CV-Model-Comparison

A simple Python project to compare two popular computer vision models for pedestrian detection: Histogram of Oriented Gradients (HoG) and Faster R-CNN.

Features

• Loads all images from the Penn Fudan Pedestrian dataset
• Detects pedestrians using:
  • HoG + SVM (OpenCV)
  • Faster R-CNN (PyTorch/Torchvision)
• Visualises detection results with bounding boxes

Requirements

• Python 3.x
• OpenCV
• scikit-image
• PyTorch
• Torchvision

Usage

1. Install dependencies:

   pip install opencv-python scikit-image torch torchvision

2. Download the Penn Fudan Pedestrian dataset and place it in a folder, e.g. c:/path/to/PennFudanPed/PNGImages.

3. Update the images_folder variable at the top of both HoG Model.py and Faster R-CNN Model.py to the path where your dataset images are stored:

   images_folder = 'c:/path/to/PennFudanPed/PNGImages'  # Update this path as needed

4. Run the HoG model:

   python "HoG Model.py"

5. Run the Faster R-CNN model:

   python "Faster R-CNN Model.py"

Each script will process and display the detection results for every image in the dataset folder.

Notes

• Ensure the dataset path is correct in both scripts.
• The HoG model uses greyscale images and classical feature-based detection.
• The Faster R-CNN model uses a deep learning approach and may require a GPU for faster inference.
• Results may vary depending on the image and environment.