply; \n Step 1: { Targets: cabinet, picture, table, Output Path: ./outputs/keep_cabinet_picture_table.ply }; Step 2: { nb_neighbors: 25, std_ratio: 0.6, Output Path: ./outputs/clean_nb25_std0.6.ply } ",
"code": "self.executor.pcd_edit(pointcloud_path='./data/source/scannet_scene0000_01.ply', edit_parts=[\"cabinet\", \"picture\", \"table\"], output_path='./outputs/keep_cabinet_picture_table.ply', edit_type='keep')\nself.executor.post_processing(input_path='./outputs/keep_cabinet_picture_table.ply', process_type='outliers_removal', output_path='./outputs/clean_nb25_std0.6.ply', parameters={\"nb_neighbors\": 25, \"std_ratio\": 0.6})",
"id": 70,
"ambiguous": "Modify the file at {input_file_path}.\nThen, Fix the noise in the output of the previous step.",
"missing_info": "Extract the objects from the point cloud at a location. Discard all other objects. Save the extracted objects to a path.\nThen, I have a noisy point cloud at a location. Please apply statistical outlier removal using a certain number of neighbors and a std_ratio of a value. Save the clean cloud to somewhere.",
"mistake": "I have a point cloud file for a room, which is placed in ./data/source/scannet_scene0000_01.png. I want to keep the cabinet, picture and table in this point cloud and remove the rest. Please save the result to ./outputs/keep_cabinet_picture_table.ply.\nThen, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 25 and a std_ratio of 0.6, then save to ./outputs/clean_nb25_std0.6.ply.",
"mistake_type": "input_format"
},
{
"complete": "I have a point cloud file for a room, which is placed in {input_file_path}. I want to delete the cabinet, desk, door and toilet in this point cloud. Please save the result to ./outputs/delete_cabinet_desk_door_toilet.ply.\nThen, clean the point cloud file the output path of the previous step. I want to remove outliers with nb_neighbors set to 45 and standard deviation ratio set to 1.3. Output result to ./outputs/clean_nb45_std1.3.ply.",
"file_path": "./data/source/scannet_scene0000_01.ply",
"ground_truth": [
{
"task": "Point Cloud Editing",
"targets": [
"cabinet",
"desk",
"door",
"toilet"
],
"edit_type": "delete",
"output_format": "ply",
"output_path": "./outputs/delete_cabinet_desk_door_toilet.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 45,
"std_ratio": 1.3,
"output_path": "./outputs/clean_nb45_std1.3.ply"
}
],
"gt_summary": " Point Cloud Editing, Post Processing \n I have a point cloud file for a room, which is placed in {input_file_path}. I want to delete the cabinet, desk, door and toilet in this point cloud. Please save the result to ./outputs/delete_cabinet_desk_door_toilet.ply.\nThen, clean the point cloud file the output path of the previous step. I want to remove outliers with nb_neighbors set to 45 and standard deviation ratio set to 1.3. Output result to ./outputs/clean_nb45_std1.3.ply. \n Input 1:./data/source/scannet_scene0000_01.ply; Input 2:./outputs/delete_cabinet_desk_door_toilet.ply; \n Step 1: { Targets: cabinet, desk, door, toilet, Output Path: ./outputs/delete_cabinet_desk_door_toilet.ply }; Step 2: { nb_neighbors: 45, std_ratio: 1.3, Output Path: ./outputs/clean_nb45_std1.3.ply } ",
"code": "self.executor.pcd_edit(pointcloud_path='./data/source/scannet_scene0000_01.ply', edit_parts=[\"cabinet\", \"desk\", \"door\", \"toilet\"], output_path='./outputs/delete_cabinet_desk_door_toilet.ply', edit_type='delete')\nself.executor.post_processing(input_path='./outputs/delete_cabinet_desk_door_toilet.ply', process_type='outliers_removal', output_path='./outputs/clean_nb45_std1.3.ply', parameters={\"nb_neighbors\": 45, \"std_ratio\": 1.3})",
"id": 71,
"ambiguous": "Do something with {input_file_path}.\nThen, Do something with the output of the previous step.",
"missing_info": "I have a point cloud file for a room, which is placed in a location. I want to delete the objects in this point cloud. Please save the result to somewhere.\nThen, Clean the point cloud file at a path. I want to remove outliers with nb_neighbors set to a number and standard deviation ratio set to a value. Output result to a location.",
"mistake": "I have a point cloud file for a room, which is placed in ./data/source/scannet_scene0000_01.png. I want to delete the cabinet, desk, door and toilet in this point cloud. Please save the result to ./outputs/delete_cabinet_desk_door_toilet.ply.\nThen, clean the point cloud file the output path of the previous step. I want to remove outliers with nb_neighbors set to 45 and standard deviation ratio set to 1.3. Output result to ./outputs/clean_nb45_std1.3.ply.",
"mistake_type": "input_format"
},
{
"complete": "I have a point cloud file for a room, which is placed in {input_file_path}. I want to delete the chair and curtain in this point cloud. Please save the result to ./outputs/delete_chair_curtain.ply.\nThen, help me filter the noise in the output path of the previous step. Use statistical removal with 40 neighbors and 1.1 std_ratio. The output should be saved at ./outputs/clean_nb40_std1.1.ply.",
"file_path": "./data/source/scannet_scene0000_01.ply",
"ground_truth": [
{
"task": "Point Cloud Editing",
"targets": [
"chair",
"curtain"
],
"edit_type": "delete",
"output_format": "ply",
"output_path": "./outputs/delete_chair_curtain.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 40,
"std_ratio": 1.1,
"output_path": "./outputs/clean_nb40_std1.1.ply"
}
],
"gt_summary": " Point Cloud Editing, Post Processing \n I have a point cloud file for a room, which is placed in {input_file_path}. I want to delete the chair and curtain in this point cloud. Please save the result to ./outputs/delete_chair_curtain.ply.\nThen, help me filter the noise in the output path of the previous step. Use statistical removal with 40 neighbors and 1.1 std_ratio. The output should be saved at ./outputs/clean_nb40_std1.1.ply. \n Input 1:./data/source/scannet_scene0000_01.ply; Input 2:./outputs/delete_chair_curtain.ply; \n Step 1: { Targets: chair, curtain, Output Path: ./outputs/delete_chair_curtain.ply }; Step 2: { nb_neighbors: 40, std_ratio: 1.1, Output Path: ./outputs/clean_nb40_std1.1.ply } ",
"code": "self.executor.pcd_edit(pointcloud_path='./data/source/scannet_scene0000_01.ply', edit_parts=[\"chair\", \"curtain\"], output_path='./outputs/delete_chair_curtain.ply', edit_type='delete')\nself.executor.post_processing(input_path='./outputs/delete_chair_curtain.ply', process_type='outliers_removal', output_path='./outputs/clean_nb40_std1.1.ply', parameters={\"nb_neighbors\": 40, \"std_ratio\": 1.1})",
"id": 72,
"ambiguous": "Modify the file at {input_file_path}.\nThen, Do something with the output of the previous step.",
"missing_info": "Remove the objects from the point cloud file at a location. Save the remaining objects to a path.\nThen, Clean the point cloud file at a path. I want to remove outliers with nb_neighbors set to a number and standard deviation ratio set to a value. Output result to a location.",
"mistake": "I have a point cloud file for a room, which is placed in ./data/source/scannet_scene0000_01.jpg. I want to delete the chair and curtain in this point cloud. Please save the result to ./outputs/delete_chair_curtain.ply.\nThen, help me filter the noise in the output path of the previous step. Use statistical removal with 40 neighbors and 1.1 std_ratio. The output should be saved at ./outputs/clean_nb40_std1.1.ply.",
"mistake_type": "input_format"
},
{
"complete": "Extract the bed, chair, table and toilet from the point cloud at {input_file_path}. Discard all other objects. Save the extracted objects to ./outputs/keep_bed_chair_table_toilet.ply.\nThen, clean the point cloud file the output path of the previous step. I want to remove outliers with nb_neighbors set to 45 and standard deviation ratio set to 1.6. Output result to ./outputs/clean_nb45_std1.6.ply.",
"file_path": "./data/source/scannet_scene0000_01.ply",
"ground_truth": [
{
"task": "Point Cloud Editing",
"targets": [
"bed",
"chair",
"table",
"toilet"
],
"edit_type": "keep",
"output_format": "ply",
"output_path": "./outputs/keep_bed_chair_table_toilet.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 45,
"std_ratio": 1.6,
"output_path": "./outputs/clean_nb45_std1.6.ply"
}
],
"gt_summary": " Point Cloud Editing, Post Processing \n Extract the bed, chair, table and toilet from the point cloud at {input_file_path}. Discard all other objects. Save the extracted objects to ./outputs/keep_bed_chair_table_toilet.ply.\nThen, clean the point cloud file the output path of the previous step. I want to remove outliers with nb_neighbors set to 45 and standard deviation ratio set to 1.6. Output result to ./outputs/clean_nb45_std1.6.ply. \n Input 1:./data/source/scannet_scene0000_01.ply; Input 2:./outputs/keep_bed_chair_table_toilet.ply; \n Step 1: { Targets: bed, chair, table, toilet, Output Path: ./outputs/keep_bed_chair_table_toilet.ply }; Step 2: { nb_neighbors: 45, std_ratio: 1.6, Output Path: ./outputs/clean_nb45_std1.6.ply } ",
"code": "self.executor.pcd_edit(pointcloud_path='./data/source/scannet_scene0000_01.ply', edit_parts=[\"bed\", \"chair\", \"table\", \"toilet\"], output_path='./outputs/keep_bed_chair_table_toilet.ply', edit_type='keep')\nself.executor.post_processing(input_path='./outputs/keep_bed_chair_table_toilet.ply', process_type='outliers_removal', output_path='./outputs/clean_nb45_std1.6.ply', parameters={\"nb_neighbors\": 45, \"std_ratio\": 1.6})",
"id": 73,
"ambiguous": "Modify the file at {input_file_path}.\nThen, Process the output of the previous step.",
"missing_info": "Extract the objects from the point cloud at a location. Discard all other objects. Save the extracted objects to a path.\nThen, I have a noisy point cloud at a location. Please apply statistical outlier removal using a certain number of neighbors and a std_ratio of a value. Save the clean cloud to somewhere.",
"mistake": "Extract the bed, chair, table and toilet from the point cloud at ./data/source/scannet_scene0000_01.jpg. Discard all other objects. Save the extracted objects to ./outputs/keep_bed_chair_table_toilet.ply.\nThen, clean the point cloud file the output path of the previous step. I want to remove outliers with nb_neighbors set to 45 and standard deviation ratio set to 1.6. Output result to ./outputs/clean_nb45_std1.6.ply.",
"mistake_type": "input_format"
},
{
"complete": "Extract the floor and picture from the point cloud at {input_file_path}. Discard all other objects. Save the extracted objects to ./outputs/keep_floor_picture.ply.\nThen, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 40 neighbors and a std_ratio of 1.1. Save the clean cloud to ./outputs/clean_nb40_std1.1.ply.",
"file_path": "./data/source/scannet_scene0000_01.ply",
"ground_truth": [
{
"task": "Point Cloud Editing",
"targets": [
"floor",
"picture"
],
"edit_type": "keep",
"output_format": "ply",
"output_path": "./outputs/keep_floor_picture.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 40,
"std_ratio": 1.1,
"output_path": "./outputs/clean_nb40_std1.1.ply"
}
],
"gt_summary": " Point Cloud Editing, Post Processing \n Extract the floor and picture from the point cloud at {input_file_path}. Discard all other objects. Save the extracted objects to ./outputs/keep_floor_picture.ply.\nThen, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 40 neighbors and a std_ratio of 1.1. Save the clean cloud to ./outputs/clean_nb40_std1.1.ply. \n Input 1:./data/source/scannet_scene0000_01.ply; Input 2:./outputs/keep_floor_picture.ply; \n Step 1: { Targets: floor, picture, Output Path: ./outputs/keep_floor_picture.ply }; Step 2: { nb_neighbors: 40, std_ratio: 1.1, Output Path: ./outputs/clean_nb40_std1.1.ply } ",
"code": "self.executor.pcd_edit(pointcloud_path='./data/source/scannet_scene0000_01.ply', edit_parts=[\"floor\", \"picture\"], output_path='./outputs/keep_floor_picture.ply', edit_type='keep')\nself.executor.post_processing(input_path='./outputs/keep_floor_picture.ply', process_type='outliers_removal', output_path='./outputs/clean_nb40_std1.1.ply', parameters={\"nb_neighbors\": 40, \"std_ratio\": 1.1})",
"id": 74,
"ambiguous": "Edit the point cloud at {input_file_path}.\nThen, Fix the noise in the output of the previous step.",
"missing_info": "I have a point cloud file for a room, which is placed in a location. I want to keep certain objects in this point cloud and remove the rest. Please save the result to somewhere.\nThen, Perform post-processing on a file to remove noise. Use a neighbor count of a number and a std_ratio of a value, then save to somewhere.",
"mistake": "Extract the floor and picture from the point cloud at ./data/source/scannet_scene0000_01.jpg. Discard all other objects. Save the extracted objects to ./outputs/keep_floor_picture.ply.\nThen, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 40 neighbors and a std_ratio of 1.1. Save the clean cloud to ./outputs/clean_nb40_std1.1.ply.",
"mistake_type": "input_format"
},
{
"complete": "Perform mesh-to-pointcloud conversion on the mesh at {input_file_path}. Set the target number of points to 30000 and save as ./outputs/example_mesh_from_ply_30000pts.ply.\nThen, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 25 and a std_ratio of 1.9, then save to ./outputs/clean_nb25_std1.9.ply.",
"file_path": "./data/source/example_mesh.ply",
"ground_truth": [
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 30000,
"output_path": "./outputs/example_mesh_from_ply_30000pts.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 25,
"std_ratio": 1.9,
"output_path": "./outputs/clean_nb25_std1.9.ply"
}
],
"gt_summary": " Post Processing, Post Processing \n Perform mesh-to-pointcloud conversion on the mesh at {input_file_path}. Set the target number of points to 30000 and save as ./outputs/example_mesh_from_ply_30000pts.ply.\nThen, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 25 and a std_ratio of 1.9, then save to ./outputs/clean_nb25_std1.9.ply. \n Input 1:./data/source/example_mesh.ply; Input 2:./outputs/example_mesh_from_ply_30000pts.ply; \n Step 1: { number_of_points: 30000; Output Path: ./outputs/example_mesh_from_ply_30000pts.ply }; Step 2: { nb_neighbors: 25, std_ratio: 1.9, Output Path: ./outputs/clean_nb25_std1.9.ply } ",
"code": "self.executor.post_processing(input_path='./data/source/example_mesh.ply', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_ply_30000pts.ply', parameters={\"number_of_points\": 30000})\nself.executor.post_processing(input_path='./outputs/example_mesh_from_ply_30000pts.ply', process_type='outliers_removal', output_path='./outputs/clean_nb25_std1.9.ply', parameters={\"nb_neighbors\": 25, \"std_ratio\": 1.9})",
"id": 75,
"ambiguous": "Do something with {input_file_path}.\nThen, Clean the point cloud at the output of the previous step.",
"missing_info": "Perform mesh-to-pointcloud conversion on the mesh at a file. Set the target number of points to a number and save as somewhere.\nThen, Clean the point cloud file at a path. I want to remove outliers with nb_neighbors set to a number and standard deviation ratio set to a value. Output result to a location.",
"mistake": "Perform mesh-to-pointcloud conversion on the mesh at {input_file_path}. Set the target number of points to 30000 and save as ./outputs/example_mesh_from_ply_30000pts.ply.\nThen, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 25 and a std_ratio of 1.9, then save to ./outputs/clean_nb25_std1.9.ply. Ensure the output format is \".pdf\"",
"mistake_type": "output_format"
},
{
"complete": "Please process the mesh at {input_file_path} to create a point cloud. Use a sampling rate of 30000 points and write the file to ./outputs/example_mesh_from_obj_30000pts.ply.\nThen, clean the point cloud file the output path of the previous step. I want to remove outliers with nb_neighbors set to 25 and standard deviation ratio set to 2.5. Output result to ./outputs/clean_nb25_std2.5.ply.",
"file_path": "./data/source/example_mesh.obj",
"ground_truth": [
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 30000,
"output_path": "./outputs/example_mesh_from_obj_30000pts.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 25,
"std_ratio": 2.5,
"output_path": "./outputs/clean_nb25_std2.5.ply"
}
],
"gt_summary": " Post Processing, Post Processing \n Please process the mesh at {input_file_path} to create a point cloud. Use a sampling rate of 30000 points and write the file to ./outputs/example_mesh_from_obj_30000pts.ply.\nThen, clean the point cloud file the output path of the previous step. I want to remove outliers with nb_neighbors set to 25 and standard deviation ratio set to 2.5. Output result to ./outputs/clean_nb25_std2.5.ply. \n Input 1:./data/source/example_mesh.obj; Input 2:./outputs/example_mesh_from_obj_30000pts.ply; \n Step 1: { number_of_points: 30000; Output Path: ./outputs/example_mesh_from_obj_30000pts.ply }; Step 2: { nb_neighbors: 25, std_ratio: 2.5, Output Path: ./outputs/clean_nb25_std2.5.ply } ",
"code": "self.executor.post_processing(input_path='./data/source/example_mesh.obj', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_obj_30000pts.ply', parameters={\"number_of_points\": 30000})\nself.executor.post_processing(input_path='./outputs/example_mesh_from_obj_30000pts.ply', process_type='outliers_removal', output_path='./outputs/clean_nb25_std2.5.ply', parameters={\"nb_neighbors\": 25, \"std_ratio\": 2.5})",
"id": 76,
"ambiguous": "Process {input_file_path}.\nThen, Clean the point cloud at the output of the previous step.",
"missing_info": "Take the mesh at a location and generate a point cloud with a certain number of samples. Store the output at a path.\nThen, Perform post-processing on a file to remove noise. Use a neighbor count of a number and a std_ratio of a value, then save to somewhere.",
"mistake": "Please process the mesh at {input_file_path} to create a point cloud. Use a sampling rate of 30000 points and write the file to ./outputs/example_mesh_from_obj_30000pts.ply.\nThen, clean the point cloud file the output path of the previous step. I want to remove outliers with nb_neighbors set to 25 and standard deviation ratio set to 2.5. Output result to ./outputs/clean_nb25_std2.5.ply. Ensure the output format is \".pdf\"",
"mistake_type": "output_format"
},
{
"complete": "Please process the mesh at {input_file_path} to create a point cloud. Use a sampling rate of 10000 points and write the file to ./outputs/example_mesh_from_glb_10000pts.ply.\nThen, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 45 neighbors and a std_ratio of 1.5. Save the clean cloud to ./outputs/clean_nb45_std1.5.ply.",
"file_path": "./data/source/example_mesh.glb",
"ground_truth": [
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 10000,
"output_path": "./outputs/example_mesh_from_glb_10000pts.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 45,
"std_ratio": 1.5,
"output_path": "./outputs/clean_nb45_std1.5.ply"
}
],
"gt_summary": " Post Processing, Post Processing \n Please process the mesh at {input_file_path} to create a point cloud. Use a sampling rate of 10000 points and write the file to ./outputs/example_mesh_from_glb_10000pts.ply.\nThen, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 45 neighbors and a std_ratio of 1.5. Save the clean cloud to ./outputs/clean_nb45_std1.5.ply. \n Input 1:./data/source/example_mesh.glb; Input 2:./outputs/example_mesh_from_glb_10000pts.ply; \n Step 1: { number_of_points: 10000; Output Path: ./outputs/example_mesh_from_glb_10000pts.ply }; Step 2: { nb_neighbors: 45, std_ratio: 1.5, Output Path: ./outputs/clean_nb45_std1.5.ply } ",
"code": "self.executor.post_processing(input_path='./data/source/example_mesh.glb', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_glb_10000pts.ply', parameters={\"number_of_points\": 10000})\nself.executor.post_processing(input_path='./outputs/example_mesh_from_glb_10000pts.ply', process_type='outliers_removal', output_path='./outputs/clean_nb45_std1.5.ply', parameters={\"nb_neighbors\": 45, \"std_ratio\": 1.5})",
"id": 77,
"ambiguous": "Do something with {input_file_path}.\nThen, Fix the noise in the output of the previous step.",
"missing_info": "Please process the mesh at a file to create a point cloud. Use a sampling rate of a number of points and write the file to somewhere.\nThen, I have a noisy point cloud at a location. Please apply statistical outlier removal using a certain number of neighbors and a std_ratio of a value. Save the clean cloud to somewhere.",
"mistake": "Please process the mesh at {input_file_path} to create a point cloud. Use a sampling rate of 10000 points and write the file to ./outputs/example_mesh_from_glb_10000pts.ply.\nThen, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 45 neighbors and a std_ratio of 1.5. Save the clean cloud to ./outputs/clean_nb45_std1.5.ply. Ensure the output format is \".pdf\"",
"mistake_type": "output_format"
},
{
"complete": "Take the mesh at {input_file_path} and generate a point cloud with 15000 samples. Store the output at ./outputs/example_mesh_from_obj_15000pts.ply.\nThen, help me filter the noise in the output path of the previous step. Use statistical removal with 35 neighbors and 2.5 std_ratio. The output should be saved at ./outputs/clean_nb35_std2.5.ply.",
"file_path": "./data/source/example_mesh.obj",
"ground_truth": [
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 15000,
"output_path": "./outputs/example_mesh_from_obj_15000pts.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 35,
"std_ratio": 2.5,
"output_path": "./outputs/clean_nb35_std2.5.ply"
}
],
"gt_summary": " Post Processing, Post Processing \n Take the mesh at {input_file_path} and generate a point cloud with 15000 samples. Store the output at ./outputs/example_mesh_from_obj_15000pts.ply.\nThen, help me filter the noise in the output path of the previous step. Use statistical removal with 35 neighbors and 2.5 std_ratio. The output should be saved at ./outputs/clean_nb35_std2.5.ply. \n Input 1:./data/source/example_mesh.obj; Input 2:./outputs/example_mesh_from_obj_15000pts.ply; \n Step 1: { number_of_points: 15000; Output Path: ./outputs/example_mesh_from_obj_15000pts.ply }; Step 2: { nb_neighbors: 35, std_ratio: 2.5, Output Path: ./outputs/clean_nb35_std2.5.ply } ",
"code": "self.executor.post_processing(input_path='./data/source/example_mesh.obj', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_obj_15000pts.ply', parameters={\"number_of_points\": 15000})\nself.executor.post_processing(input_path='./outputs/example_mesh_from_obj_15000pts.ply', process_type='outliers_removal', output_path='./outputs/clean_nb35_std2.5.ply', parameters={\"nb_neighbors\": 35, \"std_ratio\": 2.5})",
