Create dataset_infos.json

dataset_infos.json

@@ -0,0 +1,38 @@
+{
+  "IDEA-Bench": {
+    "description": "A benchmark dataset for professional-grade image generation evaluations.",
+    "citation": "@article{idea_bench2024, title={IDEA-Bench: How Far are Generative Models from Professional Designing?}, author={Chen Liang et al.}, journal={arXiv preprint}, year={2024}}",
+    "homepage": "https://github.com/ali-vilab/IDEA-Bench",
+    "license": "cc-by-nc 4.0",
+    "features": {
+      "uid": {"dtype": "string", "description": "Unique identifier for the task."},
+      "instruction": {"dtype": "string", "description": "Instruction prompt for the task."},
+      "meta": {
+        "task_name": {"dtype": "string", "description": "The name of the task."},
+        "num_of_cases": {"dtype": "int32", "description": "Number of cases in the task."},
+        "image_reference": {"dtype": "bool", "description": "Whether the task uses image references."},
+        "multi_image_reference": {"dtype": "bool", "description": "Whether the task uses multiple image references."},
+        "multi_image_output": {"dtype": "bool", "description": "Whether the task generates multiple images."},
+        "output_image_count": {"dtype": "int32", "description": "Number of expected output images."}
+      },
+      "eval": {
+        "questions": {
+          "dtype": "list",
+          "description": "List of evaluation questions with scoring standards."
+        }
+      },
+      "images": {"dtype": "list", "description": "List of associated images."},
+      "auto_eval": {
+        "dtype": "list",
+        "description": "Detailed prompt for MLLM evaluation, if available."
+      }
+    },
+    "splits": {
+      "text-to-image": {"description": "Text-to-Image tasks."},
+      "image-to-image": {"description": "Image-to-Image"},
+      "images-to-image": {"description": "Images-to-Image tasks."},
+      "text-to-images": {"description": "Text-to-Images tasks."},
+      "images-to-images": {"description": "Image(s)-to-Images tasks."}
+    }
+  }
+}