Datasets:
Tasks:
Visual Question Answering
Modalities:
Image
Formats:
imagefolder
Languages:
English
Size:
1K - 10K
License:
Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
|
@@ -1,3 +1,46 @@
|
|
| 1 |
-
---
|
| 2 |
-
license:
|
| 3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: cc-by-3.0
|
| 3 |
+
task_categories:
|
| 4 |
+
- visual-question-answering
|
| 5 |
+
language:
|
| 6 |
+
- en
|
| 7 |
+
tags:
|
| 8 |
+
- vision-language
|
| 9 |
+
- benchmark
|
| 10 |
+
- domain-specific
|
| 11 |
+
pretty_name: DomainBench
|
| 12 |
+
size_categories:
|
| 13 |
+
- 10K<n<100K
|
| 14 |
+
---
|
| 15 |
+
|
| 16 |
+
# DomainBench
|
| 17 |
+
|
| 18 |
+
DomainBench is a comprehensive benchmark for evaluating vision-language models across different domains.
|
| 19 |
+
|
| 20 |
+
## Dataset Structure
|
| 21 |
+
|
| 22 |
+
The benchmark consists of 7 domain-specific datasets:
|
| 23 |
+
- Animals
|
| 24 |
+
- Food
|
| 25 |
+
- Kitchen
|
| 26 |
+
- KITTI
|
| 27 |
+
- Person
|
| 28 |
+
- Vehicles
|
| 29 |
+
- Wildlife
|
| 30 |
+
|
| 31 |
+
Each domain contains:
|
| 32 |
+
- Images folder with visual content
|
| 33 |
+
- Question-answer pairs in JSON format
|
| 34 |
+
|
| 35 |
+
## Tasks
|
| 36 |
+
|
| 37 |
+
The benchmark includes various visual reasoning tasks:
|
| 38 |
+
1. Object Detection
|
| 39 |
+
2. Spatial Relationships
|
| 40 |
+
3. Color Analysis
|
| 41 |
+
4. Depth Perception
|
| 42 |
+
5. Object Interaction
|
| 43 |
+
6. Puzzle Solving
|
| 44 |
+
7. Image Variant Analysis
|
| 45 |
+
8. Object Condition Analysis
|
| 46 |
+
|