README.md

---
license: mit
task_categories:
  - object-detection
  - zero-shot-object-detection
language:
  - en
tags:
  - object-detection
  - open-world-detection
  - grounding-dino
  - dense-object-detection
  - coco
  - fiftyone
  - voxel51
size_categories:
  - 1K<n<10K
pretty_name: Dense Object Detection Dataset (FiftyOne Format)
---

# Dense Object Detection Dataset (FiftyOne Format)

Dense Object Detection Dataset for open-world/open-vocabulary object detection models like Grounding DINO, OWL-ViT, and GLIP.

## Statistics
- **Images**: 8,001
- **Annotations**: 344,079
- **Categories**: 1,038

## Loading with FiftyOne

```python
import fiftyone as fo
import fiftyone.utils.huggingface as fouh

dataset = fouh.load_from_hub("shubh303/dense_object_detection_voxel51")

# Launch the app
session = fo.launch_app(dataset)
```

## Important
This dataset is for **open-world detection** models only. Not suitable for closed-set detectors (YOLO, DETR, etc.).

## License
MIT