OpenCaption-Unified-10K

## **OpenCaption-Unified-10K** **OpenCaption-Unified-10K** is a dense image captioning dataset built from **10,000 images** paired with long-form synthetic captions generated using the **Qwen3.5** multimodal model. Each caption is produced through a dedicated Qwen3.5 captioning pipeline designed to generate detailed, high-fidelity descriptions of scene composition, subject attributes, spatial relationships, activities, and overall visual context rather than short, generic captions. The dataset is intended for training and evaluating image-to-text, vision-language, and dense captioning models that require rich textual grounding. The dataset is built primarily from **publicly available images**, which make up the majority of the input imagery, together with additional publicly available datasets. Every image is paired with a single comprehensive caption that attempts to describe the complete visual scene, making the dataset suitable for supervised vision-language training, instruction tuning, caption refinement, retrieval, and multimodal research. ## Dataset Statistics | Property | Value | |-----------|-------| | Number of Samples | 10,000 | | Annotation Type | Long-form Dense Caption | | Caption Generator | Qwen3.5 Multimodal | | Dataset Format | Optimized Parquet | ## Dataset Structure Each sample contains the following fields: | Column | Type | Description | |---------|------|-------------| | `image` | Image | Original input image | | `response` | String | Long-form dense caption describing the image | Example: ```python sample = ds[0] print(sample.keys()) # dict_keys([ # "image", # "response" # ]) ``` ## Loading the Dataset ```python from datasets import load_dataset dataset = load_dataset( "prithivMLmods/OpenCaption-Unified-10K", split="train" ) ``` ## Example Usage ```python from datasets import load_dataset import matplotlib.pyplot as plt dataset = load_dataset( "prithivMLmods/OpenCaption-Unified-10K", split="train" ) sample = dataset[0] image = sample["image"] caption = sample["response"] print(caption) plt.figure(figsize=(8, 8)) plt.imshow(image) plt.axis("off") plt.show() ``` ## Citation ```bibtex @misc{prithiv_sakthi_2026, author = { Prithiv Sakthi }, title = { OpenCaption-Unified-10K (Revision 75cba9f) }, year = 2026, url = { https://huggingface.co/datasets/prithivMLmods/OpenCaption-Unified-10K }, doi = { 10.57967/hf/9578 }, publisher = { Hugging Face } } ``` ## License This dataset is released under the **Apache-2.0 License**.