detection-datacollection 0.1.0

A tool for collecting object detection datasets with OpenCV

Detection Data Collection

A Python tool for collecting object detection datasets with bounding boxes using OpenCV. It allows you to draw bounding boxes on live camera/video feed and automatically saves annotated images + YOLO-format labels.

---

🚀 Installation

pip install detection-datacollection

Or install directly from GitHub:

pip install git+https://github.com/ShashwatDev-26/imageDataCollection.git

---

📦 Dependencies

This package requires:

• numpy==2.2.6
• opencv-python==4.12.0.88
• PyYAML==6.0.2

They are automatically installed with pip.

---

🔧 Usage

1. Import and initialize

from detectionDataCollection import detectionDataCollection
# Create a collector instance
collector = detectionDataCollection()
# Use a video file or camera index (0 for default webcam)
collector.set_sourceID(0)  # or "video.mp4"
# Number of samples to capture per class
collector.set_nSamples(10)
# Initialize camera
collector.camera_init_()
# Start annotation mode
collector.annotation()

2. Controls

• CTRL + Left Click → Start drawing bounding box
• Right Click → Add object/class
• CTRL + Right Click → Start capturing samples
• ESC → Stop and save dataset

---

📂 Output Structure

When you stop annotation (ESC), the dataset is stored like:

detectionDataset/
├── train/
│   ├── images/
│   │   ├── frame_000_000.jpg
│   │   └── ...
│   ├── labels/
│   │   ├── frame_000_000.txt
│   │   └── ...
├── data.yaml

The labels follow YOLO format:

class_id x_center y_center width height

---

🎥

🌍 Links

• 📦 PyPI: detection-datacollection
• 💻 Source Code: GitHub
• 🐛 Issues: Report bugs here

---

📜 License

This project is licensed under the MIT License.

To be continued ....

---