"id": 78,
"ambiguous": "Convert the file at {input_file_path}.\nThen, Clean the point cloud at the output of the previous step.",
"missing_info": "Please process the mesh at a file to create a point cloud. Use a sampling rate of a number of points and write the file to somewhere.\nThen, Perform post-processing on a file to remove noise. Use a neighbor count of a number and a std_ratio of a value, then save to somewhere.",
"mistake": "Take the mesh at {input_file_path} and generate a point cloud with 15000 samples. Store the output at ./outputs/example_mesh_from_obj_15000pts.ply.\nThen, help me filter the noise in the output path of the previous step. Use statistical removal with 35 neighbors and 2.5 std_ratio. The output should be saved at ./outputs/clean_nb35_std2.5.ply. Ensure the output format is \".pdf\"",
"mistake_type": "output_format"
},
{
"complete": "I need to sample 10000 points uniformly from the 3D model at {input_file_path}. Output the processed ply file to ./outputs/example_mesh_from_glb_10000pts.ply.\nThen, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 45 and a std_ratio of 1.8, then save to ./outputs/clean_nb45_std1.8.ply.",
"file_path": "./data/source/example_mesh.glb",
"ground_truth": [
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 10000,
"output_path": "./outputs/example_mesh_from_glb_10000pts.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 45,
"std_ratio": 1.8,
"output_path": "./outputs/clean_nb45_std1.8.ply"
}
],
"gt_summary": " Post Processing, Post Processing \n I need to sample 10000 points uniformly from the 3D model at {input_file_path}. Output the processed ply file to ./outputs/example_mesh_from_glb_10000pts.ply.\nThen, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 45 and a std_ratio of 1.8, then save to ./outputs/clean_nb45_std1.8.ply. \n Input 1:./data/source/example_mesh.glb; Input 2:./outputs/example_mesh_from_glb_10000pts.ply; \n Step 1: { number_of_points: 10000; Output Path: ./outputs/example_mesh_from_glb_10000pts.ply }; Step 2: { nb_neighbors: 45, std_ratio: 1.8, Output Path: ./outputs/clean_nb45_std1.8.ply } ",
"code": "self.executor.post_processing(input_path='./data/source/example_mesh.glb', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_glb_10000pts.ply', parameters={\"number_of_points\": 10000})\nself.executor.post_processing(input_path='./outputs/example_mesh_from_glb_10000pts.ply', process_type='outliers_removal', output_path='./outputs/clean_nb45_std1.8.ply', parameters={\"nb_neighbors\": 45, \"std_ratio\": 1.8})",
"id": 79,
"ambiguous": "Convert the file at {input_file_path}.\nThen, Fix the noise in the output of the previous step.",
"missing_info": "Convert the mesh file at a location into a point cloud by sampling a certain number of points. Save the result to somewhere.\nThen, Clean the point cloud file at a path. I want to remove outliers with nb_neighbors set to a number and standard deviation ratio set to a value. Output result to a location.",
"mistake": "I need to sample 10000 points uniformly from the 3D model at {input_file_path}. Output the processed ply file to ./outputs/example_mesh_from_glb_10000pts.ply.\nThen, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 45 and a std_ratio of 1.8, then save to ./outputs/clean_nb45_std1.8.ply. Ensure the output format is \".pdf\"",
"mistake_type": "output_format"
},
{
"complete": "Take the mesh at {input_file_path} and generate a point cloud with 35000 samples. Store the output at ./outputs/example_mesh_from_glb_35000pts.ply.\nThen, help me filter the noise in the output path of the previous step. Use statistical removal with 40 neighbors and 0.8 std_ratio. The output should be saved at ./outputs/clean_nb40_std0.8.ply.",
"file_path": "./data/source/example_mesh.glb",
"ground_truth": [
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 35000,
"output_path": "./outputs/example_mesh_from_glb_35000pts.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 40,
"std_ratio": 0.8,
"output_path": "./outputs/clean_nb40_std0.8.ply"
}
],
"gt_summary": " Post Processing, Post Processing \n Take the mesh at {input_file_path} and generate a point cloud with 35000 samples. Store the output at ./outputs/example_mesh_from_glb_35000pts.ply.\nThen, help me filter the noise in the output path of the previous step. Use statistical removal with 40 neighbors and 0.8 std_ratio. The output should be saved at ./outputs/clean_nb40_std0.8.ply. \n Input 1:./data/source/example_mesh.glb; Input 2:./outputs/example_mesh_from_glb_35000pts.ply; \n Step 1: { number_of_points: 35000; Output Path: ./outputs/example_mesh_from_glb_35000pts.ply }; Step 2: { nb_neighbors: 40, std_ratio: 0.8, Output Path: ./outputs/clean_nb40_std0.8.ply } ",
"code": "self.executor.post_processing(input_path='./data/source/example_mesh.glb', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_glb_35000pts.ply', parameters={\"number_of_points\": 35000})\nself.executor.post_processing(input_path='./outputs/example_mesh_from_glb_35000pts.ply', process_type='outliers_removal', output_path='./outputs/clean_nb40_std0.8.ply', parameters={\"nb_neighbors\": 40, \"std_ratio\": 0.8})",
"id": 80,
"ambiguous": "Convert the file at {input_file_path}.\nThen, Clean the point cloud at the output of the previous step.",
"missing_info": "Perform mesh-to-pointcloud conversion on the mesh at a file. Set the target number of points to a number and save as somewhere.\nThen, I have a noisy point cloud at a location. Please apply statistical outlier removal using a certain number of neighbors and a std_ratio of a value. Save the clean cloud to somewhere.",
"mistake": "Take the mesh at {input_file_path} and generate a point cloud with 35000 samples. Store the output at ./outputs/example_mesh_from_glb_35000pts.ply.\nThen, help me filter the noise in the output path of the previous step. Use statistical removal with 40 neighbors and 0.8 std_ratio. The output should be saved at ./outputs/clean_nb40_std0.8.ply. Ensure the output format is \".pdf\"",
"mistake_type": "output_format"
},
{
"complete": "Perform mesh-to-pointcloud conversion on the mesh at {input_file_path}. Set the target number of points to 15000 and save as ./outputs/example_mesh_from_glb_15000pts.ply.\nThen, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 25 and a std_ratio of 2.3, then save to ./outputs/clean_nb25_std2.3.ply.",
"file_path": "./data/source/example_mesh.glb",
"ground_truth": [
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 15000,
"output_path": "./outputs/example_mesh_from_glb_15000pts.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 25,
"std_ratio": 2.3,
"output_path": "./outputs/clean_nb25_std2.3.ply"
}
],
"gt_summary": " Post Processing, Post Processing \n Perform mesh-to-pointcloud conversion on the mesh at {input_file_path}. Set the target number of points to 15000 and save as ./outputs/example_mesh_from_glb_15000pts.ply.\nThen, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 25 and a std_ratio of 2.3, then save to ./outputs/clean_nb25_std2.3.ply. \n Input 1:./data/source/example_mesh.glb; Input 2:./outputs/example_mesh_from_glb_15000pts.ply; \n Step 1: { number_of_points: 15000; Output Path: ./outputs/example_mesh_from_glb_15000pts.ply }; Step 2: { nb_neighbors: 25, std_ratio: 2.3, Output Path: ./outputs/clean_nb25_std2.3.ply } ",
"code": "self.executor.post_processing(input_path='./data/source/example_mesh.glb', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_glb_15000pts.ply', parameters={\"number_of_points\": 15000})\nself.executor.post_processing(input_path='./outputs/example_mesh_from_glb_15000pts.ply', process_type='outliers_removal', output_path='./outputs/clean_nb25_std2.3.ply', parameters={\"nb_neighbors\": 25, \"std_ratio\": 2.3})",
"id": 81,
"ambiguous": "Process {input_file_path}.\nThen, Clean the point cloud at the output of the previous step.",
"missing_info": "Convert the mesh file at a location into a point cloud by sampling a certain number of points. Save the result to somewhere.\nThen, Clean the point cloud file at a path. I want to remove outliers with nb_neighbors set to a number and standard deviation ratio set to a value. Output result to a location.",
"mistake": "Perform mesh-to-pointcloud conversion on the mesh at {input_file_path}. Set the target number of points to 15000 and save as ./outputs/example_mesh_from_glb_15000pts.ply.\nThen, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 25 and a std_ratio of 2.3, then save to ./outputs/clean_nb25_std2.3.ply. Ensure the output format is \".pdf\"",
"mistake_type": "output_format"
},
{
"complete": "Convert the mesh file at {input_file_path} into a point cloud by sampling 25000 points. Save the result to ./outputs/example_mesh_from_glb_25000pts.ply.\nThen, help me filter the noise in the output path of the previous step. Use statistical removal with 45 neighbors and 2.3 std_ratio. The output should be saved at ./outputs/clean_nb45_std2.3.ply.",
"file_path": "./data/source/example_mesh.glb",
"ground_truth": [
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 25000,
"output_path": "./outputs/example_mesh_from_glb_25000pts.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 45,
"std_ratio": 2.3,
"output_path": "./outputs/clean_nb45_std2.3.ply"
}
],
"gt_summary": " Post Processing, Post Processing \n Convert the mesh file at {input_file_path} into a point cloud by sampling 25000 points. Save the result to ./outputs/example_mesh_from_glb_25000pts.ply.\nThen, help me filter the noise in the output path of the previous step. Use statistical removal with 45 neighbors and 2.3 std_ratio. The output should be saved at ./outputs/clean_nb45_std2.3.ply. \n Input 1:./data/source/example_mesh.glb; Input 2:./outputs/example_mesh_from_glb_25000pts.ply; \n Step 1: { number_of_points: 25000; Output Path: ./outputs/example_mesh_from_glb_25000pts.ply }; Step 2: { nb_neighbors: 45, std_ratio: 2.3, Output Path: ./outputs/clean_nb45_std2.3.ply } ",
"code": "self.executor.post_processing(input_path='./data/source/example_mesh.glb', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_glb_25000pts.ply', parameters={\"number_of_points\": 25000})\nself.executor.post_processing(input_path='./outputs/example_mesh_from_glb_25000pts.ply', process_type='outliers_removal', output_path='./outputs/clean_nb45_std2.3.ply', parameters={\"nb_neighbors\": 45, \"std_ratio\": 2.3})",
"id": 82,
"ambiguous": "Do something with {input_file_path}.\nThen, Do something with the output of the previous step.",
"missing_info": "Perform mesh-to-pointcloud conversion on the mesh at a file. Set the target number of points to a number and save as somewhere.\nThen, Perform post-processing on a file to remove noise. Use a neighbor count of a number and a std_ratio of a value, then save to somewhere.",
"mistake": "Convert the mesh file at {input_file_path} into a point cloud by sampling 25000 points. Save the result to ./outputs/example_mesh_from_glb_25000pts.ply.\nThen, help me filter the noise in the output path of the previous step. Use statistical removal with 45 neighbors and 2.3 std_ratio. The output should be saved at ./outputs/clean_nb45_std2.3.ply. Ensure the output format is \".pdf\"",
"mistake_type": "output_format"
},
{
"complete": "Perform mesh-to-pointcloud conversion on the mesh at {input_file_path}. Set the target number of points to 35000 and save as ./outputs/example_mesh_from_obj_35000pts.ply.\nThen, clean the point cloud file the output path of the previous step. I want to remove outliers with nb_neighbors set to 40 and standard deviation ratio set to 1.2. Output result to ./outputs/clean_nb40_std1.2.ply.",
"file_path": "./data/source/example_mesh.obj",
"ground_truth": [
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 35000,
"output_path": "./outputs/example_mesh_from_obj_35000pts.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 40,
"std_ratio": 1.2,
"output_path": "./outputs/clean_nb40_std1.2.ply"
}
],
"gt_summary": " Post Processing, Post Processing \n Perform mesh-to-pointcloud conversion on the mesh at {input_file_path}. Set the target number of points to 35000 and save as ./outputs/example_mesh_from_obj_35000pts.ply.\nThen, clean the point cloud file the output path of the previous step. I want to remove outliers with nb_neighbors set to 40 and standard deviation ratio set to 1.2. Output result to ./outputs/clean_nb40_std1.2.ply. \n Input 1:./data/source/example_mesh.obj; Input 2:./outputs/example_mesh_from_obj_35000pts.ply; \n Step 1: { number_of_points: 35000; Output Path: ./outputs/example_mesh_from_obj_35000pts.ply }; Step 2: { nb_neighbors: 40, std_ratio: 1.2, Output Path: ./outputs/clean_nb40_std1.2.ply } ",
"code": "self.executor.post_processing(input_path='./data/source/example_mesh.obj', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_obj_35000pts.ply', parameters={\"number_of_points\": 35000})\nself.executor.post_processing(input_path='./outputs/example_mesh_from_obj_35000pts.ply', process_type='outliers_removal', output_path='./outputs/clean_nb40_std1.2.ply', parameters={\"nb_neighbors\": 40, \"std_ratio\": 1.2})",
"id": 83,
"ambiguous": "Transform {input_file_path}.\nThen, Fix the noise in the output of the previous step.",
"missing_info": "Please process the mesh at a file to create a point cloud. Use a sampling rate of a number of points and write the file to somewhere.\nThen, Clean the point cloud file at a path. I want to remove outliers with nb_neighbors set to a number and standard deviation ratio set to a value. Output result to a location.",
"mistake": "Perform mesh-to-pointcloud conversion on the mesh at {input_file_path}. Set the target number of points to 35000 and save as ./outputs/example_mesh_from_obj_35000pts.ply.\nThen, clean the point cloud file the output path of the previous step. I want to remove outliers with nb_neighbors set to 40 and standard deviation ratio set to 1.2. Output result to ./outputs/clean_nb40_std1.2.ply. Ensure the output format is \".pdf\"",
"mistake_type": "output_format"
},
{
"complete": "Convert the mesh file at {input_file_path} into a point cloud by sampling 45000 points. Save the result to ./outputs/example_mesh_from_ply_45000pts.ply.\nThen, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 20 neighbors and a std_ratio of 1.7. Save the clean cloud to ./outputs/clean_nb20_std1.7.ply.",
"file_path": "./data/source/example_mesh.ply",
"ground_truth": [
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 45000,
"output_path": "./outputs/example_mesh_from_ply_45000pts.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 20,
"std_ratio": 1.7,
"output_path": "./outputs/clean_nb20_std1.7.ply"
}
],
"gt_summary": " Post Processing, Post Processing \n Convert the mesh file at {input_file_path} into a point cloud by sampling 45000 points. Save the result to ./outputs/example_mesh_from_ply_45000pts.ply.\nThen, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 20 neighbors and a std_ratio of 1.7. Save the clean cloud to ./outputs/clean_nb20_std1.7.ply. \n Input 1:./data/source/example_mesh.ply; Input 2:./outputs/example_mesh_from_ply_45000pts.ply; \n Step 1: { number_of_points: 45000; Output Path: ./outputs/example_mesh_from_ply_45000pts.ply }; Step 2: { nb_neighbors: 20, std_ratio: 1.7, Output Path: ./outputs/clean_nb20_std1.7.ply } ",
"code": "self.executor.post_processing(input_path='./data/source/example_mesh.ply', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_ply_45000pts.ply', parameters={\"number_of_points\": 45000})\nself.executor.post_processing(input_path='./outputs/example_mesh_from_ply_45000pts.ply', process_type='outliers_removal', output_path='./outputs/clean_nb20_std1.7.ply', parameters={\"nb_neighbors\": 20, \"std_ratio\": 1.7})",
"id": 84,
"ambiguous": "Do something with {input_file_path}.\nThen, Process the output of the previous step.",
"missing_info": "Perform mesh-to-pointcloud conversion on the mesh at a file. Set the target number of points to a number and save as somewhere.\nThen, Clean the point cloud file at a path. I want to remove outliers with nb_neighbors set to a number and standard deviation ratio set to a value. Output result to a location.",
"mistake": "Convert the mesh file at {input_file_path} into a point cloud by sampling 45000 points. Save the result to ./outputs/example_mesh_from_ply_45000pts.ply.\nThen, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 20 neighbors and a std_ratio of 1.7. Save the clean cloud to ./outputs/clean_nb20_std1.7.ply. Ensure the output format is \".pdf\"",
"mistake_type": "output_format"
},
{
"complete": "Generate a 3D model of a blue poison-dart frog sitting on a water lily. Ensure the result is high quality. Save the output to a glb file. It should be a high fidelity mesh. The result should be saved in ./outputs/a_blue_poison-dart_frog_sitting_on_a_water_lily.glb.\nThen, i need to sample 50000 points uniformly from the 3D model at the output path of the previous step. Output the processed ply file to ./outputs/example_mesh_from_obj_50000pts.ply.\nFinally, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 30 and a std_ratio of 1.6, then save to ./outputs/clean_nb30_std1.6.ply.",
"file_path": null,
"ground_truth": [
{
"task": "Text to 3D Generation",
"quality": "High quality",
"speed": null,
"usage": "object",
"output_format": "glb",
"output_path": "./outputs/a_blue_poison-dart_frog_sitting_on_a_water_lily.glb",
"extra_info": "High fidelity mesh"
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 50000,
"output_path": "./outputs/example_mesh_from_obj_50000pts.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 30,
"std_ratio": 1.6,
"output_path": "./outputs/clean_nb30_std1.6.ply"
}
],
"gt_summary": " Text to 3D Generation, Post Processing, Post Processing \n Generate a 3D model of a blue poison-dart frog sitting on a water lily. Ensure the result is high quality. Save the output to a glb file. It should be a high fidelity mesh. The result should be saved in ./outputs/a_blue_poison-dart_frog_sitting_on_a_water_lily.glb.\nThen, i need to sample 50000 points uniformly from the 3D model at the output path of the previous step. Output the processed ply file to ./outputs/example_mesh_from_obj_50000pts.ply.\nFinally, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 30 and a std_ratio of 1.6, then save to ./outputs/clean_nb30_std1.6.ply. \n Input 1:None; Input 2:./outputs/a_blue_poison-dart_frog_sitting_on_a_water_lily.glb; Input 3:./outputs/example_mesh_from_obj_50000pts.ply; \n Step 1: { Constraints: [quality: High quality, output_format: glb, features: High fidelity mesh]; Output Path: ./outputs/a_blue_poison-dart_frog_sitting_on_a_water_lily.glb }; Step 2: { number_of_points: 50000; Output Path: ./outputs/example_mesh_from_obj_50000pts.ply }; Step 3: { nb_neighbors: 30, std_ratio: 1.6, Output Path: ./outputs/clean_nb30_std1.6.ply } ",
"code": "self.executor.text_to_3d_generation(prompt=\"Generate a 3D model of a blue poison-dart frog sitting on a water lily\", model=\"trellis\", output_path=\"./outputs/a_blue_poison-dart_frog_sitting_on_a_water_lily.glb\")\nself.executor.post_processing(input_path='./outputs/a_blue_poison-dart_frog_sitting_on_a_water_lily.glb', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_obj_50000pts.ply', parameters={\"number_of_points\": 50000})\nself.executor.post_processing(input_path='./outputs/example_mesh_from_obj_50000pts.ply', process_type='outliers_removal', output_path='./outputs/clean_nb30_std1.6.ply', parameters={\"nb_neighbors\": 30, \"std_ratio\": 1.6})",
"id": 85,
"ambiguous": "Create something.\nThen, Transform the output of the previous step.\nFinally, Process the output of the previous step.",
"missing_info": "Generate something. Ensure the result quality is defined. Export format: 3D model. It should have specific attributes. Save the output to somewhere.\nThen, Convert the mesh file at a location into a point cloud by sampling a certain number of points. Save the result to somewhere.\nFinally, Help me filter the noise in a location. Use statistical removal with certain neighbors and a std_ratio. The output should be saved at a path.",
"mistake": "Generate a 3D model of a blue poison-dart frog sitting on a water lily. Ensure the result is high quality. Save the output to a glb file. It should be a high fidelity mesh. The result should be saved in ./outputs/a_blue_poison-dart_frog_sitting_on_a_water_lily.glb.\nThen, i need to sample 50000 points uniformly from the 3D model at the output path of the previous step. Output the processed ply file to ./outputs/example_mesh_from_obj_50000pts.ply.\nFinally, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 30 and a std_ratio of 1.6, then save to ./outputs/clean_nb30_std1.6.ply. Ensure low. Ensure low quality.",
"mistake_type": "quality_speed"
},
{
"complete": "Generate a 3D model of a glass 20-sided die. I need a glb file. Save the output to ./outputs/a_glass_20-sided_die.glb.\nThen, perform mesh-to-pointcloud conversion on the mesh at the output path of the previous step. Set the target number of points to 25000 and save as ./outputs/example_mesh_from_obj_25000pts.ply.\nFinally, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 25 and a std_ratio of 2.2, then save to ./outputs/clean_nb25_std2.2.ply.",
"file_path": null,
"ground_truth": [
{
"task": "Text to 3D Generation",
"quality": null,
"speed": null,
"usage": "object",
"output_format": "glb",
"output_path": "./outputs/a_glass_20-sided_die.glb",
"extra_info": null
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 25000,
"output_path": "./outputs/example_mesh_from_obj_25000pts.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 25,
"std_ratio": 2.2,
"output_path": "./outputs/clean_nb25_std2.2.ply"
}
],
"gt_summary": " Text to 3D Generation, Post Processing, Post Processing \n Generate a 3D model of a glass 20-sided die. I need a glb file. Save the output to ./outputs/a_glass_20-sided_die.glb.\nThen, perform mesh-to-pointcloud conversion on the mesh at the output path of the previous step. Set the target number of points to 25000 and save as ./outputs/example_mesh_from_obj_25000pts.ply.\nFinally, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 25 and a std_ratio of 2.2, then save to ./outputs/clean_nb25_std2.2.ply. \n Input 1:None; Input 2:./outputs/a_glass_20-sided_die.glb; Input 3:./outputs/example_mesh_from_obj_25000pts.ply; \n Step 1: { Constraints: [output_format: glb]; Output Path: ./outputs/a_glass_20-sided_die.glb }; Step 2: { number_of_points: 25000; Output Path: ./outputs/example_mesh_from_obj_25000pts.ply }; Step 3: { nb_neighbors: 25, std_ratio: 2.2, Output Path: ./outputs/clean_nb25_std2.2.ply } ",
