upload refined_40k/test_bench 's json, videos and frames

data/refined_40k/test_bench/data_bad_worsen_phy.json

@@ -0,0 +1,312 @@
+[
+    {
+        "id": "h104798",
+        "images": [
+            "h104798_00.jpg",
+            "h104798_01.jpg",
+            "h104798_02.jpg",
+            "h104798_03.jpg",
+            "h104798_04.jpg",
+            "h104798_05.jpg",
+            "h104798_06.jpg",
+            "h104798_07.jpg",
+            "h104798_08.jpg",
+            "h104798_09.jpg",
+            "h104798_10.jpg",
+            "h104798_11.jpg",
+            "h104798_12.jpg",
+            "h104798_13.jpg",
+            "h104798_14.jpg",
+            "h104798_15.jpg"
+        ],
+        "conversations": [
+            {
+                "from": "human",
+                "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"He was an ordinary man with an extraordinary curiosity,  \",\nall the frames of video are as follows: \n\n<image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> \n"
+            },
+            {
+                "from": "gpt",
+                "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 1\n factual consistency: 1\n"
+            }
+        ]
+    },
+    {
+        "id": "h205802",
+        "images": [
+            "h205802_00.jpg",
+            "h205802_01.jpg",
+            "h205802_02.jpg",
+            "h205802_03.jpg",
+            "h205802_04.jpg",
+            "h205802_05.jpg",
+            "h205802_06.jpg",
+            "h205802_07.jpg",
+            "h205802_08.jpg",
+            "h205802_09.jpg",
+            "h205802_10.jpg",
+            "h205802_11.jpg",
+            "h205802_12.jpg",
+            "h205802_13.jpg",
+            "h205802_14.jpg",
+            "h205802_15.jpg"
+        ],
+        "conversations": [
+            {
+                "from": "human",
+                "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Coffee burning in a crowded room in the wild wild west  \",\nall the frames of video are as follows: \n\n<image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> \n"
+            },
+            {
+                "from": "gpt",
+                "value": "visual quality: 2\n temporal consistency: 1\n dynamic degree: 2\n text-to-video alignment: 1\n factual consistency: 1\n"
+            }
+        ]
+    },
+    {
+        "id": "h802286",
+        "images": [
+            "h802286_00.jpg",
+            "h802286_01.jpg",
+            "h802286_02.jpg",
+            "h802286_03.jpg",
+            "h802286_04.jpg",
+            "h802286_05.jpg",
+            "h802286_06.jpg",
+            "h802286_07.jpg",
+            "h802286_08.jpg",
+            "h802286_09.jpg",
+            "h802286_10.jpg",
+            "h802286_11.jpg",
+            "h802286_12.jpg",
+            "h802286_13.jpg",
+            "h802286_14.jpg",
+            "h802286_15.jpg",
+            "h802286_16.jpg",
+            "h802286_17.jpg",
+            "h802286_18.jpg",
+            "h802286_19.jpg",
+            "h802286_20.jpg",
+            "h802286_21.jpg",
+            "h802286_22.jpg",
+            "h802286_23.jpg"
+        ],
+        "conversations": [
+            {
+                "from": "human",
+                "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"animation style of Studio Ghibli and Pixar, Young man sets off on an adventure with his 2 friends, walking away, camera zoom  \",\nall the frames of video are as follows: \n\n<image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> \n"
+            },
+            {
+                "from": "gpt",
+                "value": "visual quality: 1\n temporal consistency: 2\n dynamic degree: 2\n text-to-video alignment: 1\n factual consistency: 1\n"
+            }
+        ]
+    },
+    {
+        "id": "h005625",
+        "images": [
+            "h005625_00.jpg",
+            "h005625_01.jpg",
+            "h005625_02.jpg",
+            "h005625_03.jpg",
+            "h005625_04.jpg",
+            "h005625_05.jpg",
+            "h005625_06.jpg",
+            "h005625_07.jpg",
+            "h005625_08.jpg",
+            "h005625_09.jpg",
+            "h005625_10.jpg",
+            "h005625_11.jpg",
+            "h005625_12.jpg",
+            "h005625_13.jpg",
+            "h005625_14.jpg",
+            "h005625_15.jpg",
+            "h005625_16.jpg",
+            "h005625_17.jpg",
+            "h005625_18.jpg",
+            "h005625_19.jpg",
+            "h005625_20.jpg",
+            "h005625_21.jpg",
+            "h005625_22.jpg",
+            "h005625_23.jpg"
+        ],
+        "conversations": [
+            {
+                "from": "human",
+                "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"cinematic,happy girl on the beach,pixar style  \",\nall the frames of video are as follows: \n\n<image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> \n"
+            },
+            {
+                "from": "gpt",
+                "value": "visual quality: 3\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 1\n factual consistency: 1\n"
+            }
+        ]
+    },
+    {
+        "id": "h702083",
+        "images": [
+            "h702083_00.jpg",
+            "h702083_01.jpg",
+            "h702083_02.jpg",
+            "h702083_03.jpg",
+            "h702083_04.jpg",
+            "h702083_05.jpg",
+            "h702083_06.jpg",
+            "h702083_07.jpg"
+        ],
+        "conversations": [
+            {
+                "from": "human",
