YOLO Object Detection Video Annotation

This repository contains code for annotating videos using the YOLOv10 object detection model. The application processes each frame of a video, detects objects, annotates them, and saves both the annotated video and a JSON file with the detection results.

Features

• Object Detection: Uses YOLOv10 to detect objects in each frame of a video.
• Video Processing: Reads a video file, annotates each frame with bounding boxes and labels, and writes the results to a new video file.
• Results Export: Saves detection results in a JSON file, including object classes, confidence scores, and bounding boxes.

Requirements

• Python 3.7 or later
• cv2 (OpenCV) - for video processing and annotation
• ultralytics - for YOLO object detection
• torch - for PyTorch backend

You can install the required Python packages using pip:

pip install opencv-python ultralytics torch

Usage

1. Prepare the Model:

   • Download the YOLOv10 model weights (best.pt) and place it in the project directory.

2. Prepare the Video:

   • Place your video file in the test_objects directory or update the video_path variable in the code to point to your video file.

3. Run the Script:

   Save the provided code in a file named videoResNativa.py and run:

   python videoResNativa.py

4. Outputs:

   • The annotated video will be saved as annotated_output.mp4.
   • Detection results for each frame will be saved in predictions.json.

Code Overview

1. Device Setup:

   • Checks if CUDA is available and sets the device to GPU or CPU accordingly.

2. Model Loading:

   • Loads the YOLOv10 model (best.pt) and moves it to the selected device.

3. Video Processing:

   • Opens the video file, retrieves video properties, and sets up a VideoWriter to save the annotated video.
   • Processes each frame, performs object detection, annotates the frame, and saves it to the output video.

4. Results Handling:

   • Collects detection results for each frame and saves them in a JSON file.

5. Resource Management:

   • Releases video capture and writer resources, and closes any OpenCV windows.

Configuration

• Model Path: Update the YOLO constructor argument with the path to your model weights if different from best.pt.
• Video Path: Update the video_path variable with the path to your input video file.
• Output Paths: Modify output_path and JSON file paths as needed.

Troubleshooting

• Ensure that the video file path is correct and the file is accessible.
• Make sure all required packages are installed and compatible with your Python version.
• Check the console output for any errors or warnings during execution.