"code": "self.executor.text_to_3d_generation(prompt=\"Generate a 3D model of a glass 20-sided die\", model=\"trellis\", output_path=\"./outputs/a_glass_20-sided_die.glb\")\nself.executor.post_processing(input_path='./outputs/a_glass_20-sided_die.glb', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_obj_25000pts.ply', parameters={\"number_of_points\": 25000})\nself.executor.post_processing(input_path='./outputs/example_mesh_from_obj_25000pts.ply', process_type='outliers_removal', output_path='./outputs/clean_nb25_std2.2.ply', parameters={\"nb_neighbors\": 25, \"std_ratio\": 2.2})",
"id": 86,
"ambiguous": "Create something.\nThen, Convert the file at the output of the previous step.\nFinally, Clean the point cloud at the output of the previous step.",
"missing_info": "Generate something. Save the output to a file. Save the output to somewhere.\nThen, Perform mesh-to-pointcloud conversion on the mesh at a file. Set the target number of points to a number and save as somewhere.\nFinally, Clean the point cloud file at a path. I want to remove outliers with nb_neighbors set to a number and standard deviation ratio set to a value. Output result to a location.",
"mistake": "Generate a 3D model of a glass 20-sided die. I need a glb file. Save the output to ./outputs/a_glass_20-sided_die.glb.\nThen, perform mesh-to-pointcloud conversion on the mesh at the output path of the previous step. Set the target number of points to 25000 and save as ./outputs/example_mesh_from_obj_25000pts.ply.\nFinally, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 25 and a std_ratio of 2.2, then save to ./outputs/clean_nb25_std2.2.ply. Ensure the output format is \".pdf\"",
"mistake_type": "output_format"
},
{
"complete": "Generate a 3D model of a banana peeling itself. The model quality should be high quality. The result should be saved in ./outputs/a_banana_peeling_itself.ply.\nThen, please process the mesh at the output path of the previous step to create a point cloud. Use a sampling rate of 35000 points and write the file to ./outputs/example_mesh_from_ply_35000pts.ply.\nFinally, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 25 neighbors and a std_ratio of 1.9. Save the clean cloud to ./outputs/clean_nb25_std1.9.ply.",
"file_path": null,
"ground_truth": [
{
"task": "Text to 3D Generation",
"quality": "High quality",
"speed": null,
"usage": "object",
"output_format": "ply",
"output_path": "./outputs/a_banana_peeling_itself.ply",
"extra_info": null
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 35000,
"output_path": "./outputs/example_mesh_from_ply_35000pts.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 25,
"std_ratio": 1.9,
"output_path": "./outputs/clean_nb25_std1.9.ply"
}
],
"gt_summary": " Text to 3D Generation, Post Processing, Post Processing \n Generate a 3D model of a banana peeling itself. The model quality should be high quality. The result should be saved in ./outputs/a_banana_peeling_itself.ply.\nThen, please process the mesh at the output path of the previous step to create a point cloud. Use a sampling rate of 35000 points and write the file to ./outputs/example_mesh_from_ply_35000pts.ply.\nFinally, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 25 neighbors and a std_ratio of 1.9. Save the clean cloud to ./outputs/clean_nb25_std1.9.ply. \n Input 1:None; Input 2:./outputs/a_banana_peeling_itself.ply; Input 3:./outputs/example_mesh_from_ply_35000pts.ply; \n Step 1: { Constraints: [quality: High quality]; Output Path: ./outputs/a_banana_peeling_itself.ply }; Step 2: { number_of_points: 35000; Output Path: ./outputs/example_mesh_from_ply_35000pts.ply }; Step 3: { nb_neighbors: 25, std_ratio: 1.9, Output Path: ./outputs/clean_nb25_std1.9.ply } ",
"code": "self.executor.text_to_3d_generation(prompt=\"Generate a 3D model of a banana peeling itself\", model=\"trellis\", output_path=\"./outputs/a_banana_peeling_itself.ply\")\nself.executor.post_processing(input_path='./outputs/a_banana_peeling_itself.ply', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_ply_35000pts.ply', parameters={\"number_of_points\": 35000})\nself.executor.post_processing(input_path='./outputs/example_mesh_from_ply_35000pts.ply', process_type='outliers_removal', output_path='./outputs/clean_nb25_std1.9.ply', parameters={\"nb_neighbors\": 25, \"std_ratio\": 1.9})",
"id": 87,
"ambiguous": "I want a 3D asset.\nThen, Process the output of the previous step.\nFinally, Process the output of the previous step.",
"missing_info": "Generate something. The model quality should be specified. I need a file. Store the generated model.\nThen, Please process the mesh at a file to create a point cloud. Use a sampling rate of a number of points and write the file to somewhere.\nFinally, Help me filter the noise in a location. Use statistical removal with certain neighbors and a std_ratio. The output should be saved at a path.",
"mistake": "Generate a 3D model of a banana peeling itself. The model quality should be high quality. The result should be saved in ./outputs/a_banana_peeling_itself.ply.\nThen, please process the mesh at the output path of the previous step to create a point cloud. Use a sampling rate of 35000 points and write the file to ./outputs/example_mesh_from_ply_35000pts.ply.\nFinally, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 25 neighbors and a std_ratio of 1.9. Save the clean cloud to ./outputs/clean_nb25_std1.9.ply. Ensure low quality.",
"mistake_type": "quality_speed"
},
{
"complete": "Generate a 3D model of a beagle in a detective's outfit. Ensure the result is high quality. I need it generated at medium speed. Save the output to a glb file. The result should be saved in ./outputs/a_beagle_in_a_detectives_outfit.glb.\nThen, perform mesh-to-pointcloud conversion on the mesh at the output path of the previous step. Set the target number of points to 30000 and save as ./outputs/example_mesh_from_ply_30000pts.ply.\nFinally, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 45 neighbors and a std_ratio of 0.8. Save the clean cloud to ./outputs/clean_nb45_std0.8.ply.",
"file_path": null,
"ground_truth": [
{
"task": "Text to 3D Generation",
"quality": "High quality",
"speed": "Medium speed",
"usage": "object",
"output_format": "glb",
"output_path": "./outputs/a_beagle_in_a_detectives_outfit.glb",
"extra_info": null
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 30000,
"output_path": "./outputs/example_mesh_from_ply_30000pts.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 45,
"std_ratio": 0.8,
"output_path": "./outputs/clean_nb45_std0.8.ply"
}
],
"gt_summary": " Text to 3D Generation, Post Processing, Post Processing \n Generate a 3D model of a beagle in a detective's outfit. Ensure the result is high quality. I need it generated at medium speed. Save the output to a glb file. The result should be saved in ./outputs/a_beagle_in_a_detectives_outfit.glb.\nThen, perform mesh-to-pointcloud conversion on the mesh at the output path of the previous step. Set the target number of points to 30000 and save as ./outputs/example_mesh_from_ply_30000pts.ply.\nFinally, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 45 neighbors and a std_ratio of 0.8. Save the clean cloud to ./outputs/clean_nb45_std0.8.ply. \n Input 1:None; Input 2:./outputs/a_beagle_in_a_detectives_outfit.glb; Input 3:./outputs/example_mesh_from_ply_30000pts.ply; \n Step 1: { Constraints: [quality: High quality, speed: Medium speed, output_format: glb]; Output Path: ./outputs/a_beagle_in_a_detectives_outfit.glb }; Step 2: { number_of_points: 30000; Output Path: ./outputs/example_mesh_from_ply_30000pts.ply }; Step 3: { nb_neighbors: 45, std_ratio: 0.8, Output Path: ./outputs/clean_nb45_std0.8.ply } ",
"code": "self.executor.text_to_3d_generation(prompt=\"Generate a 3D model of a beagle in a detective's outfit\", model=\"trellis\", output_path=\"./outputs/a_beagle_in_a_detectives_outfit.glb\")\nself.executor.post_processing(input_path='./outputs/a_beagle_in_a_detectives_outfit.glb', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_ply_30000pts.ply', parameters={\"number_of_points\": 30000})\nself.executor.post_processing(input_path='./outputs/example_mesh_from_ply_30000pts.ply', process_type='outliers_removal', output_path='./outputs/clean_nb45_std0.8.ply', parameters={\"nb_neighbors\": 45, \"std_ratio\": 0.8})",
"id": 88,
"ambiguous": "I want a 3D asset.\nThen, Transform the output of the previous step.\nFinally, Clean the point cloud at the output of the previous step.",
"missing_info": "Generate a 3D model. The model quality should be specified. The generation speed should be specified. I need a file. Save the output to somewhere.\nThen, Please process the mesh at a file to create a point cloud. Use a sampling rate of a number of points and write the file to somewhere.\nFinally, I have a noisy point cloud at a location. Please apply statistical outlier removal using a certain number of neighbors and a std_ratio of a value. Save the clean cloud to somewhere.",
"mistake": "Generate a 3D model of a beagle in a detective's outfit. Ensure the result is high quality. I need it generated at medium speed. Save the output to a glb file. The result should be saved in ./outputs/a_beagle_in_a_detectives_outfit.glb.\nThen, perform mesh-to-pointcloud conversion on the mesh at the output path of the previous step. Set the target number of points to 30000 and save as ./outputs/example_mesh_from_ply_30000pts.ply.\nFinally, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 45 neighbors and a std_ratio of 0.8. Save the clean cloud to ./outputs/clean_nb45_std0.8.ply. Ensure extremely high speed. Ensure low quality.",
"mistake_type": "quality_speed"
},
{
"complete": "Generate a 3D model of a brightly colored mushroom growing on a log. Ensure the result is low quality. Expect fast speed processing. I need a obj file. Consider that this is a lightweight model. Save the output to ./outputs/a_brightly_colored_mushroom_growing_on_a_log.obj.\nThen, convert the mesh file at the output path of the previous step into a point cloud by sampling 50000 points. Save the result to ./outputs/example_mesh_from_glb_50000pts.ply.\nFinally, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 40 neighbors and a std_ratio of 1.8. Save the clean cloud to ./outputs/clean_nb40_std1.8.ply.",
"file_path": null,
"ground_truth": [
{
"task": "Text to 3D Generation",
"quality": "Low quality",
"speed": "Fast speed",
"usage": "object",
"output_format": "obj",
"output_path": "./outputs/a_brightly_colored_mushroom_growing_on_a_log.obj",
"extra_info": "Lightweight model"
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 50000,
"output_path": "./outputs/example_mesh_from_glb_50000pts.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 40,
"std_ratio": 1.8,
"output_path": "./outputs/clean_nb40_std1.8.ply"
}
],
"gt_summary": " Text to 3D Generation, Post Processing, Post Processing \n Generate a 3D model of a brightly colored mushroom growing on a log. Ensure the result is low quality. Expect fast speed processing. I need a obj file. Consider that this is a lightweight model. Save the output to ./outputs/a_brightly_colored_mushroom_growing_on_a_log.obj.\nThen, convert the mesh file at the output path of the previous step into a point cloud by sampling 50000 points. Save the result to ./outputs/example_mesh_from_glb_50000pts.ply.\nFinally, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 40 neighbors and a std_ratio of 1.8. Save the clean cloud to ./outputs/clean_nb40_std1.8.ply. \n Input 1:None; Input 2:./outputs/a_brightly_colored_mushroom_growing_on_a_log.obj; Input 3:./outputs/example_mesh_from_glb_50000pts.ply; \n Step 1: { Constraints: [quality: Low quality, speed: Fast speed, output_format: obj, features: Lightweight model]; Output Path: ./outputs/a_brightly_colored_mushroom_growing_on_a_log.obj }; Step 2: { number_of_points: 50000; Output Path: ./outputs/example_mesh_from_glb_50000pts.ply }; Step 3: { nb_neighbors: 40, std_ratio: 1.8, Output Path: ./outputs/clean_nb40_std1.8.ply } ",
"code": "self.executor.text_to_3d_generation(prompt=\"Generate a 3D model of a brightly colored mushroom growing on a log\", model=\"triplaneturbo\", output_path=\"./outputs/a_brightly_colored_mushroom_growing_on_a_log.obj\")\nself.executor.post_processing(input_path='./outputs/a_brightly_colored_mushroom_growing_on_a_log.obj', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_glb_50000pts.ply', parameters={\"number_of_points\": 50000})\nself.executor.post_processing(input_path='./outputs/example_mesh_from_glb_50000pts.ply', process_type='outliers_removal', output_path='./outputs/clean_nb40_std1.8.ply', parameters={\"nb_neighbors\": 40, \"std_ratio\": 1.8})",
"id": 89,
"ambiguous": "Generate an object.\nThen, Convert the file at the output of the previous step.\nFinally, Process the output of the previous step.",
"missing_info": "Generate something. Ensure the result quality is defined. Expect appropriate processing speed. Save the output to a file. It should have specific attributes. Save the output to somewhere.\nThen, Convert the mesh file at a location into a point cloud by sampling a certain number of points. Save the result to somewhere.\nFinally, Perform post-processing on a file to remove noise. Use a neighbor count of a number and a std_ratio of a value, then save to somewhere.",
"mistake": "Generate a 3D model of a brightly colored mushroom growing on a log. Ensure the result is low quality. Expect fast speed processing. I need a obj file. Consider that this is a lightweight model. Save the output to ./outputs/a_brightly_colored_mushroom_growing_on_a_log.obj.\nThen, convert the mesh file at the output path of the previous step into a point cloud by sampling 50000 points. Save the result to ./outputs/example_mesh_from_glb_50000pts.ply.\nFinally, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 40 neighbors and a std_ratio of 1.8. Save the clean cloud to ./outputs/clean_nb40_std1.8.ply. Ensure heavy model. Ensure slow speed. Ensure high quality.",
"mistake_type": "quality_speed"
},
{
"complete": "Generate a 3D model of a glass 20-sided die. Ensure the result is low quality. I need it generated at fast speed. Store the generated model at ./outputs/a_glass_20-sided_die.obj.\nThen, convert the mesh file at the output path of the previous step into a point cloud by sampling 25000 points. Save the result to ./outputs/example_mesh_from_glb_25000pts.ply.\nFinally, help me filter the noise in the output path of the previous step. Use statistical removal with 25 neighbors and 0.9 std_ratio. The output should be saved at ./outputs/clean_nb25_std0.9.ply.",
"file_path": null,
"ground_truth": [
{
"task": "Text to 3D Generation",
"quality": "Low quality",
"speed": "Fast speed",
"usage": "object",
"output_format": "obj",
"output_path": "./outputs/a_glass_20-sided_die.obj",
"extra_info": null
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 25000,
"output_path": "./outputs/example_mesh_from_glb_25000pts.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 25,
"std_ratio": 0.9,
"output_path": "./outputs/clean_nb25_std0.9.ply"
}
],
"gt_summary": " Text to 3D Generation, Post Processing, Post Processing \n Generate a 3D model of a glass 20-sided die. Ensure the result is low quality. I need it generated at fast speed. Store the generated model at ./outputs/a_glass_20-sided_die.obj.\nThen, convert the mesh file at the output path of the previous step into a point cloud by sampling 25000 points. Save the result to ./outputs/example_mesh_from_glb_25000pts.ply.\nFinally, help me filter the noise in the output path of the previous step. Use statistical removal with 25 neighbors and 0.9 std_ratio. The output should be saved at ./outputs/clean_nb25_std0.9.ply. \n Input 1:None; Input 2:./outputs/a_glass_20-sided_die.obj; Input 3:./outputs/example_mesh_from_glb_25000pts.ply; \n Step 1: { Constraints: [quality: Low quality, speed: Fast speed]; Output Path: ./outputs/a_glass_20-sided_die.obj }; Step 2: { number_of_points: 25000; Output Path: ./outputs/example_mesh_from_glb_25000pts.ply }; Step 3: { nb_neighbors: 25, std_ratio: 0.9, Output Path: ./outputs/clean_nb25_std0.9.ply } ",
"code": "self.executor.text_to_3d_generation(prompt=\"Generate a 3D model of a glass 20-sided die\", model=\"triplaneturbo\", output_path=\"./outputs/a_glass_20-sided_die.obj\")\nself.executor.post_processing(input_path='./outputs/a_glass_20-sided_die.obj', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_glb_25000pts.ply', parameters={\"number_of_points\": 25000})\nself.executor.post_processing(input_path='./outputs/example_mesh_from_glb_25000pts.ply', process_type='outliers_removal', output_path='./outputs/clean_nb25_std0.9.ply', parameters={\"nb_neighbors\": 25, \"std_ratio\": 0.9})",
"id": 90,
"ambiguous": "Make a 3D thing.\nThen, Convert the file at the output of the previous step.\nFinally, Fix the noise in the output of the previous step.",
"missing_info": "Generate something. The model quality should be specified. I need it generated at a certain speed. Save the output to a file. The result should be saved in a location.\nThen, I need to sample a number of points uniformly from the 3D model at a path. Output the processed ply file to a location.\nFinally, I have a noisy point cloud at a location. Please apply statistical outlier removal using a certain number of neighbors and a std_ratio of a value. Save the clean cloud to somewhere.",
"mistake": "Generate a 3D model of a glass 20-sided die. Ensure the result is low quality. I need it generated at fast speed. Store the generated model at ./outputs/a_glass_20-sided_die.obj.\nThen, convert the mesh file at the output path of the previous step into a point cloud by sampling 25000 points. Save the result to ./outputs/example_mesh_from_glb_25000pts.ply.\nFinally, help me filter the noise in the output path of the previous step. Use statistical removal with 25 neighbors and 0.9 std_ratio. The output should be saved at ./outputs/clean_nb25_std0.9.ply. Ensure slow speed. Ensure high quality.",
"mistake_type": "quality_speed"
},
{
"complete": "Generate a 3D model of a bichon frise wearing academic regalia. I require low quality output. Consider that this is a lightweight model. Store the generated model at ./outputs/a_bichon_frise_wearing_academic_regalia.obj.\nThen, perform mesh-to-pointcloud conversion on the mesh at the output path of the previous step. Set the target number of points to 30000 and save as ./outputs/example_mesh_from_obj_30000pts.ply.\nFinally, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 45 neighbors and a std_ratio of 2.2. Save the clean cloud to ./outputs/clean_nb45_std2.2.ply.",
"file_path": null,
"ground_truth": [
{
"task": "Text to 3D Generation",
"quality": "Low quality",
"speed": null,
"usage": "object",
"output_format": "obj",
"output_path": "./outputs/a_bichon_frise_wearing_academic_regalia.obj",
"extra_info": "Lightweight model"
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 30000,
"output_path": "./outputs/example_mesh_from_obj_30000pts.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 45,
"std_ratio": 2.2,
"output_path": "./outputs/clean_nb45_std2.2.ply"
}
],
"gt_summary": " Text to 3D Generation, Post Processing, Post Processing \n Generate a 3D model of a bichon frise wearing academic regalia. I require low quality output. Consider that this is a lightweight model. Store the generated model at ./outputs/a_bichon_frise_wearing_academic_regalia.obj.\nThen, perform mesh-to-pointcloud conversion on the mesh at the output path of the previous step. Set the target number of points to 30000 and save as ./outputs/example_mesh_from_obj_30000pts.ply.\nFinally, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 45 neighbors and a std_ratio of 2.2. Save the clean cloud to ./outputs/clean_nb45_std2.2.ply. \n Input 1:None; Input 2:./outputs/a_bichon_frise_wearing_academic_regalia.obj; Input 3:./outputs/example_mesh_from_obj_30000pts.ply; \n Step 1: { Constraints: [quality: Low quality, features: Lightweight model]; Output Path: ./outputs/a_bichon_frise_wearing_academic_regalia.obj }; Step 2: { number_of_points: 30000; Output Path: ./outputs/example_mesh_from_obj_30000pts.ply }; Step 3: { nb_neighbors: 45, std_ratio: 2.2, Output Path: ./outputs/clean_nb45_std2.2.ply } ",
"code": "self.executor.text_to_3d_generation(prompt=\"Generate a 3D model of a bichon frise wearing academic regalia\", model=\"triplaneturbo\", output_path=\"./outputs/a_bichon_frise_wearing_academic_regalia.obj\")\nself.executor.post_processing(input_path='./outputs/a_bichon_frise_wearing_academic_regalia.obj', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_obj_30000pts.ply', parameters={\"number_of_points\": 30000})\nself.executor.post_processing(input_path='./outputs/example_mesh_from_obj_30000pts.ply', process_type='outliers_removal', output_path='./outputs/clean_nb45_std2.2.ply', parameters={\"nb_neighbors\": 45, \"std_ratio\": 2.2})",
"id": 91,
"ambiguous": "Create something.\nThen, Do something with the output of the previous step.\nFinally, Process the output of the previous step.",
"missing_info": "Generate something. Ensure the result quality is defined. Export format: 3D model. Consider that this has particular characteristics. The result should be saved in a location.\nThen, I need to sample a number of points uniformly from the 3D model at a path. Output the processed ply file to a location.\nFinally, Clean the point cloud file at a path. I want to remove outliers with nb_neighbors set to a number and standard deviation ratio set to a value. Output result to a location.",
"mistake": "Generate a 3D model of a bichon frise wearing academic regalia. I require low quality output. Consider that this is a lightweight model. Store the generated model at ./outputs/a_bichon_frise_wearing_academic_regalia.obj.\nThen, perform mesh-to-pointcloud conversion on the mesh at the output path of the previous step. Set the target number of points to 30000 and save as ./outputs/example_mesh_from_obj_30000pts.ply.\nFinally, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 45 neighbors and a std_ratio of 2.2. Save the clean cloud to ./outputs/clean_nb45_std2.2.ply. Ensure heavy model. Ensure high quality.",
"mistake_type": "quality_speed"
},
{