+                "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"a group of hikers reported an encounter that would forever change their perception of the supernatural  \",\nall the frames of video are as follows: \n\n<image> <image> <image> <image> <image> <image> <image> <image> \n"
+            },
+            {
+                "from": "gpt",
+                "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 1\n factual consistency: 1\n"
+            }
+        ]
+    },
+    {
+        "id": "h104844",
+        "images": [
+            "h104844_00.jpg",
+            "h104844_01.jpg",
+            "h104844_02.jpg",
+            "h104844_03.jpg",
+            "h104844_04.jpg",
+            "h104844_05.jpg",
+            "h104844_06.jpg",
+            "h104844_07.jpg",
+            "h104844_08.jpg",
+            "h104844_09.jpg",
+            "h104844_10.jpg",
+            "h104844_11.jpg",
+            "h104844_12.jpg",
+            "h104844_13.jpg",
+            "h104844_14.jpg",
+            "h104844_15.jpg"
+        ],
+        "conversations": [
+            {
+                "from": "human",
+                "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"On a top high above the Futuristic Architectur Building, a young boy stands with flame. In his hand, he jump on road. blck and white video.  \",\nall the frames of video are as follows: \n\n<image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> \n"
+            },
+            {
+                "from": "gpt",
+                "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 1\n"
+            }
+        ]
+    },
+    {
+        "id": "h603195",
+        "images": [
+            "h603195_00.jpg",
+            "h603195_01.jpg",
+            "h603195_02.jpg",
+            "h603195_03.jpg",
+            "h603195_04.jpg",
+            "h603195_05.jpg",
+            "h603195_06.jpg",
+            "h603195_07.jpg",
+            "h603195_08.jpg",
+            "h603195_09.jpg",
+            "h603195_10.jpg",
+            "h603195_11.jpg",
+            "h603195_12.jpg",
+            "h603195_13.jpg",
+            "h603195_14.jpg",
+            "h603195_15.jpg"
+        ],
+        "conversations": [
+            {
+                "from": "human",
+                "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"video who prrsent nfc cards and in this video peoples stikcs her cards to their phone  \",\nall the frames of video are as follows: \n\n<image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> \n"
+            },
+            {
+                "from": "gpt",
+                "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 1\n"
+            }
+        ]
+    },
+    {
+        "id": "h703494",
+        "images": [
+            "h703494_00.jpg",
+            "h703494_01.jpg",
+            "h703494_02.jpg",
+            "h703494_03.jpg",
+            "h703494_04.jpg",
+            "h703494_05.jpg",
+            "h703494_06.jpg",
+            "h703494_07.jpg"
+        ],
+        "conversations": [
+            {
+                "from": "human",
+                "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"drop of water in a rose with sun reflection  \",\nall the frames of video are as follows: \n\n<image> <image> <image> <image> <image> <image> <image> <image> \n"
+            },
+            {
+                "from": "gpt",
+                "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 1\n factual consistency: 1\n"
+            }
+        ]
+    },
+    {
+        "id": "h204576",
+        "images": [
+            "h204576_00.jpg",
+            "h204576_01.jpg",
+            "h204576_02.jpg",
+            "h204576_03.jpg",
+            "h204576_04.jpg",
+            "h204576_05.jpg",
+            "h204576_06.jpg",
+            "h204576_07.jpg",
+            "h204576_08.jpg",
+            "h204576_09.jpg",
+            "h204576_10.jpg",
+            "h204576_11.jpg",
+            "h204576_12.jpg",
+            "h204576_13.jpg",
+            "h204576_14.jpg",
+            "h204576_15.jpg"
+        ],
+        "conversations": [
+            {
+                "from": "human",
+                "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"On the mountain, watching the soldiers fighting below.  \",\nall the frames of video are as follows: \n\n<image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> \n"
+            },
+            {
+                "from": "gpt",
+                "value": "visual quality: 2\n temporal consistency: 1\n dynamic degree: 1\n text-to-video alignment: 2\n factual consistency: 1\n"
+            }
+        ]
+    },
+    {
+        "id": "h304430",
+        "images": [
+            "h304430_00.jpg",
+            "h304430_01.jpg",
+            "h304430_02.jpg",
+            "h304430_03.jpg",
+            "h304430_04.jpg",
+            "h304430_05.jpg",
+            "h304430_06.jpg",
+            "h304430_07.jpg",
+            "h304430_08.jpg",
+            "h304430_09.jpg",
+            "h304430_10.jpg",
+            "h304430_11.jpg",
+            "h304430_12.jpg",
+            "h304430_13.jpg",
+            "h304430_14.jpg",
+            "h304430_15.jpg"
+        ],
+        "conversations": [
+            {
+                "from": "human",
+                "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Cartoon style, people live with robots.  \",\nall the frames of video are as follows: \n\n<image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> \n"
+            },
+            {
+                "from": "gpt",
+                "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 2\n factual consistency: 1\n"
+            }
+        ]
+    }
+]