PyTorch: How to Detect and Track Objects in a Video using YOLOv5

Object detection and tracking in videos are essential tasks in computer vision for various applications such as surveillance, self-driving cars, and sports analytics. YOLOv5 is a popular object detection algorithm built using PyTorch which provides fast and accurate results.

Step 1: Install PyTorch and YOLOv5

To get started, make sure you have PyTorch and YOLOv5 installed:

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

Step 2: Process the Video

Before detecting and tracking objects in the video, you need to preprocess the video:

import cv2

video_path = "path/to/video.mp4"
cap = cv2.VideoCapture(video_path)

while cap.isOpened():
    ret, frame = cap.read()
    if not ret:
        break

    # Process the frame here
    
cap.release()

Step 3: Detect Objects using YOLOv5

Now, you can use YOLOv5 to detect objects in each frame of the video:

import torch
from models.experimental import attempt_load
from utils.general import non_max_suppression

model = attempt_load("yolov5m.pt", map_location="cpu")

while cap.isOpened():
    ret, frame = cap.read()
    if not ret:
        break

    # Preprocess the frame
    img = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)

    # Perform object detection
    results = model(torch.Tensor(img).unsqueeze(0))

    # Apply non-maximum suppression
    detections = non_max_suppression(results, 0.4, 0.5)

Step 4: Track Objects

Finally, you can track the detected objects across frames to get their motion paths:

from tracker import Tracker

tracker = Tracker()

for detection in detections:
    tracker.update(detection)

motion_paths = tracker.get_motion_paths()

Conclusion

By following these steps, you can easily detect and track objects in a video using YOLOv5 in PyTorch. This opens up a wide range of possibilities for applications in computer vision and beyond.