Table of Contents

• detect_video.ipynb: detect videos from your own source
• evaluate_result.ipynb: evaluate performance of various detection models
• faster_rcnn.ipynb: use the Faster R-CNN model for detection
• ssd.ipynb: use the SSH model for detection
• training.ipynb: train the YOLOv3 model using custom dataset, annotation file and anchor file

Environment Setup

Install Anaconda with Python 3.7

https://www.anaconda.com/distribution/#download-section

Setup Conda Environment

conda env create -f environment.yml

Activate Conda Environment

conda activate tensorflow

Start Jupyter Notebook

jupyter notebook

Detect Pedestrians in Videos from own Source

1. Open detect_video.ipynb in your browser
2. Change the video_path argument in the second cell to the path of your own input video
3. Change the output_path argument in the second cell to the desired output path
4. Run all cells

See report.pdf for result analysis