https://github.com/grantgasser/vehicle-detection

Detecting and tracking vehicles with traditional computer vision techniques as well as deep learning
https://github.com/grantgasser/vehicle-detection

computer-vision feature-extraction vehicle-detection yolov5

Last synced: 6 days ago
JSON representation

Detecting and tracking vehicles with traditional computer vision techniques as well as deep learning

• Host: GitHub
• URL: https://github.com/grantgasser/vehicle-detection
• Owner: grantgasser
• License: mit
• Created: 2020-12-03T22:35:53.000Z (about 4 years ago)
• Default Branch: master
• Last Pushed: 2021-02-10T02:58:16.000Z (almost 4 years ago)
• Last Synced: 2024-10-28T16:59:49.009Z (about 2 months ago)
• Topics: computer-vision, feature-extraction, vehicle-detection, yolov5
• Language: Python
• Homepage:
• Size: 5.88 MB
• Stars: 1
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# Vehicle Detection and Tracking

## Final Results via Deep Learning

Click to play video:

[![YOLOv5 Video](docs/0000000005.png)](https://youtu.be/sitoSZPr8HQ)

## 1st Implementation: Manual Feature Extraction + Classifier

1. Color Features

    - Color Histogram

    - Color Spaces (HSV, LUV, etc.)

    - Spatial Binning (`cv2.resize()`)

2. Gradient Features

    - Histogram of Oriented Gradients (HOG)

### Feature Extraction

- Normalize each feature, then combine into one feature vector

- Train-Test split

- Could even use a decision tree for feature selection

- Be careful of time dependencies, even w/ a random train-test split

### Train Vehicle Classifier

Train SVM to classify Car/Not Car.

### Sliding Window

Slide a window (at different scales) over the frames and for each window, classify car/not-car.

### Manual Feature Extraction Results

Click to play video:

[![Manual Feature Extraction](docs/0000000215.png)](https://youtu.be/N94dRl46f8k)

## Deep Learning Implementation (YOLOv5)

Learning the features directly. 

Click to play video:

[![YOLOv5 Video](docs/0000000005.png)](https://youtu.be/sitoSZPr8HQ)