"complete": "Generate a 3D model of Sydney opera house, aerial view, zoomed out, in the style of a DSLR photo. I require medium quality output. The generation speed should be slow speed. I need a obj file. Consider that this is a mesh with detailed geometry. Save the output to ./outputs/Sydney_opera_house.obj.\nThen, take the mesh at the output path of the previous step and generate a point cloud with 30000 samples. Store the output at ./outputs/example_mesh_from_obj_30000pts.ply.\nFinally, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 50 and a std_ratio of 2.5, then save to ./outputs/clean_nb50_std2.5.ply.",
"file_path": null,
"ground_truth": [
{
"task": "Text to 3D Generation",
"quality": "Medium quality",
"speed": "Slow speed",
"usage": "scene",
"output_format": "obj",
"output_path": "./outputs/Sydney_opera_house.obj",
"extra_info": "Mesh with detailed geometry"
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 30000,
"output_path": "./outputs/example_mesh_from_obj_30000pts.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 50,
"std_ratio": 2.5,
"output_path": "./outputs/clean_nb50_std2.5.ply"
}
],
"gt_summary": " Text to 3D Generation, Post Processing, Post Processing \n Generate a 3D model of Sydney opera house, aerial view, zoomed out, in the style of a DSLR photo. I require medium quality output. The generation speed should be slow speed. I need a obj file. Consider that this is a mesh with detailed geometry. Save the output to ./outputs/Sydney_opera_house.obj.\nThen, take the mesh at the output path of the previous step and generate a point cloud with 30000 samples. Store the output at ./outputs/example_mesh_from_obj_30000pts.ply.\nFinally, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 50 and a std_ratio of 2.5, then save to ./outputs/clean_nb50_std2.5.ply. \n Input 1:None; Input 2:./outputs/Sydney_opera_house.obj; Input 3:./outputs/example_mesh_from_obj_30000pts.ply; \n Step 1: { Constraints: [quality: Medium quality, speed: Slow speed, output_format: obj, features: Mesh with detailed geometry]; Output Path: ./outputs/Sydney_opera_house.obj }; Step 2: { number_of_points: 30000; Output Path: ./outputs/example_mesh_from_obj_30000pts.ply }; Step 3: { nb_neighbors: 50, std_ratio: 2.5, Output Path: ./outputs/clean_nb50_std2.5.ply } ",
"code": "self.executor.text_to_3d_generation(prompt=\"Generate a 3D model of Sydney opera house, aerial view, zoomed out, in the style of a DSLR photo\", model=\"worldgen\", output_path=\"./outputs/Sydney_opera_house.obj\")\nself.executor.post_processing(input_path='./outputs/Sydney_opera_house.obj', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_obj_30000pts.ply', parameters={\"number_of_points\": 30000})\nself.executor.post_processing(input_path='./outputs/example_mesh_from_obj_30000pts.ply', process_type='outliers_removal', output_path='./outputs/clean_nb50_std2.5.ply', parameters={\"nb_neighbors\": 50, \"std_ratio\": 2.5})",
"id": 92,
"ambiguous": "Create something.\nThen, Transform the output of the previous step.\nFinally, Fix the noise in the output of the previous step.",
"missing_info": "Create a 3D asset. The model quality should be specified. I need it generated at a certain speed. Save the output to a file. Focus on certain features. The result should be saved in a location.\nThen, Perform mesh-to-pointcloud conversion on the mesh at a file. Set the target number of points to a number and save as somewhere.\nFinally, Help me filter the noise in a location. Use statistical removal with certain neighbors and a std_ratio. The output should be saved at a path.",
"mistake": "Generate a 3D model of Sydney opera house, aerial view, zoomed out, in the style of a DSLR photo. I require medium quality output. The generation speed should be slow speed. I need a obj file. Consider that this is a mesh with detailed geometry. Save the output to ./outputs/Sydney_opera_house.obj.\nThen, take the mesh at the output path of the previous step and generate a point cloud with 30000 samples. Store the output at ./outputs/example_mesh_from_obj_30000pts.ply.\nFinally, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 50 and a std_ratio of 2.5, then save to ./outputs/clean_nb50_std2.5.ply. Ensure mesh with simplified geometry. Ensure fast speed. Ensure extremely high quality.",
"mistake_type": "quality_speed"
},
{
"complete": "Generate a 3D model of Mount Fuji, aerial view, in the style of a DSLR photo. The model quality should be medium quality. Export format: glb. Consider that this is a mesh with detailed geometry. Store the generated model at ./outputs/Mount_Fuji.glb.\nThen, i need to sample 35000 points uniformly from the 3D model at the output path of the previous step. Output the processed ply file to ./outputs/example_mesh_from_glb_35000pts.ply.\nFinally, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 40 neighbors and a std_ratio of 1.3. Save the clean cloud to ./outputs/clean_nb40_std1.3.ply.",
"file_path": null,
"ground_truth": [
{
"task": "Text to 3D Generation",
"quality": "Medium quality",
"speed": null,
"usage": "scene",
"output_format": "glb",
"output_path": "./outputs/Mount_Fuji.glb",
"extra_info": "Mesh with detailed geometry"
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 35000,
"output_path": "./outputs/example_mesh_from_glb_35000pts.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 40,
"std_ratio": 1.3,
"output_path": "./outputs/clean_nb40_std1.3.ply"
}
],
"gt_summary": " Text to 3D Generation, Post Processing, Post Processing \n Generate a 3D model of Mount Fuji, aerial view, in the style of a DSLR photo. The model quality should be medium quality. Export format: glb. Consider that this is a mesh with detailed geometry. Store the generated model at ./outputs/Mount_Fuji.glb.\nThen, i need to sample 35000 points uniformly from the 3D model at the output path of the previous step. Output the processed ply file to ./outputs/example_mesh_from_glb_35000pts.ply.\nFinally, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 40 neighbors and a std_ratio of 1.3. Save the clean cloud to ./outputs/clean_nb40_std1.3.ply. \n Input 1:None; Input 2:./outputs/Mount_Fuji.glb; Input 3:./outputs/example_mesh_from_glb_35000pts.ply; \n Step 1: { Constraints: [quality: Medium quality, output_format: glb, features: Mesh with detailed geometry]; Output Path: ./outputs/Mount_Fuji.glb }; Step 2: { number_of_points: 35000; Output Path: ./outputs/example_mesh_from_glb_35000pts.ply }; Step 3: { nb_neighbors: 40, std_ratio: 1.3, Output Path: ./outputs/clean_nb40_std1.3.ply } ",
"code": "self.executor.text_to_3d_generation(prompt=\"Generate a 3D model of Mount Fuji, aerial view, in the style of a DSLR photo\", model=\"worldgen\", output_path=\"./outputs/Mount_Fuji.glb\")\nself.executor.post_processing(input_path='./outputs/Mount_Fuji.glb', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_glb_35000pts.ply', parameters={\"number_of_points\": 35000})\nself.executor.post_processing(input_path='./outputs/example_mesh_from_glb_35000pts.ply', process_type='outliers_removal', output_path='./outputs/clean_nb40_std1.3.ply', parameters={\"nb_neighbors\": 40, \"std_ratio\": 1.3})",
"id": 93,
"ambiguous": "Make a 3D thing.\nThen, Transform the output of the previous step.\nFinally, Fix the noise in the output of the previous step.",
"missing_info": "Create a 3D asset. I require a certain quality. Save the output to a file. Consider that this has particular characteristics. Save the output to somewhere.\nThen, Take the mesh at a location and generate a point cloud with a certain number of samples. Store the output at a path.\nFinally, I have a noisy point cloud at a location. Please apply statistical outlier removal using a certain number of neighbors and a std_ratio of a value. Save the clean cloud to somewhere.",
"mistake": "Generate a 3D model of Mount Fuji, aerial view, in the style of a DSLR photo. The model quality should be medium quality. Export format: glb. Consider that this is a mesh with detailed geometry. Store the generated model at ./outputs/Mount_Fuji.glb.\nThen, i need to sample 35000 points uniformly from the 3D model at the output path of the previous step. Output the processed ply file to ./outputs/example_mesh_from_glb_35000pts.ply.\nFinally, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 40 neighbors and a std_ratio of 1.3. Save the clean cloud to ./outputs/clean_nb40_std1.3.ply. Ensure mesh with simplified geometry. Ensure extremely high quality.",
"mistake_type": "quality_speed"
},
{
"complete": "Generate a 3D model of the leaning tower of Pisa, aerial view, in the style of a DSLR photo. Save the output to a glb file. The result should be saved in ./outputs/the_leaning_tower_of_Pisa.glb.\nThen, convert the mesh file at the output path of the previous step into a point cloud by sampling 10000 points. Save the result to ./outputs/example_mesh_from_obj_10000pts.ply.\nFinally, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 25 neighbors and a std_ratio of 0.9. Save the clean cloud to ./outputs/clean_nb25_std0.9.ply.",
"file_path": null,
"ground_truth": [
{
"task": "Text to 3D Generation",
"quality": null,
"speed": null,
"usage": "scene",
"output_format": "glb",
"output_path": "./outputs/the_leaning_tower_of_Pisa.glb",
"extra_info": null
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 10000,
"output_path": "./outputs/example_mesh_from_obj_10000pts.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 25,
"std_ratio": 0.9,
"output_path": "./outputs/clean_nb25_std0.9.ply"
}
],
"gt_summary": " Text to 3D Generation, Post Processing, Post Processing \n Generate a 3D model of the leaning tower of Pisa, aerial view, in the style of a DSLR photo. Save the output to a glb file. The result should be saved in ./outputs/the_leaning_tower_of_Pisa.glb.\nThen, convert the mesh file at the output path of the previous step into a point cloud by sampling 10000 points. Save the result to ./outputs/example_mesh_from_obj_10000pts.ply.\nFinally, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 25 neighbors and a std_ratio of 0.9. Save the clean cloud to ./outputs/clean_nb25_std0.9.ply. \n Input 1:None; Input 2:./outputs/the_leaning_tower_of_Pisa.glb; Input 3:./outputs/example_mesh_from_obj_10000pts.ply; \n Step 1: { Constraints: [output_format: glb]; Output Path: ./outputs/the_leaning_tower_of_Pisa.glb }; Step 2: { number_of_points: 10000; Output Path: ./outputs/example_mesh_from_obj_10000pts.ply }; Step 3: { nb_neighbors: 25, std_ratio: 0.9, Output Path: ./outputs/clean_nb25_std0.9.ply } ",
"code": "self.executor.text_to_3d_generation(prompt=\"Generate a 3D model of the leaning tower of Pisa, aerial view, in the style of a DSLR photo\", model=\"worldgen\", output_path=\"./outputs/the_leaning_tower_of_Pisa.glb\")\nself.executor.post_processing(input_path='./outputs/the_leaning_tower_of_Pisa.glb', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_obj_10000pts.ply', parameters={\"number_of_points\": 10000})\nself.executor.post_processing(input_path='./outputs/example_mesh_from_obj_10000pts.ply', process_type='outliers_removal', output_path='./outputs/clean_nb25_std0.9.ply', parameters={\"nb_neighbors\": 25, \"std_ratio\": 0.9})",
"id": 94,
"ambiguous": "Generate an object.\nThen, Transform the output of the previous step.\nFinally, Clean the point cloud at the output of the previous step.",
"missing_info": "Create an object. Save the output to a file. Save the output to somewhere.\nThen, Take the mesh at a location and generate a point cloud with a certain number of samples. Store the output at a path.\nFinally, Perform post-processing on a file to remove noise. Use a neighbor count of a number and a std_ratio of a value, then save to somewhere.",
"mistake": "Generate a 3D model of the leaning tower of Pisa, aerial view, in the style of a DSLR photo. Save the output to a glb file. The result should be saved in ./outputs/the_leaning_tower_of_Pisa.glb.\nThen, convert the mesh file at the output path of the previous step into a point cloud by sampling 10000 points. Save the result to ./outputs/example_mesh_from_obj_10000pts.ply.\nFinally, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 25 neighbors and a std_ratio of 0.9. Save the clean cloud to ./outputs/clean_nb25_std0.9.ply. Ensure the output format is \".pdf\"",
"mistake_type": "output_format"
},
{
"complete": "Process the multi-view folder located at {input_file_path} to create a 3D representation. The model quality should be ultra-high quality. I need a ply file. Store the result at ./outputs/Courtroom.ply.\nThen, i have a point cloud file for a room, which is placed in the output path of the previous step. I want to delete the chair and table in this point cloud. Please save the result to ./outputs/delete_chair_table.ply.\nFinally, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 25 neighbors and a std_ratio of 2.1. Save the clean cloud to ./outputs/clean_nb25_std2.1.ply.",
"file_path": "./data/source/multiview/scene/Courtroom",
"ground_truth": [
{
"task": "Multi-View 3D Reconstruction",
"usage": "scene",
"quality": "Ultra-High quality",
"speed": null,
"conf_thres": null,
"output_format": "ply",
"output_path": "./outputs/Courtroom.ply",
"extra_info": null
},
{
"task": "Point Cloud Editing",
"targets": [
"chair",
"table"
],
"edit_type": "delete",
"output_format": "ply",
"output_path": "./outputs/delete_chair_table.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 25,
"std_ratio": 2.1,
"output_path": "./outputs/clean_nb25_std2.1.ply"
}
],
"gt_summary": " Multi-View 3D Reconstruction, Point Cloud Editing, Post Processing \n Process the multi-view folder located at {input_file_path} to create a 3D representation. The model quality should be ultra-high quality. I need a ply file. Store the result at ./outputs/Courtroom.ply.\nThen, i have a point cloud file for a room, which is placed in the output path of the previous step. I want to delete the chair and table in this point cloud. Please save the result to ./outputs/delete_chair_table.ply.\nFinally, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 25 neighbors and a std_ratio of 2.1. Save the clean cloud to ./outputs/clean_nb25_std2.1.ply. \n Input 1:./data/source/multiview/scene/Courtroom; Input 2:./outputs/Courtroom.ply; Input 3:./outputs/delete_chair_table.ply; \n Step 1: { Constraints: [quality: Ultra-High quality, output_format: ply]; Output Path: ./outputs/Courtroom.ply }; Step 2: { Targets: chair, table, Output Path: ./outputs/delete_chair_table.ply }; Step 3: { nb_neighbors: 25, std_ratio: 2.1, Output Path: ./outputs/clean_nb25_std2.1.ply } ",
"code": "self.executor.multi_view_reconstruction(video_path='./data/source/multiview/scene/Courtroom', output_path='./outputs/Courtroom.ply', model='vggt', conf_thres=20.0)\nself.executor.pcd_edit(pointcloud_path='./outputs/Courtroom.ply', edit_parts=[\"chair\", \"table\"], output_path='./outputs/delete_chair_table.ply', edit_type='delete')\nself.executor.post_processing(input_path='./outputs/delete_chair_table.ply', process_type='outliers_removal', output_path='./outputs/clean_nb25_std2.1.ply', parameters={\"nb_neighbors\": 25, \"std_ratio\": 2.1})",
"id": 95,
"ambiguous": "Do something with {input_file_path}.\nThen, Edit the point cloud at the output of the previous step.\nFinally, Clean the point cloud at the output of the previous step.",
"missing_info": "Perform multi-view 3D reconstruction using a file path as the input source. The model quality should be specified. Save the output to a file. Save the output to somewhere.\nThen, I have a point cloud file for a room, which is placed in a location. I want to delete the objects in this point cloud. Please save the result to somewhere.\nFinally, Help me filter the noise in a location. Use statistical removal with certain neighbors and a std_ratio. The output should be saved at a path.",
"mistake": "Process the multi-view folder located at {input_file_path} to create a 3D representation. The model quality should be ultra-high quality. I need a ply file. Store the result at ./outputs/Courtroom.ply.\nThen, i have a point cloud file for a room, which is placed in the output path of the previous step. I want to delete the chair and table in this point cloud. Please save the result to ./outputs/delete_chair_table.ply.\nFinally, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 25 neighbors and a std_ratio of 2.1. Save the clean cloud to ./outputs/clean_nb25_std2.1.ply. Ensure low quality.",
"mistake_type": "quality_speed"
},
{
"complete": "Generate a 3D asset from the views provided in {input_file_path}. The reconstruction speed should be fast speed. Save the output as a ply file. Use a confidence cutoff of 21. Save the output to ./outputs/Courtroom.ply.\nThen, i have a point cloud file for a room, which is placed in the output path of the previous step. I want to keep the floor in this point cloud and remove the rest. Please save the result to ./outputs/keep_floor.ply.\nFinally, clean the point cloud file the output path of the previous step. I want to remove outliers with nb_neighbors set to 25 and standard deviation ratio set to 2.5. Output result to ./outputs/clean_nb25_std2.5.ply.",
"file_path": "./data/source/multiview/scene/Courtroom",
"ground_truth": [
{
"task": "Multi-View 3D Reconstruction",
"usage": "scene",
"quality": null,
"speed": "Fast speed",
"conf_thres": 21,
"output_format": "ply",
"output_path": "./outputs/Courtroom.ply",
"extra_info": null
},
{
"task": "Point Cloud Editing",
"targets": [
"floor"
],
"edit_type": "keep",
"output_format": "ply",
"output_path": "./outputs/keep_floor.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 25,
"std_ratio": 2.5,
"output_path": "./outputs/clean_nb25_std2.5.ply"
}
],
"gt_summary": " Multi-View 3D Reconstruction, Point Cloud Editing, Post Processing \n Generate a 3D asset from the views provided in {input_file_path}. The reconstruction speed should be fast speed. Save the output as a ply file. Use a confidence cutoff of 21. Save the output to ./outputs/Courtroom.ply.\nThen, i have a point cloud file for a room, which is placed in the output path of the previous step. I want to keep the floor in this point cloud and remove the rest. Please save the result to ./outputs/keep_floor.ply.\nFinally, clean the point cloud file the output path of the previous step. I want to remove outliers with nb_neighbors set to 25 and standard deviation ratio set to 2.5. Output result to ./outputs/clean_nb25_std2.5.ply. \n Input 1:./data/source/multiview/scene/Courtroom; Input 2:./outputs/Courtroom.ply; Input 3:./outputs/keep_floor.ply; \n Step 1: { Constraints: [speed: Fast speed, output_format: ply, conf_thres: 21]; Output Path: ./outputs/Courtroom.ply }; Step 2: { Targets: floor, Output Path: ./outputs/keep_floor.ply }; Step 3: { nb_neighbors: 25, std_ratio: 2.5, Output Path: ./outputs/clean_nb25_std2.5.ply } ",
"code": "self.executor.multi_view_reconstruction(video_path='./data/source/multiview/scene/Courtroom', output_path='./outputs/Courtroom.ply', model='vggt', conf_thres=21)\nself.executor.pcd_edit(pointcloud_path='./outputs/Courtroom.ply', edit_parts=[\"floor\"], output_path='./outputs/keep_floor.ply', edit_type='keep')\nself.executor.post_processing(input_path='./outputs/keep_floor.ply', process_type='outliers_removal', output_path='./outputs/clean_nb25_std2.5.ply', parameters={\"nb_neighbors\": 25, \"std_ratio\": 2.5})",
"id": 96,
"ambiguous": "Make 3D from the views in {input_file_path}.\nThen, Do something with the output of the previous step.\nFinally, Process the output of the previous step.",
"missing_info": "Perform multi-view 3D reconstruction using a file path as the input source. I need it generated at a certain speed. Filter out points with confidence below a value. Export format: 3D model. Store the generated model.\nThen, Extract the objects from the point cloud at a location. Discard all other objects. Save the extracted objects to a path.\nFinally, Perform post-processing on a file to remove noise. Use a neighbor count of a number and a std_ratio of a value, then save to somewhere.",
"mistake": "Generate a 3D asset from the views provided in {input_file_path}. The reconstruction speed should be fast speed. Save the output as a ply file. Use a confidence cutoff of -21. Save the output to ./outputs/Courtroom.ply.\nThen, i have a point cloud file for a room, which is placed in the output path of the previous step. I want to keep the floor in this point cloud and remove the rest. Please save the result to ./outputs/keep_floor.ply.\nFinally, clean the point cloud file the output path of the previous step. I want to remove outliers with nb_neighbors set to 25 and standard deviation ratio set to 2.5. Output result to ./outputs/clean_nb25_std2.5.ply.",
"mistake_type": "parameter"
},
{
"complete": "Generate a 3D asset from the views provided in {input_file_path}. I require ultra-high quality output. I need it processed with fast speed. I need a ply file. Save the output to ./outputs/bonsai.ply.\nThen, i have a point cloud file for a room, which is placed in the output path of the previous step. I want to keep the floor in this point cloud and remove the rest. Please save the result to ./outputs/keep_floor.ply.\nFinally, clean the point cloud file the output path of the previous step. I want to remove outliers with nb_neighbors set to 40 and standard deviation ratio set to 1.2. Output result to ./outputs/clean_nb40_std1.2.ply.",
"file_path": "./data/source/multiview/scene/bonsai",
"ground_truth": [
{
"task": "Multi-View 3D Reconstruction",
"usage": "scene",
"quality": "Ultra-High quality",
"speed": "Fast speed",
"conf_thres": null,
"output_format": "ply",
"output_path": "./outputs/bonsai.ply",
"extra_info": null
},
{
"task": "Point Cloud Editing",
"targets": [
"floor"
],
"edit_type": "keep",
"output_format": "ply",
"output_path": "./outputs/keep_floor.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 40,
"std_ratio": 1.2,
"output_path": "./outputs/clean_nb40_std1.2.ply"
}
],
"gt_summary": " Multi-View 3D Reconstruction, Point Cloud Editing, Post Processing \n Generate a 3D asset from the views provided in {input_file_path}. I require ultra-high quality output. I need it processed with fast speed. I need a ply file. Save the output to ./outputs/bonsai.ply.\nThen, i have a point cloud file for a room, which is placed in the output path of the previous step. I want to keep the floor in this point cloud and remove the rest. Please save the result to ./outputs/keep_floor.ply.\nFinally, clean the point cloud file the output path of the previous step. I want to remove outliers with nb_neighbors set to 40 and standard deviation ratio set to 1.2. Output result to ./outputs/clean_nb40_std1.2.ply. \n Input 1:./data/source/multiview/scene/bonsai; Input 2:./outputs/bonsai.ply; Input 3:./outputs/keep_floor.ply; \n Step 1: { Constraints: [quality: Ultra-High quality, speed: Fast speed, output_format: ply]; Output Path: ./outputs/bonsai.ply }; Step 2: { Targets: floor, Output Path: ./outputs/keep_floor.ply }; Step 3: { nb_neighbors: 40, std_ratio: 1.2, Output Path: ./outputs/clean_nb40_std1.2.ply } ",
