Dataset · phiyodr
InpaintCOCO for fine-grained understanding of color, size, and objects
phiyodr/InpaintCOCO
Multimodal COCO-based dataset with images and English texts for evaluating and training models on fine details.
What's inside
The dataset contains 1,000 to 10,000 examples (images and text) in English. It is based on COCO and focuses on detailed understanding of multimodal concepts such as color, size, and specific objects. The card also mentions the existence of a multilingual version, MultiInpaintCOCO, in 14 languages including Russian.
What you can do with it
- Evaluation and fine-tuning of models linking text and image (image-to-text, text-to-image)
- Testing VLM sensitivity to object attributes (color and size)
- Testing image classification models based on text descriptions
- Tasks
- Image → text, Text → image, Image classification
- Languages
- English
- Size
- 1K–10K rows
- Downloads this month
- 6.7K
- Likes
- 7
How to load it
from datasets import load_dataset
ds = load_dataset("phiyodr/InpaintCOCO")