README.md

---
dataset_info:
  features:
  - name: image
    dtype: image
  - name: objects
    dtype: string
  - name: annotated_image
    dtype: image
  splits:
  - name: train
    num_bytes: 6891140314.688933
    num_examples: 14911
  download_size: 6870251562
  dataset_size: 6891140314.688933
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
tags:
- object-detection
- vision
- image
- bounding-boxes
- multimodal
- detection
- machine-learning
- deep-learning
- coco-format
- open-vocabulary-detection
- auto-annotation
- vlm
license: apache-2.0
task_categories:
- object-detection
language:
- en
pretty_name: OpenDetection
size_categories:
- 10K<n<100K
---

## **OpenDetection-15K-Dense-v2.0**

**OpenDetection-15K-Dense-v2.0** is an object detection dataset built primarily from general, publicly available images, which make up the majority of the input imagery, together with additional publicly available datasets. The dataset contains high-quality object detection annotations generated using modern automated computer vision pipelines, providing structured object labels, confidence scores, and bounding box coordinates for every detected object. Each sample includes the original image, machine-readable detection annotations, and a rendered visualization with all detected objects overlaid on the image. Version 2.0 contains a different curated subset of images while maintaining the same annotation format and dataset structure, making it suitable for training, evaluation, benchmarking, and multimodal computer vision research. The dataset is distributed using the Hugging Face Datasets format with optimized Parquet files for efficient loading and large-scale machine learning workflows.

## Dataset Statistics

| Property | Value |
|-----------|-------|
| Number of Samples | 14,911 |
| Image Format | RGB |
| Annotation Format | JSON |
| Visualization | Annotated Image |
| Dataset Format | Optimized Parquet |

## Dataset Structure

Each sample contains the following fields:

| Column | Type | Description |
|---------|------|-------------|
| `image` | Image | Original input image |
| `objects` | List | Object detection annotations containing labels, label IDs, confidence scores, and bounding boxes |
| `annotated_image` | Image | Visualization of the image with rendered bounding boxes |

Example:

```python
sample = ds[0]

print(sample.keys())

# dict_keys([
#     "image",
#     "objects",
#     "annotated_image"
# ])
```

## Loading the Dataset

```python
from datasets import load_dataset

ds = load_dataset(
    "prithivMLmods/OpenDetection-15K-Dense-v2.0",
    split="train"
)
```

## Example Usage

```python
from datasets import load_dataset
import matplotlib.pyplot as plt

ds = load_dataset(
    "prithivMLmods/OpenDetection-15K-Dense-v2.0",
    split="train"
)

sample = ds[0]

image = sample["image"]
objects = sample["objects"]
annotated = sample["annotated_image"]

print("Detected Objects:")
print(objects)

fig, axes = plt.subplots(1, 2, figsize=(12, 6))

axes[0].imshow(image)
axes[0].set_title("Image")
axes[0].axis("off")

axes[1].imshow(annotated)
axes[1].set_title("Annotated Image")
axes[1].axis("off")

plt.show()
```

## Dataset Features

- High-quality object detection annotations
- Structured JSON object annotations
- Bounding box coordinates for every detected object
- Confidence scores for each detection
- Annotated visualization for every image
- Optimized Parquet format for efficient loading
- Compatible with the Hugging Face Datasets library
- Suitable for object detection training, evaluation, benchmarking, and multimodal computer vision research

## License

This dataset is released under the **Apache-2.0 License**.