Skip to content

Latest commit

 

History

History
100 lines (64 loc) · 2.55 KB

Session_5.md

File metadata and controls

100 lines (64 loc) · 2.55 KB

交大密院Deep learning学习手册

UM-SJTU JI Deep learning Hands-on Tutorial

Session 5 - Object Detection with YOLOv5 on Custom Video Data (Reduced Scope)

Table of Contents

Introduction

In this truncated session, we will focus solely on running YOLOv5 for object detection on a custom video and visualizing the results.

Prerequisites

To run the code, you'll need:

  • PyTorch
  • OpenCV (cv2)
  • YOLOv5 (from GitHub)

You can clone the YOLOv5 repository from GitHub:

git clone https://github.com/ultralytics/yolov5.git
cd yolov5
pip install -U -r requirements.txt

Data Preparation

Assume you have a video file named my_video.mp4. Place it in a directory of your choosing.

Object Detection with YOLOv5

Step 1: Run Object Detection on Video

Run the following command to perform object detection:

python detect.py --source /path/to/my_video.mp4 --weights yolov5s.pt --conf 0.4

Replace /path/to/my_video.mp4 with the path to your video.

This command will generate an output video with bounding boxes around detected objects.

Visualization

Step 2: View the Output Video

You can use any video player to view the output video generated by YOLOv5, or you can use OpenCV to visualize it programmatically.

Here's a Python code snippet that uses OpenCV to play the output video:

import cv2

# Initialize OpenCV video capture
video = cv2.VideoCapture('runs/detect/exp/your_output_video.mp4')  # Change the path to your output video

while video.isOpened():
    ret, frame = video.read()
    if not ret:
        print("Failed to grab frame or video has ended. Exiting.")
        break

    # Display the frame
    cv2.imshow('YOLOv5 Object Detection', frame)

    # Exit if 'q' is pressed
    if cv2.waitKey(1) & 0xFF == ord('q'):
        break

# Release the video and destroy OpenCV windows
video.release()
cv2.destroyAllWindows()

In this code, OpenCV will read the output video generated by YOLOv5 (your_output_video.mp4) and play it in a window. Press 'q' to close the window and stop the video.

Conclusion

In this simplified session, you learned how to run YOLOv5 object detection on a custom video and visualize the results. You can further explore the various configurations and options YOLOv5 offers to fine-tune the detection to your specific needs.