README.md

---
dataset_info:
  features:
  - name: image
    dtype: image
  - name: objects
    dtype: string
  - name: annotated_image
    dtype: image
  splits:
  - name: train
    num_bytes: 32518816505.639652
    num_examples: 52557
  download_size: 32487827237
  dataset_size: 32518816505.639652
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
tags:
- computer-vision
- 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
size_categories:
- 10K<n<100K
pretty_name: OpenDetection
---

## **OpenDetection-50K-Remastered-Cleaned**

**OpenDetection-50K-Remastered-Cleaned** is the cleaned version of **[OpenDetection-50K-Remastered](https://huggingface.co/datasets/prithivMLmods/OpenDetection-50K-Remastered)**, created by removing every sample that contains no detected objects. This ensures that every image in the dataset includes at least one valid object annotation, making the dataset more suitable for training, evaluation, and benchmarking object detection models. The dataset is built primarily from general, publicly available images, which make up the majority of the input imagery, together with additional publicly available datasets. Each sample contains the original image, structured object detection annotations including class labels, confidence scores, and bounding boxes, along with a rendered visualization showing all detected objects. The dataset is distributed using the Hugging Face Datasets format with optimized Parquet files for efficient loading and large-scale training workflows.

> **52,557 rows (81 rows with null or empty `objects` were removed) from `prithivMLmods/OpenDetection-50K-Remastered`.**

## Dataset Statistics

| Property | Value |
|-----------|-------|
| Number of Samples | 52,557 |
| 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, confidence scores, label IDs, 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-50K-Remastered-Cleaned",
    split="train"
)
```

## Example Usage

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

ds = load_dataset(
    "prithivMLmods/OpenDetection-50K-Remastered-Cleaned",
    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

- Cleaned version of OpenDetection-50K-Remastered
- All empty annotations removed
- Every image contains at least one valid object
- High-quality object detection annotations
- Bounding box visualizations for every sample
- Optimized Parquet format for efficient loading
- Compatible with the Hugging Face Datasets library
- Suitable for training, evaluation, benchmarking, and multimodal computer vision research

## License

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