"code": "self.executor.multi_view_reconstruction(video_path='./data/source/multiview/scene/bonsai', output_path='./outputs/bonsai.ply', model='vggt', conf_thres=20.0)\nself.executor.pcd_edit(pointcloud_path='./outputs/bonsai.ply', edit_parts=[\"floor\"], output_path='./outputs/keep_floor.ply', edit_type='keep')\nself.executor.post_processing(input_path='./outputs/keep_floor.ply', process_type='outliers_removal', output_path='./outputs/clean_nb40_std1.2.ply', parameters={\"nb_neighbors\": 40, \"std_ratio\": 1.2})",
"id": 97,
"ambiguous": "Generate 3D from {input_file_path}.\nThen, Modify the file at the output of the previous step.\nFinally, Fix the noise in the output of the previous step.",
"missing_info": "Reconstruct a 3D scene from the image sequence at a location. The model quality should be specified. I need it generated at a certain speed. Save the output to a file. The result should be saved in a location.\nThen, Extract the objects from the point cloud at a location. Discard all other objects. Save the extracted objects to a path.\nFinally, I have a noisy point cloud at a location. Please apply statistical outlier removal using a certain number of neighbors and a std_ratio of a value. Save the clean cloud to somewhere.",
"mistake": "Generate a 3D asset from the views provided in {input_file_path}. I require ultra-high quality output. I need it processed with fast speed. I need a ply file. Save the output to ./outputs/bonsai.ply.\nThen, i have a point cloud file for a room, which is placed in the output path of the previous step. I want to keep the floor in this point cloud and remove the rest. Please save the result to ./outputs/keep_floor.ply.\nFinally, clean the point cloud file the output path of the previous step. I want to remove outliers with nb_neighbors set to 40 and standard deviation ratio set to 1.2. Output result to ./outputs/clean_nb40_std1.2.ply. Ensure slow speed. Ensure low quality.",
"mistake_type": "quality_speed"
},
{
"complete": "Load the video or image folder at {input_file_path} and reconstruct it into a 3D asset. I require ultra-high quality output. Export format: ply. Use a confidence cutoff of 29. The result should be saved in ./outputs/room.ply.\nThen, i have a point cloud file for a room, which is placed in the output path of the previous step. I want to delete the chair and table in this point cloud. Please save the result to ./outputs/delete_chair_table.ply.\nFinally, help me filter the noise in the output path of the previous step. Use statistical removal with 35 neighbors and 1.3 std_ratio. The output should be saved at ./outputs/clean_nb35_std1.3.ply.",
"file_path": "./data/source/multiview/scene/room",
"ground_truth": [
{
"task": "Multi-View 3D Reconstruction",
"usage": "scene",
"quality": "Ultra-High quality",
"speed": null,
"conf_thres": 29,
"output_format": "ply",
"output_path": "./outputs/room.ply",
"extra_info": null
},
{
"task": "Point Cloud Editing",
"targets": [
"chair",
"table"
],
"edit_type": "delete",
"output_format": "ply",
"output_path": "./outputs/delete_chair_table.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 35,
"std_ratio": 1.3,
"output_path": "./outputs/clean_nb35_std1.3.ply"
}
],
"gt_summary": " Multi-View 3D Reconstruction, Point Cloud Editing, Post Processing \n Load the video or image folder at {input_file_path} and reconstruct it into a 3D asset. I require ultra-high quality output. Export format: ply. Use a confidence cutoff of 29. The result should be saved in ./outputs/room.ply.\nThen, i have a point cloud file for a room, which is placed in the output path of the previous step. I want to delete the chair and table in this point cloud. Please save the result to ./outputs/delete_chair_table.ply.\nFinally, help me filter the noise in the output path of the previous step. Use statistical removal with 35 neighbors and 1.3 std_ratio. The output should be saved at ./outputs/clean_nb35_std1.3.ply. \n Input 1:./data/source/multiview/scene/room; Input 2:./outputs/room.ply; Input 3:./outputs/delete_chair_table.ply; \n Step 1: { Constraints: [quality: Ultra-High quality, output_format: ply, conf_thres: 29]; Output Path: ./outputs/room.ply }; Step 2: { Targets: chair, table, Output Path: ./outputs/delete_chair_table.ply }; Step 3: { nb_neighbors: 35, std_ratio: 1.3, Output Path: ./outputs/clean_nb35_std1.3.ply } ",
"code": "self.executor.multi_view_reconstruction(video_path='./data/source/multiview/scene/room', output_path='./outputs/room.ply', model='vggt', conf_thres=29)\nself.executor.pcd_edit(pointcloud_path='./outputs/room.ply', edit_parts=[\"chair\", \"table\"], output_path='./outputs/delete_chair_table.ply', edit_type='delete')\nself.executor.post_processing(input_path='./outputs/delete_chair_table.ply', process_type='outliers_removal', output_path='./outputs/clean_nb35_std1.3.ply', parameters={\"nb_neighbors\": 35, \"std_ratio\": 1.3})",
"id": 98,
"ambiguous": "Generate 3D from {input_file_path}.\nThen, Edit the point cloud at the output of the previous step.\nFinally, Fix the noise in the output of the previous step.",
"missing_info": "Reconstruct a 3D scene from the image sequence at a location. Ensure the result quality is defined. Set the confidence threshold to a value. Export format: 3D model. The result should be saved in a location.\nThen, Remove the objects from the point cloud file at a location. Save the remaining objects to a path.\nFinally, Help me filter the noise in a location. Use statistical removal with certain neighbors and a std_ratio. The output should be saved at a path.",
"mistake": "Load the video or image folder at {input_file_path} and reconstruct it into a 3D asset. I require ultra-high quality output. Export format: ply. Use a confidence cutoff of -29. The result should be saved in ./outputs/room.ply.\nThen, i have a point cloud file for a room, which is placed in the output path of the previous step. I want to delete the chair and table in this point cloud. Please save the result to ./outputs/delete_chair_table.ply.\nFinally, help me filter the noise in the output path of the previous step. Use statistical removal with 35 neighbors and 1.3 std_ratio. The output should be saved at ./outputs/clean_nb35_std1.3.ply.",
"mistake_type": "parameter"
},
{
"complete": "Generate a 3D asset from the views provided in {input_file_path}. Save the output as a ply file. It should be a lightweight point cloud. Filter out points with confidence below 30. Save the output to ./outputs/Ballroom.ply.\nThen, extract the wall from the point cloud at the output path of the previous step. Discard all other objects. Save the extracted objects to ./outputs/keep_wall.ply.\nFinally, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 50 and a std_ratio of 2.5, then save to ./outputs/clean_nb50_std2.5.ply.",
"file_path": "./data/source/multiview/scene/Ballroom",
"ground_truth": [
{
"task": "Multi-View 3D Reconstruction",
"usage": "scene",
"quality": null,
"speed": null,
"conf_thres": 30,
"output_format": "ply",
"output_path": "./outputs/Ballroom.ply",
"extra_info": "Lightweight point cloud"
},
{
"task": "Point Cloud Editing",
"targets": [
"wall"
],
"edit_type": "keep",
"output_format": "ply",
"output_path": "./outputs/keep_wall.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 50,
"std_ratio": 2.5,
"output_path": "./outputs/clean_nb50_std2.5.ply"
}
],
"gt_summary": " Multi-View 3D Reconstruction, Point Cloud Editing, Post Processing \n Generate a 3D asset from the views provided in {input_file_path}. Save the output as a ply file. It should be a lightweight point cloud. Filter out points with confidence below 30. Save the output to ./outputs/Ballroom.ply.\nThen, extract the wall from the point cloud at the output path of the previous step. Discard all other objects. Save the extracted objects to ./outputs/keep_wall.ply.\nFinally, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 50 and a std_ratio of 2.5, then save to ./outputs/clean_nb50_std2.5.ply. \n Input 1:./data/source/multiview/scene/Ballroom; Input 2:./outputs/Ballroom.ply; Input 3:./outputs/keep_wall.ply; \n Step 1: { Constraints: [output_format: ply, features: Lightweight point cloud, conf_thres: 30]; Output Path: ./outputs/Ballroom.ply }; Step 2: { Targets: wall, Output Path: ./outputs/keep_wall.ply }; Step 3: { nb_neighbors: 50, std_ratio: 2.5, Output Path: ./outputs/clean_nb50_std2.5.ply } ",
"code": "self.executor.multi_view_reconstruction(video_path='./data/source/multiview/scene/Ballroom', output_path='./outputs/Ballroom.ply', model='fast3r', conf_thres=30)\nself.executor.pcd_edit(pointcloud_path='./outputs/Ballroom.ply', edit_parts=[\"wall\"], output_path='./outputs/keep_wall.ply', edit_type='keep')\nself.executor.post_processing(input_path='./outputs/keep_wall.ply', process_type='outliers_removal', output_path='./outputs/clean_nb50_std2.5.ply', parameters={\"nb_neighbors\": 50, \"std_ratio\": 2.5})",
"id": 99,
"ambiguous": "Generate 3D from {input_file_path}.\nThen, Edit the point cloud at the output of the previous step.\nFinally, Process the output of the previous step.",
"missing_info": "Generate a 3D asset from the views provided in a location. Use a confidence cutoff of a value. Focus on certain features. Export format: 3D model. The result should be saved in a location.\nThen, I have a point cloud file for a room, which is placed in a location. I want to keep certain objects in this point cloud and remove the rest. Please save the result to somewhere.\nFinally, Perform post-processing on a file to remove noise. Use a neighbor count of a number and a std_ratio of a value, then save to somewhere.",
"mistake": "Generate a 3D asset from the views provided in {input_file_path}. Save the output as a ply file. It should be a lightweight point cloud. Filter out points with confidence below 3000. Save the output to ./outputs/Ballroom.ply.\nThen, extract the wall from the point cloud at the output path of the previous step. Discard all other objects. Save the extracted objects to ./outputs/keep_wall.ply.\nFinally, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 50 and a std_ratio of 2.5, then save to ./outputs/clean_nb50_std2.5.ply.",
"mistake_type": "parameter"
},
{
"complete": "Reconstruct a 3D scene from the image sequence at {input_file_path}. The reconstruction speed should be medium speed. I need a ply file. Store the result at ./outputs/scan11.ply.\nThen, take the mesh at the output path of the previous step and generate a point cloud with 1000 samples. Store the output at ./outputs/example_mesh_from_glb_1000pts.ply.\nFinally, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 20 neighbors and a std_ratio of 0.6. Save the clean cloud to ./outputs/clean_nb20_std0.6.ply.",
"file_path": "./data/source/multiview/object/scan11",
"ground_truth": [
{
"task": "Multi-View 3D Reconstruction",
"usage": "object",
"quality": null,
"speed": "Medium speed",
"conf_thres": null,
"output_format": "ply",
"output_path": "./outputs/scan11.ply",
"extra_info": null
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 1000,
"output_path": "./outputs/example_mesh_from_glb_1000pts.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 20,
"std_ratio": 0.6,
"output_path": "./outputs/clean_nb20_std0.6.ply"
}
],
"gt_summary": " Multi-View 3D Reconstruction, Post Processing, Post Processing \n Reconstruct a 3D scene from the image sequence at {input_file_path}. The reconstruction speed should be medium speed. I need a ply file. Store the result at ./outputs/scan11.ply.\nThen, take the mesh at the output path of the previous step and generate a point cloud with 1000 samples. Store the output at ./outputs/example_mesh_from_glb_1000pts.ply.\nFinally, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 20 neighbors and a std_ratio of 0.6. Save the clean cloud to ./outputs/clean_nb20_std0.6.ply. \n Input 1:./data/source/multiview/object/scan11; Input 2:./outputs/scan11.ply; Input 3:./outputs/example_mesh_from_glb_1000pts.ply; \n Step 1: { Constraints: [speed: Medium speed, output_format: ply]; Output Path: ./outputs/scan11.ply }; Step 2: { number_of_points: 1000; Output Path: ./outputs/example_mesh_from_glb_1000pts.ply }; Step 3: { nb_neighbors: 20, std_ratio: 0.6, Output Path: ./outputs/clean_nb20_std0.6.ply } ",
"code": "self.executor.multi_view_reconstruction(video_path='./data/source/multiview/object/scan11', output_path='./outputs/scan11.ply', model='trellis')\nself.executor.post_processing(input_path='./outputs/scan11.ply', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_glb_1000pts.ply', parameters={\"number_of_points\": 1000})\nself.executor.post_processing(input_path='./outputs/example_mesh_from_glb_1000pts.ply', process_type='outliers_removal', output_path='./outputs/clean_nb20_std0.6.ply', parameters={\"nb_neighbors\": 20, \"std_ratio\": 0.6})",
"id": 100,
"ambiguous": "Process {input_file_path} into 3D.\nThen, Transform the output of the previous step.\nFinally, Do something with the output of the previous step.",
"missing_info": "Reconstruct a 3D scene from the image sequence at a location. Expect appropriate processing speed. Export format: 3D model. Save the output to somewhere.\nThen, Please process the mesh at a file to create a point cloud. Use a sampling rate of a number of points and write the file to somewhere.\nFinally, Clean the point cloud file at a path. I want to remove outliers with nb_neighbors set to a number and standard deviation ratio set to a value. Output result to a location.",
"mistake": "Reconstruct a 3D scene from the image sequence at {input_file_path}. The reconstruction speed should be medium speed. I need a ply file. Store the result at ./outputs/scan11.ply.\nThen, take the mesh at the output path of the previous step and generate a point cloud with 1000 samples. Store the output at ./outputs/example_mesh_from_glb_1000pts.ply.\nFinally, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 20 neighbors and a std_ratio of 0.6. Save the clean cloud to ./outputs/clean_nb20_std0.6.ply. Ensure extremely high speed.",
"mistake_type": "quality_speed"
},
{
"complete": "Generate a 3D asset from the views provided in {input_file_path}. Export format: obj. Ensure high-fidelity mesh. The result should be saved in ./outputs/scan10.obj.\nThen, convert the mesh file at the output path of the previous step into a point cloud by sampling 35000 points. Save the result to ./outputs/example_mesh_from_glb_35000pts.ply.\nFinally, help me filter the noise in the output path of the previous step. Use statistical removal with 25 neighbors and 2.5 std_ratio. The output should be saved at ./outputs/clean_nb25_std2.5.ply.",
"file_path": "./data/source/multiview/object/scan10",
"ground_truth": [
{
"task": "Multi-View 3D Reconstruction",
"usage": "object",
"quality": null,
"speed": null,
"conf_thres": null,
"output_format": "obj",
"output_path": "./outputs/scan10.obj",
"extra_info": "High-fidelity mesh"
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 35000,
"output_path": "./outputs/example_mesh_from_glb_35000pts.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 25,
"std_ratio": 2.5,
"output_path": "./outputs/clean_nb25_std2.5.ply"
}
],
"gt_summary": " Multi-View 3D Reconstruction, Post Processing, Post Processing \n Generate a 3D asset from the views provided in {input_file_path}. Export format: obj. Ensure high-fidelity mesh. The result should be saved in ./outputs/scan10.obj.\nThen, convert the mesh file at the output path of the previous step into a point cloud by sampling 35000 points. Save the result to ./outputs/example_mesh_from_glb_35000pts.ply.\nFinally, help me filter the noise in the output path of the previous step. Use statistical removal with 25 neighbors and 2.5 std_ratio. The output should be saved at ./outputs/clean_nb25_std2.5.ply. \n Input 1:./data/source/multiview/object/scan10; Input 2:./outputs/scan10.obj; Input 3:./outputs/example_mesh_from_glb_35000pts.ply; \n Step 1: { Constraints: [output_format: obj, features: High-fidelity mesh]; Output Path: ./outputs/scan10.obj }; Step 2: { number_of_points: 35000; Output Path: ./outputs/example_mesh_from_glb_35000pts.ply }; Step 3: { nb_neighbors: 25, std_ratio: 2.5, Output Path: ./outputs/clean_nb25_std2.5.ply } ",
"code": "self.executor.multi_view_reconstruction(video_path='./data/source/multiview/object/scan10', output_path='./outputs/scan10.obj', model='trellis')\nself.executor.post_processing(input_path='./outputs/scan10.obj', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_glb_35000pts.ply', parameters={\"number_of_points\": 35000})\nself.executor.post_processing(input_path='./outputs/example_mesh_from_glb_35000pts.ply', process_type='outliers_removal', output_path='./outputs/clean_nb25_std2.5.ply', parameters={\"nb_neighbors\": 25, \"std_ratio\": 2.5})",
"id": 101,
"ambiguous": "Generate 3D from {input_file_path}.\nThen, Transform the output of the previous step.\nFinally, Fix the noise in the output of the previous step.",
"missing_info": "Perform multi-view 3D reconstruction using a file path as the input source. It should have specific attributes. Save the output to a file. The result should be saved in a location.\nThen, Perform mesh-to-pointcloud conversion on the mesh at a file. Set the target number of points to a number and save as somewhere.\nFinally, Perform post-processing on a file to remove noise. Use a neighbor count of a number and a std_ratio of a value, then save to somewhere.",
"mistake": "Generate a 3D asset from the views provided in ./data/source/multiview/object/scan10.ply. Export format: obj. Ensure high-fidelity mesh. The result should be saved in ./outputs/scan10.obj.\nThen, convert the mesh file at the output path of the previous step into a point cloud by sampling 35000 points. Save the result to ./outputs/example_mesh_from_glb_35000pts.ply.\nFinally, help me filter the noise in the output path of the previous step. Use statistical removal with 25 neighbors and 2.5 std_ratio. The output should be saved at ./outputs/clean_nb25_std2.5.ply.",
"mistake_type": "input_format"
},
{
"complete": "Perform multi-view 3D reconstruction using {input_file_path} as the input source. I need it processed with medium speed. Focus on high-fidelity mesh. The result should be saved in ./outputs/scan4.ply.\nThen, take the mesh at the output path of the previous step and generate a point cloud with 15000 samples. Store the output at ./outputs/example_mesh_from_ply_15000pts.ply.\nFinally, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 45 neighbors and a std_ratio of 0.8. Save the clean cloud to ./outputs/clean_nb45_std0.8.ply.",
"file_path": "./data/source/multiview/object/scan4",
"ground_truth": [
{
"task": "Multi-View 3D Reconstruction",
"usage": "object",
"quality": null,
"speed": "Medium speed",
"conf_thres": null,
"output_format": null,
"output_path": "./outputs/scan4.ply",
"extra_info": "High-fidelity mesh"
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 15000,
"output_path": "./outputs/example_mesh_from_ply_15000pts.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 45,
"std_ratio": 0.8,
"output_path": "./outputs/clean_nb45_std0.8.ply"
}
],
"gt_summary": " Multi-View 3D Reconstruction, Post Processing, Post Processing \n Perform multi-view 3D reconstruction using {input_file_path} as the input source. I need it processed with medium speed. Focus on high-fidelity mesh. The result should be saved in ./outputs/scan4.ply.\nThen, take the mesh at the output path of the previous step and generate a point cloud with 15000 samples. Store the output at ./outputs/example_mesh_from_ply_15000pts.ply.\nFinally, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 45 neighbors and a std_ratio of 0.8. Save the clean cloud to ./outputs/clean_nb45_std0.8.ply. \n Input 1:./data/source/multiview/object/scan4; Input 2:./outputs/scan4.ply; Input 3:./outputs/example_mesh_from_ply_15000pts.ply; \n Step 1: { Constraints: [speed: Medium speed, features: High-fidelity mesh]; Output Path: ./outputs/scan4.ply }; Step 2: { number_of_points: 15000; Output Path: ./outputs/example_mesh_from_ply_15000pts.ply }; Step 3: { nb_neighbors: 45, std_ratio: 0.8, Output Path: ./outputs/clean_nb45_std0.8.ply } ",
"code": "self.executor.multi_view_reconstruction(video_path='./data/source/multiview/object/scan4', output_path='./outputs/scan4.ply', model='trellis')\nself.executor.post_processing(input_path='./outputs/scan4.ply', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_ply_15000pts.ply', parameters={\"number_of_points\": 15000})\nself.executor.post_processing(input_path='./outputs/example_mesh_from_ply_15000pts.ply', process_type='outliers_removal', output_path='./outputs/clean_nb45_std0.8.ply', parameters={\"nb_neighbors\": 45, \"std_ratio\": 0.8})",
"id": 102,
"ambiguous": "Make 3D from the views in {input_file_path}.\nThen, Process the output of the previous step.\nFinally, Fix the noise in the output of the previous step.",
"missing_info": "Reconstruct a 3D scene from the image sequence at a location. Expect appropriate processing speed. It should have specific attributes. The result should be saved in a location.\nThen, Perform mesh-to-pointcloud conversion on the mesh at a file. Set the target number of points to a number and save as somewhere.\nFinally, Help me filter the noise in a location. Use statistical removal with certain neighbors and a std_ratio. The output should be saved at a path.",
"mistake": "Perform multi-view 3D reconstruction using {input_file_path} as the input source. I need it processed with medium speed. Focus on high-fidelity mesh. The result should be saved in ./outputs/scan4.ply.\nThen, take the mesh at the output path of the previous step and generate a point cloud with 15000 samples. Store the output at ./outputs/example_mesh_from_ply_15000pts.ply.\nFinally, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 45 neighbors and a std_ratio of 0.8. Save the clean cloud to ./outputs/clean_nb45_std0.8.ply. Ensure low. Ensure extremely high speed.",
"mistake_type": "quality_speed"
},
{
