Skip to content

Latest commit

 

History

History
82 lines (64 loc) · 3.19 KB

README.md

File metadata and controls

82 lines (64 loc) · 3.19 KB

Object Detection and Segmentation Script

This script enables object detection and segmentation on images and videos using YOLO and a segmentation model. It provides functionalities to process images/videos and display the results in real-time.

Demo Video - Youtube Link

Features

  • YOLO Object Detection: Detect objects using a trained YOLO model.
  • Segmentation Model: Perform image segmentation with a TensorFlow Keras model.
  • Drivable Area Detection: Identify drivable areas using YOLOPv2.
  • Supports Image & Video Processing: Apply models to both images and videos.
  • Real-time Processing: Uses OpenCV to visualize processed frames in real-time.

Requirements

Ensure you have the following dependencies installed:

pip install opencv-python numpy pillow tensorflow torch ultralytics
  • Note - Install Tensorflow and Torch according to your Cuda version.

Usage

Run the script with appropriate arguments to process an image or video.

Arguments

  • --mode: Choose between img (image) or video (video) processing.
  • --source: Path to the image or video file (or camera index for live feed).
  • --model: Choose between yolo, seg, or drivable.

Cuda is compulsary, if not avialable, use "--devices cpu" (not recommended too slow for segmentation).

Running Image Processing

python main.py --mode img --source path/to/image.jpg --model yolo

Running Video Processing

python main.py --mode video --source path/to/video.mp4 --model seg

Running Drivable Area Detection

python main.py --mode video --source path/to/video.mp4 --model drivable

Functions Overview

run_inference(frame, model_name)

  • Rotates the input frame.
  • Runs inference using either the YOLO model (yolo_model) or segmentation model (seg_model).
  • Returns the processed frame with bounding boxes or segmentation mask.

process_video(source, model)

  • Captures frames from a video source.
  • Processes each frame using the selected model.
  • Displays the output in a real-time window.

process_image(source, model)

  • Reads an image from the specified path.
  • Runs inference using the chosen model.
  • Displays the processed image.

detect2(source)

  • Loads YOLOPv2 for drivable area and lane line detection.
  • Performs inference using the model.
  • Applies non-max suppression and segmentation masks.
  • Displays results in a real-time window.

Notes

  • Press q to exit the visualization windows.
  • Ensure the required models (seg.h5, best.pt, drivable.pt) are available in the working directory.
  • If you are seeing a rotated image, comment out line 25 in main.py.
  • You can download weights from this particular drive link :- Drive Link.
  • The weights file have to be in the same directory as of the "main.py".
  • Drivable model has not been tested without cuda.

License

This script is open-source and free to use. Modify it as needed for your projects.

Author

Developed by Bhavik Ostwal, Lavish Singal, Nishant Nehra, Piyush Roy. Team - Alt + F4