"complete": "Process the multi-view folder located at {input_file_path} to create a 3D representation. I need a obj file. Save the output to ./outputs/scan13.obj.\nThen, perform mesh-to-pointcloud conversion on the mesh at the output path of the previous step. Set the target number of points to 45000 and save as ./outputs/example_mesh_from_glb_45000pts.ply.\nFinally, clean the point cloud file the output path of the previous step. I want to remove outliers with nb_neighbors set to 20 and standard deviation ratio set to 0.7. Output result to ./outputs/clean_nb20_std0.7.ply.",
"file_path": "./data/source/multiview/object/scan13",
"ground_truth": [
{
"task": "Multi-View 3D Reconstruction",
"usage": "object",
"quality": null,
"speed": null,
"conf_thres": null,
"output_format": "obj",
"output_path": "./outputs/scan13.obj",
"extra_info": null
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 45000,
"output_path": "./outputs/example_mesh_from_glb_45000pts.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 20,
"std_ratio": 0.7,
"output_path": "./outputs/clean_nb20_std0.7.ply"
}
],
"gt_summary": " Multi-View 3D Reconstruction, Post Processing, Post Processing \n Process the multi-view folder located at {input_file_path} to create a 3D representation. I need a obj file. Save the output to ./outputs/scan13.obj.\nThen, perform mesh-to-pointcloud conversion on the mesh at the output path of the previous step. Set the target number of points to 45000 and save as ./outputs/example_mesh_from_glb_45000pts.ply.\nFinally, clean the point cloud file the output path of the previous step. I want to remove outliers with nb_neighbors set to 20 and standard deviation ratio set to 0.7. Output result to ./outputs/clean_nb20_std0.7.ply. \n Input 1:./data/source/multiview/object/scan13; Input 2:./outputs/scan13.obj; Input 3:./outputs/example_mesh_from_glb_45000pts.ply; \n Step 1: { Constraints: [output_format: obj]; Output Path: ./outputs/scan13.obj }; Step 2: { number_of_points: 45000; Output Path: ./outputs/example_mesh_from_glb_45000pts.ply }; Step 3: { nb_neighbors: 20, std_ratio: 0.7, Output Path: ./outputs/clean_nb20_std0.7.ply } ",
"code": "self.executor.multi_view_reconstruction(video_path='./data/source/multiview/object/scan13', output_path='./outputs/scan13.obj', model='trellis')\nself.executor.post_processing(input_path='./outputs/scan13.obj', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_glb_45000pts.ply', parameters={\"number_of_points\": 45000})\nself.executor.post_processing(input_path='./outputs/example_mesh_from_glb_45000pts.ply', process_type='outliers_removal', output_path='./outputs/clean_nb20_std0.7.ply', parameters={\"nb_neighbors\": 20, \"std_ratio\": 0.7})",
"id": 103,
"ambiguous": "Do something with {input_file_path}.\nThen, Convert the file at the output of the previous step.\nFinally, Do something with the output of the previous step.",
"missing_info": "Perform multi-view 3D reconstruction using a file path as the input source. Export format: 3D model. The result should be saved in a location.\nThen, Perform mesh-to-pointcloud conversion on the mesh at a file. Set the target number of points to a number and save as somewhere.\nFinally, Clean the point cloud file at a path. I want to remove outliers with nb_neighbors set to a number and standard deviation ratio set to a value. Output result to a location.",
"mistake": "Process the multi-view folder located at ./data/source/multiview/object/scan13.obj to create a 3D representation. I need a obj file. Save the output to ./outputs/scan13.obj.\nThen, perform mesh-to-pointcloud conversion on the mesh at the output path of the previous step. Set the target number of points to 45000 and save as ./outputs/example_mesh_from_glb_45000pts.ply.\nFinally, clean the point cloud file the output path of the previous step. I want to remove outliers with nb_neighbors set to 20 and standard deviation ratio set to 0.7. Output result to ./outputs/clean_nb20_std0.7.ply.",
"mistake_type": "input_format"
},
{
"complete": "Perform multi-view 3D reconstruction using {input_file_path} as the input source. Save the output as a obj file. Ensure high-fidelity mesh. Save the output to ./outputs/scan75.obj.\nThen, perform mesh-to-pointcloud conversion on the mesh at the output path of the previous step. Set the target number of points to 30000 and save as ./outputs/example_mesh_from_ply_30000pts.ply.\nFinally, help me filter the noise in the output path of the previous step. Use statistical removal with 50 neighbors and 0.6 std_ratio. The output should be saved at ./outputs/clean_nb50_std0.6.ply.",
"file_path": "./data/source/multiview/object/scan75",
"ground_truth": [
{
"task": "Multi-View 3D Reconstruction",
"usage": "object",
"quality": null,
"speed": null,
"conf_thres": null,
"output_format": "obj",
"output_path": "./outputs/scan75.obj",
"extra_info": "High-fidelity mesh"
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 30000,
"output_path": "./outputs/example_mesh_from_ply_30000pts.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 50,
"std_ratio": 0.6,
"output_path": "./outputs/clean_nb50_std0.6.ply"
}
],
"gt_summary": " Multi-View 3D Reconstruction, Post Processing, Post Processing \n Perform multi-view 3D reconstruction using {input_file_path} as the input source. Save the output as a obj file. Ensure high-fidelity mesh. Save the output to ./outputs/scan75.obj.\nThen, perform mesh-to-pointcloud conversion on the mesh at the output path of the previous step. Set the target number of points to 30000 and save as ./outputs/example_mesh_from_ply_30000pts.ply.\nFinally, help me filter the noise in the output path of the previous step. Use statistical removal with 50 neighbors and 0.6 std_ratio. The output should be saved at ./outputs/clean_nb50_std0.6.ply. \n Input 1:./data/source/multiview/object/scan75; Input 2:./outputs/scan75.obj; Input 3:./outputs/example_mesh_from_ply_30000pts.ply; \n Step 1: { Constraints: [output_format: obj, features: High-fidelity mesh]; Output Path: ./outputs/scan75.obj }; Step 2: { number_of_points: 30000; Output Path: ./outputs/example_mesh_from_ply_30000pts.ply }; Step 3: { nb_neighbors: 50, std_ratio: 0.6, Output Path: ./outputs/clean_nb50_std0.6.ply } ",
"code": "self.executor.multi_view_reconstruction(video_path='./data/source/multiview/object/scan75', output_path='./outputs/scan75.obj', model='trellis')\nself.executor.post_processing(input_path='./outputs/scan75.obj', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_ply_30000pts.ply', parameters={\"number_of_points\": 30000})\nself.executor.post_processing(input_path='./outputs/example_mesh_from_ply_30000pts.ply', process_type='outliers_removal', output_path='./outputs/clean_nb50_std0.6.ply', parameters={\"nb_neighbors\": 50, \"std_ratio\": 0.6})",
"id": 104,
"ambiguous": "Do something with {input_file_path}.\nThen, Convert the file at the output of the previous step.\nFinally, Process the output of the previous step.",
"missing_info": "Process the multi-view folder located somewhere to create a 3D representation. It should have specific attributes. I need a file. Store the generated model.\nThen, I need to sample a number of points uniformly from the 3D model at a path. Output the processed ply file to a location.\nFinally, Perform post-processing on a file to remove noise. Use a neighbor count of a number and a std_ratio of a value, then save to somewhere.",
"mistake": "Perform multi-view 3D reconstruction using ./data/source/multiview/object/scan75.stl as the input source. Save the output as a obj file. Ensure high-fidelity mesh. Save the output to ./outputs/scan75.obj.\nThen, perform mesh-to-pointcloud conversion on the mesh at the output path of the previous step. Set the target number of points to 30000 and save as ./outputs/example_mesh_from_ply_30000pts.ply.\nFinally, help me filter the noise in the output path of the previous step. Use statistical removal with 50 neighbors and 0.6 std_ratio. The output should be saved at ./outputs/clean_nb50_std0.6.ply.",
"mistake_type": "input_format"
},
{
"complete": "Generate a 3D asset from the views provided in {input_file_path}. I need it processed with medium speed. Ensure high-fidelity mesh. Store the result at ./outputs/scan114.glb.\nThen, convert the mesh file at the output path of the previous step into a point cloud by sampling 25000 points. Save the result to ./outputs/example_mesh_from_glb_25000pts.ply.\nFinally, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 45 and a std_ratio of 1.8, then save to ./outputs/clean_nb45_std1.8.ply.",
"file_path": "./data/source/multiview/object/scan114",
"ground_truth": [
{
"task": "Multi-View 3D Reconstruction",
"usage": "object",
"quality": null,
"speed": "Medium speed",
"conf_thres": null,
"output_format": null,
"output_path": "./outputs/scan114.glb",
"extra_info": "High-fidelity mesh"
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 25000,
"output_path": "./outputs/example_mesh_from_glb_25000pts.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 45,
"std_ratio": 1.8,
"output_path": "./outputs/clean_nb45_std1.8.ply"
}
],
"gt_summary": " Multi-View 3D Reconstruction, Post Processing, Post Processing \n Generate a 3D asset from the views provided in {input_file_path}. I need it processed with medium speed. Ensure high-fidelity mesh. Store the result at ./outputs/scan114.glb.\nThen, convert the mesh file at the output path of the previous step into a point cloud by sampling 25000 points. Save the result to ./outputs/example_mesh_from_glb_25000pts.ply.\nFinally, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 45 and a std_ratio of 1.8, then save to ./outputs/clean_nb45_std1.8.ply. \n Input 1:./data/source/multiview/object/scan114; Input 2:./outputs/scan114.glb; Input 3:./outputs/example_mesh_from_glb_25000pts.ply; \n Step 1: { Constraints: [speed: Medium speed, features: High-fidelity mesh]; Output Path: ./outputs/scan114.glb }; Step 2: { number_of_points: 25000; Output Path: ./outputs/example_mesh_from_glb_25000pts.ply }; Step 3: { nb_neighbors: 45, std_ratio: 1.8, Output Path: ./outputs/clean_nb45_std1.8.ply } ",
"code": "self.executor.multi_view_reconstruction(video_path='./data/source/multiview/object/scan114', output_path='./outputs/scan114.glb', model='trellis')\nself.executor.post_processing(input_path='./outputs/scan114.glb', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_glb_25000pts.ply', parameters={\"number_of_points\": 25000})\nself.executor.post_processing(input_path='./outputs/example_mesh_from_glb_25000pts.ply', process_type='outliers_removal', output_path='./outputs/clean_nb45_std1.8.ply', parameters={\"nb_neighbors\": 45, \"std_ratio\": 1.8})",
"id": 105,
"ambiguous": "Make 3D from the views in {input_file_path}.\nThen, Convert the file at the output of the previous step.\nFinally, Clean the point cloud at the output of the previous step.",
"missing_info": "Load the video or image folder from somewhere and reconstruct it into a 3D asset. Expect appropriate processing speed. Focus on certain features. Store the generated model.\nThen, Perform mesh-to-pointcloud conversion on the mesh at a file. Set the target number of points to a number and save as somewhere.\nFinally, Clean the point cloud file at a path. I want to remove outliers with nb_neighbors set to a number and standard deviation ratio set to a value. Output result to a location.",
"mistake": "Generate a 3D asset from the views provided in {input_file_path}. I need it processed with medium speed. Ensure high-fidelity mesh. Store the result at ./outputs/scan114.glb.\nThen, convert the mesh file at the output path of the previous step into a point cloud by sampling 25000 points. Save the result to ./outputs/example_mesh_from_glb_25000pts.ply.\nFinally, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 45 and a std_ratio of 1.8, then save to ./outputs/clean_nb45_std1.8.ply. Ensure low. Ensure extremely high speed.",
"mistake_type": "quality_speed"
},
{
"complete": "Generate a 3D asset from the views provided in {input_file_path}. Export format: obj. The result should be saved in ./outputs/scan114.obj.\nThen, perform mesh-to-pointcloud conversion on the mesh at the output path of the previous step. Set the target number of points to 25000 and save as ./outputs/example_mesh_from_obj_25000pts.ply.\nFinally, clean the point cloud file the output path of the previous step. I want to remove outliers with nb_neighbors set to 40 and standard deviation ratio set to 2.5. Output result to ./outputs/clean_nb40_std2.5.ply.",
"file_path": "./data/source/multiview/object/scan114",
"ground_truth": [
{
"task": "Multi-View 3D Reconstruction",
"usage": "object",
"quality": null,
"speed": null,
"conf_thres": null,
"output_format": "obj",
"output_path": "./outputs/scan114.obj",
"extra_info": null
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 25000,
"output_path": "./outputs/example_mesh_from_obj_25000pts.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 40,
"std_ratio": 2.5,
"output_path": "./outputs/clean_nb40_std2.5.ply"
}
],
"gt_summary": " Multi-View 3D Reconstruction, Post Processing, Post Processing \n Generate a 3D asset from the views provided in {input_file_path}. Export format: obj. The result should be saved in ./outputs/scan114.obj.\nThen, perform mesh-to-pointcloud conversion on the mesh at the output path of the previous step. Set the target number of points to 25000 and save as ./outputs/example_mesh_from_obj_25000pts.ply.\nFinally, clean the point cloud file the output path of the previous step. I want to remove outliers with nb_neighbors set to 40 and standard deviation ratio set to 2.5. Output result to ./outputs/clean_nb40_std2.5.ply. \n Input 1:./data/source/multiview/object/scan114; Input 2:./outputs/scan114.obj; Input 3:./outputs/example_mesh_from_obj_25000pts.ply; \n Step 1: { Constraints: [output_format: obj]; Output Path: ./outputs/scan114.obj }; Step 2: { number_of_points: 25000; Output Path: ./outputs/example_mesh_from_obj_25000pts.ply }; Step 3: { nb_neighbors: 40, std_ratio: 2.5, Output Path: ./outputs/clean_nb40_std2.5.ply } ",
"code": "self.executor.multi_view_reconstruction(video_path='./data/source/multiview/object/scan114', output_path='./outputs/scan114.obj', model='trellis')\nself.executor.post_processing(input_path='./outputs/scan114.obj', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_obj_25000pts.ply', parameters={\"number_of_points\": 25000})\nself.executor.post_processing(input_path='./outputs/example_mesh_from_obj_25000pts.ply', process_type='outliers_removal', output_path='./outputs/clean_nb40_std2.5.ply', parameters={\"nb_neighbors\": 40, \"std_ratio\": 2.5})",
"id": 106,
"ambiguous": "Make 3D from the views in {input_file_path}.\nThen, Convert the file at the output of the previous step.\nFinally, Clean the point cloud at the output of the previous step.",
"missing_info": "Load the video or image folder from somewhere and reconstruct it into a 3D asset. I need a file. The result should be saved in a location.\nThen, Perform mesh-to-pointcloud conversion on the mesh at a file. Set the target number of points to a number and save as somewhere.\nFinally, Help me filter the noise in a location. Use statistical removal with certain neighbors and a std_ratio. The output should be saved at a path.",
"mistake": "Generate a 3D asset from the views provided in {input_file_path}. Export format: obj. The result should be saved in ./outputs/scan114.obj.\nThen, perform mesh-to-pointcloud conversion on the mesh at the output path of the previous step. Set the target number of points to 25000 and save as ./outputs/example_mesh_from_obj_25000pts.ply.\nFinally, clean the point cloud file the output path of the previous step. I want to remove outliers with nb_neighbors set to 40 and standard deviation ratio set to 2.5. Output result to ./outputs/clean_nb40_std2.5.ply. Ensure the output format is \".pdf\"",
"mistake_type": "output_format"
},
{
"complete": "Perform multi-view 3D reconstruction using {input_file_path} as the input source. I require high quality output. Ensure high-fidelity mesh. Save the output to ./outputs/scan62.ply.\nThen, convert the mesh file at the output path of the previous step into a point cloud by sampling 10000 points. Save the result to ./outputs/example_mesh_from_obj_10000pts.ply.\nFinally, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 25 and a std_ratio of 0.6, then save to ./outputs/clean_nb25_std0.6.ply.",
"file_path": "./data/source/multiview/object/scan62",
"ground_truth": [
{
"task": "Multi-View 3D Reconstruction",
"usage": "object",
"quality": "High quality",
"speed": null,
"conf_thres": null,
"output_format": null,
"output_path": "./outputs/scan62.ply",
"extra_info": "High-fidelity mesh"
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 10000,
"output_path": "./outputs/example_mesh_from_obj_10000pts.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 25,
"std_ratio": 0.6,
"output_path": "./outputs/clean_nb25_std0.6.ply"
}
],
"gt_summary": " Multi-View 3D Reconstruction, Post Processing, Post Processing \n Perform multi-view 3D reconstruction using {input_file_path} as the input source. I require high quality output. Ensure high-fidelity mesh. Save the output to ./outputs/scan62.ply.\nThen, convert the mesh file at the output path of the previous step into a point cloud by sampling 10000 points. Save the result to ./outputs/example_mesh_from_obj_10000pts.ply.\nFinally, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 25 and a std_ratio of 0.6, then save to ./outputs/clean_nb25_std0.6.ply. \n Input 1:./data/source/multiview/object/scan62; Input 2:./outputs/scan62.ply; Input 3:./outputs/example_mesh_from_obj_10000pts.ply; \n Step 1: { Constraints: [quality: High quality, features: High-fidelity mesh]; Output Path: ./outputs/scan62.ply }; Step 2: { number_of_points: 10000; Output Path: ./outputs/example_mesh_from_obj_10000pts.ply }; Step 3: { nb_neighbors: 25, std_ratio: 0.6, Output Path: ./outputs/clean_nb25_std0.6.ply } ",
"code": "self.executor.multi_view_reconstruction(video_path='./data/source/multiview/object/scan62', output_path='./outputs/scan62.ply', model='trellis')\nself.executor.post_processing(input_path='./outputs/scan62.ply', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_obj_10000pts.ply', parameters={\"number_of_points\": 10000})\nself.executor.post_processing(input_path='./outputs/example_mesh_from_obj_10000pts.ply', process_type='outliers_removal', output_path='./outputs/clean_nb25_std0.6.ply', parameters={\"nb_neighbors\": 25, \"std_ratio\": 0.6})",
"id": 107,
"ambiguous": "Make 3D from the views in {input_file_path}.\nThen, Transform the output of the previous step.\nFinally, Clean the point cloud at the output of the previous step.",
"missing_info": "Load the video or image folder from somewhere and reconstruct it into a 3D asset. I require a certain quality. Consider that this has particular characteristics. Save the output to somewhere.\nThen, Please process the mesh at a file to create a point cloud. Use a sampling rate of a number of points and write the file to somewhere.\nFinally, Help me filter the noise in a location. Use statistical removal with certain neighbors and a std_ratio. The output should be saved at a path.",
"mistake": "Perform multi-view 3D reconstruction using ./data/source/multiview/object/scan62.stl as the input source. I require high quality output. Ensure high-fidelity mesh. Save the output to ./outputs/scan62.ply.\nThen, convert the mesh file at the output path of the previous step into a point cloud by sampling 10000 points. Save the result to ./outputs/example_mesh_from_obj_10000pts.ply.\nFinally, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 25 and a std_ratio of 0.6, then save to ./outputs/clean_nb25_std0.6.ply.",
"mistake_type": "input_format"
},
{
"complete": "Load the video or image folder at {input_file_path} and reconstruct it into a 3D asset. The reconstruction speed should be medium speed. Export format: obj. Save the output to ./outputs/scan48.obj.\nThen, convert the mesh file at the output path of the previous step into a point cloud by sampling 30000 points. Save the result to ./outputs/example_mesh_from_obj_30000pts.ply.\nFinally, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 45 neighbors and a std_ratio of 1.5. Save the clean cloud to ./outputs/clean_nb45_std1.5.ply.",
"file_path": "./data/source/multiview/object/scan48",
"ground_truth": [
{
"task": "Multi-View 3D Reconstruction",
"usage": "object",
"quality": null,
"speed": "Medium speed",
"conf_thres": null,
"output_format": "obj",
"output_path": "./outputs/scan48.obj",
"extra_info": null
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 30000,
"output_path": "./outputs/example_mesh_from_obj_30000pts.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 45,
"std_ratio": 1.5,
"output_path": "./outputs/clean_nb45_std1.5.ply"
}
],
"gt_summary": " Multi-View 3D Reconstruction, Post Processing, Post Processing \n Load the video or image folder at {input_file_path} and reconstruct it into a 3D asset. The reconstruction speed should be medium speed. Export format: obj. Save the output to ./outputs/scan48.obj.\nThen, convert the mesh file at the output path of the previous step into a point cloud by sampling 30000 points. Save the result to ./outputs/example_mesh_from_obj_30000pts.ply.\nFinally, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 45 neighbors and a std_ratio of 1.5. Save the clean cloud to ./outputs/clean_nb45_std1.5.ply. \n Input 1:./data/source/multiview/object/scan48; Input 2:./outputs/scan48.obj; Input 3:./outputs/example_mesh_from_obj_30000pts.ply; \n Step 1: { Constraints: [speed: Medium speed, output_format: obj]; Output Path: ./outputs/scan48.obj }; Step 2: { number_of_points: 30000; Output Path: ./outputs/example_mesh_from_obj_30000pts.ply }; Step 3: { nb_neighbors: 45, std_ratio: 1.5, Output Path: ./outputs/clean_nb45_std1.5.ply } ",
"code": "self.executor.multi_view_reconstruction(video_path='./data/source/multiview/object/scan48', output_path='./outputs/scan48.obj', model='trellis')\nself.executor.post_processing(input_path='./outputs/scan48.obj', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_obj_30000pts.ply', parameters={\"number_of_points\": 30000})\nself.executor.post_processing(input_path='./outputs/example_mesh_from_obj_30000pts.ply', process_type='outliers_removal', output_path='./outputs/clean_nb45_std1.5.ply', parameters={\"nb_neighbors\": 45, \"std_ratio\": 1.5})",
"id": 108,
"ambiguous": "Make 3D from the views in {input_file_path}.\nThen, Convert the file at the output of the previous step.\nFinally, Clean the point cloud at the output of the previous step.",
"missing_info": "Process the multi-view folder located somewhere to create a 3D representation. Expect appropriate processing speed. Save the output to a file. Store the generated model.\nThen, Please process the mesh at a file to create a point cloud. Use a sampling rate of a number of points and write the file to somewhere.\nFinally, Perform post-processing on a file to remove noise. Use a neighbor count of a number and a std_ratio of a value, then save to somewhere.",
"mistake": "Load the video or image folder at {input_file_path} and reconstruct it into a 3D asset. The reconstruction speed should be medium speed. Export format: obj. Save the output to ./outputs/scan48.obj.\nThen, convert the mesh file at the output path of the previous step into a point cloud by sampling 30000 points. Save the result to ./outputs/example_mesh_from_obj_30000pts.ply.\nFinally, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 45 neighbors and a std_ratio of 1.5. Save the clean cloud to ./outputs/clean_nb45_std1.5.ply. Ensure extremely high speed.",
"mistake_type": "quality_speed"
},
{
"complete": "Generate a 3D asset from the views provided in {input_file_path}. I need a obj file. Store the result at ./outputs/scan11.obj.\nThen, please process the mesh at the output path of the previous step to create a point cloud. Use a sampling rate of 30000 points and write the file to ./outputs/example_mesh_from_obj_30000pts.ply.\nFinally, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 20 neighbors and a std_ratio of 1.7. Save the clean cloud to ./outputs/clean_nb20_std1.7.ply.",
"file_path": "./data/source/multiview/object/scan11",
"ground_truth": [
{
"task": "Multi-View 3D Reconstruction",
"usage": "object",
"quality": null,
"speed": null,
"conf_thres": null,
"output_format": "obj",
"output_path": "./outputs/scan11.obj",
"extra_info": null
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 30000,
"output_path": "./outputs/example_mesh_from_obj_30000pts.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 20,
"std_ratio": 1.7,
"output_path": "./outputs/clean_nb20_std1.7.ply"
}
],
"gt_summary": " Multi-View 3D Reconstruction, Post Processing, Post Processing \n Generate a 3D asset from the views provided in {input_file_path}. I need a obj file. Store the result at ./outputs/scan11.obj.\nThen, please process the mesh at the output path of the previous step to create a point cloud. Use a sampling rate of 30000 points and write the file to ./outputs/example_mesh_from_obj_30000pts.ply.\nFinally, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 20 neighbors and a std_ratio of 1.7. Save the clean cloud to ./outputs/clean_nb20_std1.7.ply. \n Input 1:./data/source/multiview/object/scan11; Input 2:./outputs/scan11.obj; Input 3:./outputs/example_mesh_from_obj_30000pts.ply; \n Step 1: { Constraints: [output_format: obj]; Output Path: ./outputs/scan11.obj }; Step 2: { number_of_points: 30000; Output Path: ./outputs/example_mesh_from_obj_30000pts.ply }; Step 3: { nb_neighbors: 20, std_ratio: 1.7, Output Path: ./outputs/clean_nb20_std1.7.ply } ",
"code": "self.executor.multi_view_reconstruction(video_path='./data/source/multiview/object/scan11', output_path='./outputs/scan11.obj', model='trellis')\nself.executor.post_processing(input_path='./outputs/scan11.obj', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_obj_30000pts.ply', parameters={\"number_of_points\": 30000})\nself.executor.post_processing(input_path='./outputs/example_mesh_from_obj_30000pts.ply', process_type='outliers_removal', output_path='./outputs/clean_nb20_std1.7.ply', parameters={\"nb_neighbors\": 20, \"std_ratio\": 1.7})",
"id": 109,
"ambiguous": "Generate 3D from {input_file_path}.\nThen, Process the output of the previous step.\nFinally, Do something with the output of the previous step.",
"missing_info": "Reconstruct a 3D scene from the image sequence at a location. Save the output to a file. The result should be saved in a location.\nThen, Perform mesh-to-pointcloud conversion on the mesh at a file. Set the target number of points to a number and save as somewhere.\nFinally, Clean the point cloud file at a path. I want to remove outliers with nb_neighbors set to a number and standard deviation ratio set to a value. Output result to a location.",
"mistake": "Generate a 3D asset from the views provided in ./data/source/multiview/object/scan11.obj. I need a obj file. Store the result at ./outputs/scan11.obj.\nThen, please process the mesh at the output path of the previous step to create a point cloud. Use a sampling rate of 30000 points and write the file to ./outputs/example_mesh_from_obj_30000pts.ply.\nFinally, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 20 neighbors and a std_ratio of 1.7. Save the clean cloud to ./outputs/clean_nb20_std1.7.ply.",
"mistake_type": "input_format"
},
{
"complete": "Perform single-view 3D reconstruction using {input_file_path} as input. I require high quality output. Save the output as a obj file. Ensure high-fidelity mesh. Store the result at ./outputs/scan75.obj.\nThen, perform mesh-to-pointcloud conversion on the mesh at the output path of the previous step. Set the target number of points to 45000 and save as ./outputs/example_mesh_from_glb_45000pts.ply.\nFinally, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 20 neighbors and a std_ratio of 1.7. Save the clean cloud to ./outputs/clean_nb20_std1.7.ply.",
"file_path": "./data/source/singleview/object/scan75.jpg",
"ground_truth": [
{
"task": "Single-View 3D Reconstruction",
"usage": "object",
"quality": "High quality",
"speed": null,
"conf_thres": null,
"output_format": "obj",
"output_path": "./outputs/scan75.obj",
"extra_info": "High-fidelity mesh"
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 45000,
"output_path": "./outputs/example_mesh_from_glb_45000pts.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 20,
"std_ratio": 1.7,
"output_path": "./outputs/clean_nb20_std1.7.ply"
}
],
"gt_summary": " Single-View 3D Reconstruction, Post Processing, Post Processing \n Perform single-view 3D reconstruction using {input_file_path} as input. I require high quality output. Save the output as a obj file. Ensure high-fidelity mesh. Store the result at ./outputs/scan75.obj.\nThen, perform mesh-to-pointcloud conversion on the mesh at the output path of the previous step. Set the target number of points to 45000 and save as ./outputs/example_mesh_from_glb_45000pts.ply.\nFinally, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 20 neighbors and a std_ratio of 1.7. Save the clean cloud to ./outputs/clean_nb20_std1.7.ply. \n Input 1:./data/source/singleview/object/scan75.jpg; Input 2:./outputs/scan75.obj; Input 3:./outputs/example_mesh_from_glb_45000pts.ply; \n Step 1: { Constraints: [quality: High quality, output_format: obj, features: High-fidelity mesh]; Output Path: ./outputs/scan75.obj }; Step 2: { number_of_points: 45000; Output Path: ./outputs/example_mesh_from_glb_45000pts.ply }; Step 3: { nb_neighbors: 20, std_ratio: 1.7, Output Path: ./outputs/clean_nb20_std1.7.ply } ",
"code": "self.executor.single_view_reconstruction(image_path='./data/source/singleview/object/scan75.jpg', output_path='./outputs/scan75.obj', model='trellis')\nself.executor.post_processing(input_path='./outputs/scan75.obj', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_glb_45000pts.ply', parameters={\"number_of_points\": 45000})\nself.executor.post_processing(input_path='./outputs/example_mesh_from_glb_45000pts.ply', process_type='outliers_removal', output_path='./outputs/clean_nb20_std1.7.ply', parameters={\"nb_neighbors\": 20, \"std_ratio\": 1.7})",
"id": 110,
"ambiguous": "Make 3D from {input_file_path}.\nThen, Process the output of the previous step.\nFinally, Fix the noise in the output of the previous step.",
"missing_info": "Process the single-view image to create a 3D representation. The model quality should be specified. Focus on certain features. Save the output to a file. Save the output to somewhere.\nThen, Take the mesh at a location and generate a point cloud with a certain number of samples. Store the output at a path.\nFinally, I have a noisy point cloud at a location. Please apply statistical outlier removal using a certain number of neighbors and a std_ratio of a value. Save the clean cloud to somewhere.",
"mistake": "Perform single-view 3D reconstruction using {input_file_path} as input. I require high quality output. Save the output as a obj file. Ensure high-fidelity mesh. Store the result at ./outputs/scan75.obj.\nThen, perform mesh-to-pointcloud conversion on the mesh at the output path of the previous step. Set the target number of points to 45000 and save as ./outputs/example_mesh_from_glb_45000pts.ply.\nFinally, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 20 neighbors and a std_ratio of 1.7. Save the clean cloud to ./outputs/clean_nb20_std1.7.ply. Ensure low. Ensure low quality.",
"mistake_type": "quality_speed"
},
{
"complete": "Load image {input_file_path} and reconstruct it into 3D asset. The reconstruction speed should be medium speed. I need a glb file. Ensure high-fidelity mesh. Save the output to ./outputs/scan118.glb.\nThen, take the mesh at the output path of the previous step and generate a point cloud with 30000 samples. Store the output at ./outputs/example_mesh_from_glb_30000pts.ply.\nFinally, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 45 neighbors and a std_ratio of 2.2. Save the clean cloud to ./outputs/clean_nb45_std2.2.ply.",
"file_path": "./data/source/singleview/object/scan118.jpg",
"ground_truth": [
{
"task": "Single-View 3D Reconstruction",
"usage": "object",
"quality": null,
"speed": "Medium speed",
"conf_thres": null,
"output_format": "glb",
"output_path": "./outputs/scan118.glb",
"extra_info": "High-fidelity mesh"
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 30000,
"output_path": "./outputs/example_mesh_from_glb_30000pts.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 45,
"std_ratio": 2.2,
"output_path": "./outputs/clean_nb45_std2.2.ply"
}
],
"gt_summary": " Single-View 3D Reconstruction, Post Processing, Post Processing \n Load image {input_file_path} and reconstruct it into 3D asset. The reconstruction speed should be medium speed. I need a glb file. Ensure high-fidelity mesh. Save the output to ./outputs/scan118.glb.\nThen, take the mesh at the output path of the previous step and generate a point cloud with 30000 samples. Store the output at ./outputs/example_mesh_from_glb_30000pts.ply.\nFinally, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 45 neighbors and a std_ratio of 2.2. Save the clean cloud to ./outputs/clean_nb45_std2.2.ply. \n Input 1:./data/source/singleview/object/scan118.jpg; Input 2:./outputs/scan118.glb; Input 3:./outputs/example_mesh_from_glb_30000pts.ply; \n Step 1: { Constraints: [speed: Medium speed, output_format: glb, features: High-fidelity mesh]; Output Path: ./outputs/scan118.glb }; Step 2: { number_of_points: 30000; Output Path: ./outputs/example_mesh_from_glb_30000pts.ply }; Step 3: { nb_neighbors: 45, std_ratio: 2.2, Output Path: ./outputs/clean_nb45_std2.2.ply } ",
"code": "self.executor.single_view_reconstruction(image_path='./data/source/singleview/object/scan118.jpg', output_path='./outputs/scan118.glb', model='trellis')\nself.executor.post_processing(input_path='./outputs/scan118.glb', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_glb_30000pts.ply', parameters={\"number_of_points\": 30000})\nself.executor.post_processing(input_path='./outputs/example_mesh_from_glb_30000pts.ply', process_type='outliers_removal', output_path='./outputs/clean_nb45_std2.2.ply', parameters={\"nb_neighbors\": 45, \"std_ratio\": 2.2})",
"id": 111,
"ambiguous": "Fix {input_file_path} to 3D.\nThen, Transform the output of the previous step.\nFinally, Fix the noise in the output of the previous step.",
"missing_info": "Load image and reconstruct it into 3D asset. I need it generated at a certain speed. Consider that this has particular characteristics. I need a file. Save the output to somewhere.\nThen, Take the mesh at a location and generate a point cloud with a certain number of samples. Store the output at a path.\nFinally, Help me filter the noise in a location. Use statistical removal with certain neighbors and a std_ratio. The output should be saved at a path.",
"mistake": "Load image ./data/source/singleview/object/scan118.obj and reconstruct it into 3D asset. The reconstruction speed should be medium speed. I need a glb file. Ensure high-fidelity mesh. Save the output to ./outputs/scan118.glb.\nThen, take the mesh at the output path of the previous step and generate a point cloud with 30000 samples. Store the output at ./outputs/example_mesh_from_glb_30000pts.ply.\nFinally, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 45 neighbors and a std_ratio of 2.2. Save the clean cloud to ./outputs/clean_nb45_std2.2.ply.",
"mistake_type": "input_format"
},
{
"complete": "Process the single-view image located at {input_file_path} to create a 3D representation. The model quality should be high quality. The reconstruction speed should be medium speed. I need a ply file. The result should be saved in ./outputs/scan75.ply.\nThen, take the mesh at the output path of the previous step and generate a point cloud with 1000 samples. Store the output at ./outputs/example_mesh_from_glb_1000pts.ply.\nFinally, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 25 and a std_ratio of 2.2, then save to ./outputs/clean_nb25_std2.2.ply.",
"file_path": "./data/source/singleview/object/scan75.jpg",
"ground_truth": [
{
"task": "Single-View 3D Reconstruction",
"usage": "object",
"quality": "High quality",
"speed": "Medium speed",
"conf_thres": null,
"output_format": "ply",
"output_path": "./outputs/scan75.ply",
"extra_info": null
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 1000,
"output_path": "./outputs/example_mesh_from_glb_1000pts.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 25,
"std_ratio": 2.2,
"output_path": "./outputs/clean_nb25_std2.2.ply"
}
],
"gt_summary": " Single-View 3D Reconstruction, Post Processing, Post Processing \n Process the single-view image located at {input_file_path} to create a 3D representation. The model quality should be high quality. The reconstruction speed should be medium speed. I need a ply file. The result should be saved in ./outputs/scan75.ply.\nThen, take the mesh at the output path of the previous step and generate a point cloud with 1000 samples. Store the output at ./outputs/example_mesh_from_glb_1000pts.ply.\nFinally, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 25 and a std_ratio of 2.2, then save to ./outputs/clean_nb25_std2.2.ply. \n Input 1:./data/source/singleview/object/scan75.jpg; Input 2:./outputs/scan75.ply; Input 3:./outputs/example_mesh_from_glb_1000pts.ply; \n Step 1: { Constraints: [quality: High quality, speed: Medium speed, output_format: ply]; Output Path: ./outputs/scan75.ply }; Step 2: { number_of_points: 1000; Output Path: ./outputs/example_mesh_from_glb_1000pts.ply }; Step 3: { nb_neighbors: 25, std_ratio: 2.2, Output Path: ./outputs/clean_nb25_std2.2.ply } ",
"code": "self.executor.single_view_reconstruction(image_path='./data/source/singleview/object/scan75.jpg', output_path='./outputs/scan75.ply', model='trellis')\nself.executor.post_processing(input_path='./outputs/scan75.ply', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_glb_1000pts.ply', parameters={\"number_of_points\": 1000})\nself.executor.post_processing(input_path='./outputs/example_mesh_from_glb_1000pts.ply', process_type='outliers_removal', output_path='./outputs/clean_nb25_std2.2.ply', parameters={\"nb_neighbors\": 25, \"std_ratio\": 2.2})",
"id": 112,
"ambiguous": "Fix {input_file_path} to 3D.\nThen, Transform the output of the previous step.\nFinally, Do something with the output of the previous step.",
"missing_info": "Load image and reconstruct it into 3D asset. The model quality should be specified. The generation speed should be specified. I need a file. The result should be saved in a location.\nThen, Convert the mesh file at a location into a point cloud by sampling a certain number of points. Save the result to somewhere.\nFinally, I have a noisy point cloud at a location. Please apply statistical outlier removal using a certain number of neighbors and a std_ratio of a value. Save the clean cloud to somewhere.",
"mistake": "Process the single-view image located at {input_file_path} to create a 3D representation. The model quality should be high quality. The reconstruction speed should be medium speed. I need a ply file. The result should be saved in ./outputs/scan75.ply.\nThen, take the mesh at the output path of the previous step and generate a point cloud with 1000 samples. Store the output at ./outputs/example_mesh_from_glb_1000pts.ply.\nFinally, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 25 and a std_ratio of 2.2, then save to ./outputs/clean_nb25_std2.2.ply. Ensure extremely high speed. Ensure low quality.",
"mistake_type": "quality_speed"
},
{
"complete": "Load image {input_file_path} and reconstruct it into 3D asset. The reconstruction speed should be medium speed. Save the output as a ply file. Save the output to ./outputs/scan10.ply.\nThen, please process the mesh at the output path of the previous step to create a point cloud. Use a sampling rate of 1000 points and write the file to ./outputs/example_mesh_from_obj_1000pts.ply.\nFinally, help me filter the noise in the output path of the previous step. Use statistical removal with 25 neighbors and 2.5 std_ratio. The output should be saved at ./outputs/clean_nb25_std2.5.ply.",
"file_path": "./data/source/singleview/object/scan10.jpg",
"ground_truth": [
{
"task": "Single-View 3D Reconstruction",
"usage": "object",
"quality": null,
"speed": "Medium speed",
"conf_thres": null,
"output_format": "ply",
"output_path": "./outputs/scan10.ply",
"extra_info": null
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 1000,
"output_path": "./outputs/example_mesh_from_obj_1000pts.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 25,
"std_ratio": 2.5,
"output_path": "./outputs/clean_nb25_std2.5.ply"
}
],
"gt_summary": " Single-View 3D Reconstruction, Post Processing, Post Processing \n Load image {input_file_path} and reconstruct it into 3D asset. The reconstruction speed should be medium speed. Save the output as a ply file. Save the output to ./outputs/scan10.ply.\nThen, please process the mesh at the output path of the previous step to create a point cloud. Use a sampling rate of 1000 points and write the file to ./outputs/example_mesh_from_obj_1000pts.ply.\nFinally, help me filter the noise in the output path of the previous step. Use statistical removal with 25 neighbors and 2.5 std_ratio. The output should be saved at ./outputs/clean_nb25_std2.5.ply. \n Input 1:./data/source/singleview/object/scan10.jpg; Input 2:./outputs/scan10.ply; Input 3:./outputs/example_mesh_from_obj_1000pts.ply; \n Step 1: { Constraints: [speed: Medium speed, output_format: ply]; Output Path: ./outputs/scan10.ply }; Step 2: { number_of_points: 1000; Output Path: ./outputs/example_mesh_from_obj_1000pts.ply }; Step 3: { nb_neighbors: 25, std_ratio: 2.5, Output Path: ./outputs/clean_nb25_std2.5.ply } ",
"code": "self.executor.single_view_reconstruction(image_path='./data/source/singleview/object/scan10.jpg', output_path='./outputs/scan10.ply', model='trellis')\nself.executor.post_processing(input_path='./outputs/scan10.ply', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_obj_1000pts.ply', parameters={\"number_of_points\": 1000})\nself.executor.post_processing(input_path='./outputs/example_mesh_from_obj_1000pts.ply', process_type='outliers_removal', output_path='./outputs/clean_nb25_std2.5.ply', parameters={\"nb_neighbors\": 25, \"std_ratio\": 2.5})",
"id": 113,
"ambiguous": "Do something with {input_file_path}.\nThen, Convert the file at the output of the previous step.\nFinally, Process the output of the previous step.",
"missing_info": "Load image and reconstruct it into 3D asset. The generation speed should be specified. Save the output to a file. The result should be saved in a location.\nThen, Please process the mesh at a file to create a point cloud. Use a sampling rate of a number of points and write the file to somewhere.\nFinally, Help me filter the noise in a location. Use statistical removal with certain neighbors and a std_ratio. The output should be saved at a path.",
"mistake": "Load image {input_file_path} and reconstruct it into 3D asset. The reconstruction speed should be medium speed. Save the output as a ply file. Save the output to ./outputs/scan10.ply.\nThen, please process the mesh at the output path of the previous step to create a point cloud. Use a sampling rate of 1000 points and write the file to ./outputs/example_mesh_from_obj_1000pts.ply.\nFinally, help me filter the noise in the output path of the previous step. Use statistical removal with 25 neighbors and 2.5 std_ratio. The output should be saved at ./outputs/clean_nb25_std2.5.ply. Ensure extremely high speed.",
"mistake_type": "quality_speed"
},
{
"complete": "Process the single-view image located at {input_file_path} to create a 3D representation. Expect medium speed processing. I need a glb file. Store the result at ./outputs/scan11.glb.\nThen, convert the mesh file at the output path of the previous step into a point cloud by sampling 10000 points. Save the result to ./outputs/example_mesh_from_obj_10000pts.ply.\nFinally, help me filter the noise in the output path of the previous step. Use statistical removal with 20 neighbors and 1.4 std_ratio. The output should be saved at ./outputs/clean_nb20_std1.4.ply.",
"file_path": "./data/source/singleview/object/scan11.jpg",
"ground_truth": [
{
"task": "Single-View 3D Reconstruction",
"usage": "object",
"quality": null,
"speed": "Medium speed",
"conf_thres": null,
"output_format": "glb",
"output_path": "./outputs/scan11.glb",
"extra_info": null
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 10000,
"output_path": "./outputs/example_mesh_from_obj_10000pts.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 20,
"std_ratio": 1.4,
"output_path": "./outputs/clean_nb20_std1.4.ply"
}
],
"gt_summary": " Single-View 3D Reconstruction, Post Processing, Post Processing \n Process the single-view image located at {input_file_path} to create a 3D representation. Expect medium speed processing. I need a glb file. Store the result at ./outputs/scan11.glb.\nThen, convert the mesh file at the output path of the previous step into a point cloud by sampling 10000 points. Save the result to ./outputs/example_mesh_from_obj_10000pts.ply.\nFinally, help me filter the noise in the output path of the previous step. Use statistical removal with 20 neighbors and 1.4 std_ratio. The output should be saved at ./outputs/clean_nb20_std1.4.ply. \n Input 1:./data/source/singleview/object/scan11.jpg; Input 2:./outputs/scan11.glb; Input 3:./outputs/example_mesh_from_obj_10000pts.ply; \n Step 1: { Constraints: [speed: Medium speed, output_format: glb]; Output Path: ./outputs/scan11.glb }; Step 2: { number_of_points: 10000; Output Path: ./outputs/example_mesh_from_obj_10000pts.ply }; Step 3: { nb_neighbors: 20, std_ratio: 1.4, Output Path: ./outputs/clean_nb20_std1.4.ply } ",
"code": "self.executor.single_view_reconstruction(image_path='./data/source/singleview/object/scan11.jpg', output_path='./outputs/scan11.glb', model='trellis')\nself.executor.post_processing(input_path='./outputs/scan11.glb', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_obj_10000pts.ply', parameters={\"number_of_points\": 10000})\nself.executor.post_processing(input_path='./outputs/example_mesh_from_obj_10000pts.ply', process_type='outliers_removal', output_path='./outputs/clean_nb20_std1.4.ply', parameters={\"nb_neighbors\": 20, \"std_ratio\": 1.4})",
"id": 114,
"ambiguous": "Do something with {input_file_path}.\nThen, Convert the file at the output of the previous step.\nFinally, Fix the noise in the output of the previous step.",
"missing_info": "Reconstruct a 3D asset from the image. Expect appropriate processing speed. Save the output to a file. Save the output to somewhere.\nThen, Take the mesh at a location and generate a point cloud with a certain number of samples. Store the output at a path.\nFinally, Clean the point cloud file at a path. I want to remove outliers with nb_neighbors set to a number and standard deviation ratio set to a value. Output result to a location.",
"mistake": "Process the single-view image located at ./data/source/singleview/object/scan11.stl to create a 3D representation. Expect medium speed processing. I need a glb file. Store the result at ./outputs/scan11.glb.\nThen, convert the mesh file at the output path of the previous step into a point cloud by sampling 10000 points. Save the result to ./outputs/example_mesh_from_obj_10000pts.ply.\nFinally, help me filter the noise in the output path of the previous step. Use statistical removal with 20 neighbors and 1.4 std_ratio. The output should be saved at ./outputs/clean_nb20_std1.4.ply.",
"mistake_type": "input_format"
},
{
"complete": "Load image {input_file_path} and reconstruct it into 3D asset. It should be a mesh with detailed geometry. Save the output to ./outputs/scan118.glb.\nThen, i need to sample 45000 points uniformly from the 3D model at the output path of the previous step. Output the processed ply file to ./outputs/example_mesh_from_glb_45000pts.ply.\nFinally, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 20 and a std_ratio of 1.2, then save to ./outputs/clean_nb20_std1.2.ply.",
"file_path": "./data/source/singleview/object/scan118.jpg",
"ground_truth": [
{
"task": "Single-View 3D Reconstruction",
"usage": "object",
"quality": null,
"speed": null,
"conf_thres": null,
"output_format": null,
"output_path": "./outputs/scan118.glb",
"extra_info": "Mesh with detailed geometry"
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 45000,
"output_path": "./outputs/example_mesh_from_glb_45000pts.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 20,
"std_ratio": 1.2,
"output_path": "./outputs/clean_nb20_std1.2.ply"
}
],
"gt_summary": " Single-View 3D Reconstruction, Post Processing, Post Processing \n Load image {input_file_path} and reconstruct it into 3D asset. It should be a mesh with detailed geometry. Save the output to ./outputs/scan118.glb.\nThen, i need to sample 45000 points uniformly from the 3D model at the output path of the previous step. Output the processed ply file to ./outputs/example_mesh_from_glb_45000pts.ply.\nFinally, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 20 and a std_ratio of 1.2, then save to ./outputs/clean_nb20_std1.2.ply. \n Input 1:./data/source/singleview/object/scan118.jpg; Input 2:./outputs/scan118.glb; Input 3:./outputs/example_mesh_from_glb_45000pts.ply; \n Step 1: { Constraints: [features: Mesh with detailed geometry]; Output Path: ./outputs/scan118.glb }; Step 2: { number_of_points: 45000; Output Path: ./outputs/example_mesh_from_glb_45000pts.ply }; Step 3: { nb_neighbors: 20, std_ratio: 1.2, Output Path: ./outputs/clean_nb20_std1.2.ply } ",
"code": "self.executor.single_view_reconstruction(image_path='./data/source/singleview/object/scan118.jpg', output_path='./outputs/scan118.glb', model='hunyuan')\nself.executor.post_processing(input_path='./outputs/scan118.glb', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_glb_45000pts.ply', parameters={\"number_of_points\": 45000})\nself.executor.post_processing(input_path='./outputs/example_mesh_from_glb_45000pts.ply', process_type='outliers_removal', output_path='./outputs/clean_nb20_std1.2.ply', parameters={\"nb_neighbors\": 20, \"std_ratio\": 1.2})",
"id": 115,
"ambiguous": "Fix {input_file_path} to 3D.\nThen, Do something with the output of the previous step.\nFinally, Do something with the output of the previous step.",
"missing_info": "Load image and reconstruct it into 3D asset. Consider that this has particular characteristics. The result should be saved in a location.\nThen, Please process the mesh at a file to create a point cloud. Use a sampling rate of a number of points and write the file to somewhere.\nFinally, Perform post-processing on a file to remove noise. Use a neighbor count of a number and a std_ratio of a value, then save to somewhere.",
"mistake": "Load image {input_file_path} and reconstruct it into 3D asset. It should be a mesh with detailed geometry. Save the output to ./outputs/scan118.glb.\nThen, i need to sample 45000 points uniformly from the 3D model at the output path of the previous step. Output the processed ply file to ./outputs/example_mesh_from_glb_45000pts.ply.\nFinally, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 20 and a std_ratio of 1.2, then save to ./outputs/clean_nb20_std1.2.ply. Ensure mesh with simplified geometry.",
"mistake_type": "quality_speed"
},
{
"complete": "Reconstruct a 3D asset from the image at {input_file_path}. Ensure the mesh is ultra-high quality. The reconstruction speed should be slow speed. Export format: glb. Save the output to ./outputs/scan10.glb.\nThen, perform mesh-to-pointcloud conversion on the mesh at the output path of the previous step. Set the target number of points to 15000 and save as ./outputs/example_mesh_from_glb_15000pts.ply.\nFinally, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 40 neighbors and a std_ratio of 1.3. Save the clean cloud to ./outputs/clean_nb40_std1.3.ply.",
"file_path": "./data/source/singleview/object/scan10.jpg",
"ground_truth": [
{
"task": "Single-View 3D Reconstruction",
"usage": "object",
"quality": "Ultra-High quality",
"speed": "Slow speed",
"conf_thres": null,
"output_format": "glb",
"output_path": "./outputs/scan10.glb",
"extra_info": null
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 15000,
"output_path": "./outputs/example_mesh_from_glb_15000pts.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 40,
"std_ratio": 1.3,
"output_path": "./outputs/clean_nb40_std1.3.ply"
}
],
"gt_summary": " Single-View 3D Reconstruction, Post Processing, Post Processing \n Reconstruct a 3D asset from the image at {input_file_path}. Ensure the mesh is ultra-high quality. The reconstruction speed should be slow speed. Export format: glb. Save the output to ./outputs/scan10.glb.\nThen, perform mesh-to-pointcloud conversion on the mesh at the output path of the previous step. Set the target number of points to 15000 and save as ./outputs/example_mesh_from_glb_15000pts.ply.\nFinally, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 40 neighbors and a std_ratio of 1.3. Save the clean cloud to ./outputs/clean_nb40_std1.3.ply. \n Input 1:./data/source/singleview/object/scan10.jpg; Input 2:./outputs/scan10.glb; Input 3:./outputs/example_mesh_from_glb_15000pts.ply; \n Step 1: { Constraints: [quality: Ultra-High quality, speed: Slow speed, output_format: glb]; Output Path: ./outputs/scan10.glb }; Step 2: { number_of_points: 15000; Output Path: ./outputs/example_mesh_from_glb_15000pts.ply }; Step 3: { nb_neighbors: 40, std_ratio: 1.3, Output Path: ./outputs/clean_nb40_std1.3.ply } ",
"code": "self.executor.single_view_reconstruction(image_path='./data/source/singleview/object/scan10.jpg', output_path='./outputs/scan10.glb', model='hunyuan')\nself.executor.post_processing(input_path='./outputs/scan10.glb', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_glb_15000pts.ply', parameters={\"number_of_points\": 15000})\nself.executor.post_processing(input_path='./outputs/example_mesh_from_glb_15000pts.ply', process_type='outliers_removal', output_path='./outputs/clean_nb40_std1.3.ply', parameters={\"nb_neighbors\": 40, \"std_ratio\": 1.3})",
"id": 116,
"ambiguous": "Do something with {input_file_path}.\nThen, Do something with the output of the previous step.\nFinally, Do something with the output of the previous step.",
"missing_info": "Load image and reconstruct it into 3D asset. Ensure the result quality is defined. Expect appropriate processing speed. I need a file. Store the generated model.\nThen, Please process the mesh at a file to create a point cloud. Use a sampling rate of a number of points and write the file to somewhere.\nFinally, Help me filter the noise in a location. Use statistical removal with certain neighbors and a std_ratio. The output should be saved at a path.",
"mistake": "Reconstruct a 3D asset from the image at ./data/source/singleview/object/scan10.stl. Ensure the mesh is ultra-high quality. The reconstruction speed should be slow speed. Export format: glb. Save the output to ./outputs/scan10.glb.\nThen, perform mesh-to-pointcloud conversion on the mesh at the output path of the previous step. Set the target number of points to 15000 and save as ./outputs/example_mesh_from_glb_15000pts.ply.\nFinally, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 40 neighbors and a std_ratio of 1.3. Save the clean cloud to ./outputs/clean_nb40_std1.3.ply.",
"mistake_type": "input_format"
},
{
"complete": "Perform single-view 3D reconstruction using {input_file_path} as input. Export format: glb. It should be a mesh with detailed geometry. Save the output to ./outputs/scan110.glb.\nThen, take the mesh at the output path of the previous step and generate a point cloud with 35000 samples. Store the output at ./outputs/example_mesh_from_ply_35000pts.ply.\nFinally, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 50 neighbors and a std_ratio of 1.8. Save the clean cloud to ./outputs/clean_nb50_std1.8.ply.",
"file_path": "./data/source/singleview/object/scan110.jpg",
"ground_truth": [
{
"task": "Single-View 3D Reconstruction",
"usage": "object",
"quality": null,
"speed": null,
"conf_thres": null,
"output_format": "glb",
"output_path": "./outputs/scan110.glb",
"extra_info": "Mesh with detailed geometry"
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 35000,
"output_path": "./outputs/example_mesh_from_ply_35000pts.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 50,
"std_ratio": 1.8,
"output_path": "./outputs/clean_nb50_std1.8.ply"
}
],
"gt_summary": " Single-View 3D Reconstruction, Post Processing, Post Processing \n Perform single-view 3D reconstruction using {input_file_path} as input. Export format: glb. It should be a mesh with detailed geometry. Save the output to ./outputs/scan110.glb.\nThen, take the mesh at the output path of the previous step and generate a point cloud with 35000 samples. Store the output at ./outputs/example_mesh_from_ply_35000pts.ply.\nFinally, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 50 neighbors and a std_ratio of 1.8. Save the clean cloud to ./outputs/clean_nb50_std1.8.ply. \n Input 1:./data/source/singleview/object/scan110.jpg; Input 2:./outputs/scan110.glb; Input 3:./outputs/example_mesh_from_ply_35000pts.ply; \n Step 1: { Constraints: [output_format: glb, features: Mesh with detailed geometry]; Output Path: ./outputs/scan110.glb }; Step 2: { number_of_points: 35000; Output Path: ./outputs/example_mesh_from_ply_35000pts.ply }; Step 3: { nb_neighbors: 50, std_ratio: 1.8, Output Path: ./outputs/clean_nb50_std1.8.ply } ",
"code": "self.executor.single_view_reconstruction(image_path='./data/source/singleview/object/scan110.jpg', output_path='./outputs/scan110.glb', model='hunyuan')\nself.executor.post_processing(input_path='./outputs/scan110.glb', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_ply_35000pts.ply', parameters={\"number_of_points\": 35000})\nself.executor.post_processing(input_path='./outputs/example_mesh_from_ply_35000pts.ply', process_type='outliers_removal', output_path='./outputs/clean_nb50_std1.8.ply', parameters={\"nb_neighbors\": 50, \"std_ratio\": 1.8})",
"id": 117,
"ambiguous": "Fix {input_file_path} to 3D.\nThen, Do something with the output of the previous step.\nFinally, Process the output of the previous step.",
"missing_info": "Process the single-view image to create a 3D representation. Consider that this has particular characteristics. Export format: 3D model. Store the generated model.\nThen, Perform mesh-to-pointcloud conversion on the mesh at a file. Set the target number of points to a number and save as somewhere.\nFinally, Clean the point cloud file at a path. I want to remove outliers with nb_neighbors set to a number and standard deviation ratio set to a value. Output result to a location.",
"mistake": "Perform single-view 3D reconstruction using ./data/source/singleview/object/scan110.glb as input. Export format: glb. It should be a mesh with detailed geometry. Save the output to ./outputs/scan110.glb.\nThen, take the mesh at the output path of the previous step and generate a point cloud with 35000 samples. Store the output at ./outputs/example_mesh_from_ply_35000pts.ply.\nFinally, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 50 neighbors and a std_ratio of 1.8. Save the clean cloud to ./outputs/clean_nb50_std1.8.ply.",
"mistake_type": "input_format"
},
{
"complete": "Process the single-view image located at {input_file_path} to create a 3D representation. Ensure the mesh is ultra-high quality. It should be a mesh with detailed geometry. The result should be saved in ./outputs/scan75.glb.\nThen, convert the mesh file at the output path of the previous step into a point cloud by sampling 1000 points. Save the result to ./outputs/example_mesh_from_ply_1000pts.ply.\nFinally, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 50 and a std_ratio of 2.5, then save to ./outputs/clean_nb50_std2.5.ply.",
"file_path": "./data/source/singleview/object/scan75.jpg",
"ground_truth": [
{
"task": "Single-View 3D Reconstruction",
"usage": "object",
"quality": "Ultra-High quality",
"speed": null,
"conf_thres": null,
"output_format": null,
"output_path": "./outputs/scan75.glb",
"extra_info": "Mesh with detailed geometry"
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 1000,
"output_path": "./outputs/example_mesh_from_ply_1000pts.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 50,
"std_ratio": 2.5,
"output_path": "./outputs/clean_nb50_std2.5.ply"
}
],
"gt_summary": " Single-View 3D Reconstruction, Post Processing, Post Processing \n Process the single-view image located at {input_file_path} to create a 3D representation. Ensure the mesh is ultra-high quality. It should be a mesh with detailed geometry. The result should be saved in ./outputs/scan75.glb.\nThen, convert the mesh file at the output path of the previous step into a point cloud by sampling 1000 points. Save the result to ./outputs/example_mesh_from_ply_1000pts.ply.\nFinally, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 50 and a std_ratio of 2.5, then save to ./outputs/clean_nb50_std2.5.ply. \n Input 1:./data/source/singleview/object/scan75.jpg; Input 2:./outputs/scan75.glb; Input 3:./outputs/example_mesh_from_ply_1000pts.ply; \n Step 1: { Constraints: [quality: Ultra-High quality, features: Mesh with detailed geometry]; Output Path: ./outputs/scan75.glb }; Step 2: { number_of_points: 1000; Output Path: ./outputs/example_mesh_from_ply_1000pts.ply }; Step 3: { nb_neighbors: 50, std_ratio: 2.5, Output Path: ./outputs/clean_nb50_std2.5.ply } ",
"code": "self.executor.single_view_reconstruction(image_path='./data/source/singleview/object/scan75.jpg', output_path='./outputs/scan75.glb', model='hunyuan')\nself.executor.post_processing(input_path='./outputs/scan75.glb', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_ply_1000pts.ply', parameters={\"number_of_points\": 1000})\nself.executor.post_processing(input_path='./outputs/example_mesh_from_ply_1000pts.ply', process_type='outliers_removal', output_path='./outputs/clean_nb50_std2.5.ply', parameters={\"nb_neighbors\": 50, \"std_ratio\": 2.5})",
"id": 118,
"ambiguous": "Fix {input_file_path} to 3D.\nThen, Convert the file at the output of the previous step.\nFinally, Do something with the output of the previous step.",
"missing_info": "Perform single-view 3D reconstruction. Ensure the result quality is defined. It should have specific attributes. The result should be saved in a location.\nThen, I need to sample a number of points uniformly from the 3D model at a path. Output the processed ply file to a location.\nFinally, Perform post-processing on a file to remove noise. Use a neighbor count of a number and a std_ratio of a value, then save to somewhere.",
"mistake": "Process the single-view image located at {input_file_path} to create a 3D representation. Ensure the mesh is ultra-high quality. It should be a mesh with detailed geometry. The result should be saved in ./outputs/scan75.glb.\nThen, convert the mesh file at the output path of the previous step into a point cloud by sampling 1000 points. Save the result to ./outputs/example_mesh_from_ply_1000pts.ply.\nFinally, perform post-processing on the output path of the previous step to remove noise. Use a neighbor count of 50 and a std_ratio of 2.5, then save to ./outputs/clean_nb50_std2.5.ply. Ensure mesh with simplified geometry. Ensure low quality.",
"mistake_type": "quality_speed"
},
{
"complete": "Load image {input_file_path} and reconstruct it into 3D asset. I need it processed with slow speed. Export format: obj. Save the output to ./outputs/scan62.obj.\nThen, take the mesh at the output path of the previous step and generate a point cloud with 50000 samples. Store the output at ./outputs/example_mesh_from_glb_50000pts.ply.\nFinally, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 45 neighbors and a std_ratio of 1.5. Save the clean cloud to ./outputs/clean_nb45_std1.5.ply.",
"file_path": "./data/source/singleview/object/scan62.jpg",
"ground_truth": [
{
"task": "Single-View 3D Reconstruction",
"usage": "object",
"quality": null,
"speed": "Slow speed",
"conf_thres": null,
"output_format": "obj",
"output_path": "./outputs/scan62.obj",
"extra_info": null
},
{
"task": "Post Processing",
"process_type": "mesh_to_pointcloud",
"number_of_points": 50000,
"output_path": "./outputs/example_mesh_from_glb_50000pts.ply"
},
{
"task": "Post Processing",
"process_type": "outlier_removal",
"nb_neighbors": 45,
"std_ratio": 1.5,
"output_path": "./outputs/clean_nb45_std1.5.ply"
}
],
"gt_summary": " Single-View 3D Reconstruction, Post Processing, Post Processing \n Load image {input_file_path} and reconstruct it into 3D asset. I need it processed with slow speed. Export format: obj. Save the output to ./outputs/scan62.obj.\nThen, take the mesh at the output path of the previous step and generate a point cloud with 50000 samples. Store the output at ./outputs/example_mesh_from_glb_50000pts.ply.\nFinally, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 45 neighbors and a std_ratio of 1.5. Save the clean cloud to ./outputs/clean_nb45_std1.5.ply. \n Input 1:./data/source/singleview/object/scan62.jpg; Input 2:./outputs/scan62.obj; Input 3:./outputs/example_mesh_from_glb_50000pts.ply; \n Step 1: { Constraints: [speed: Slow speed, output_format: obj]; Output Path: ./outputs/scan62.obj }; Step 2: { number_of_points: 50000; Output Path: ./outputs/example_mesh_from_glb_50000pts.ply }; Step 3: { nb_neighbors: 45, std_ratio: 1.5, Output Path: ./outputs/clean_nb45_std1.5.ply } ",
"code": "self.executor.single_view_reconstruction(image_path='./data/source/singleview/object/scan62.jpg', output_path='./outputs/scan62.obj', model='hunyuan')\nself.executor.post_processing(input_path='./outputs/scan62.obj', process_type='mesh_to_pointcloud', output_path='./outputs/example_mesh_from_glb_50000pts.ply', parameters={\"number_of_points\": 50000})\nself.executor.post_processing(input_path='./outputs/example_mesh_from_glb_50000pts.ply', process_type='outliers_removal', output_path='./outputs/clean_nb45_std1.5.ply', parameters={\"nb_neighbors\": 45, \"std_ratio\": 1.5})",
"id": 119,
"ambiguous": "Do something with {input_file_path}.\nThen, Do something with the output of the previous step.\nFinally, Do something with the output of the previous step.",
"missing_info": "Load image and reconstruct it into 3D asset. The generation speed should be specified. I need a file. Store the generated model.\nThen, I need to sample a number of points uniformly from the 3D model at a path. Output the processed ply file to a location.\nFinally, Perform post-processing on a file to remove noise. Use a neighbor count of a number and a std_ratio of a value, then save to somewhere.",
"mistake": "Load image {input_file_path} and reconstruct it into 3D asset. I need it processed with slow speed. Export format: obj. Save the output to ./outputs/scan62.obj.\nThen, take the mesh at the output path of the previous step and generate a point cloud with 50000 samples. Store the output at ./outputs/example_mesh_from_glb_50000pts.ply.\nFinally, i have a noisy point cloud at the output path of the previous step. Please apply statistical outlier removal using 45 neighbors and a std_ratio of 1.5. Save the clean cloud to ./outputs/clean_nb45_std1.5.ply. Ensure fast speed.",
"mistake_type": "quality_speed"
}
]