Human Detector

tech: OpenCV
version: 1.0.1

Overview

This application detects and boxes humans in a video feed using MobileNet-SSD pre-trained model.
The model was obtained from this repo : https://github.com/chuanqi305/MobileNet-SSD

Inference time (how quick the network processes the feed) is displayed at the bottom left corner of the video.

Dependencies

1. Create a virtual environment :

    python -m venv virtual
   

2. Activate the virtual environment :

   on Linux:

    source virtual/Scripts/activate
   

   on Windows:

    .\virtual\Scripts\activate
   

3. Install the necessary librairies :

    pip install opencv-python
   

Improvements

The current version of the application requires one video input and outputs the corresponding human-detected video. Following improvements are planned in the future :

v1.2:

• allowing use of webcam feed

v2:

• deploying web application on Flask
• allowing user to upload video or paste video url

v3:

• allowing triming of uploaded video
• implementing output format options (for e.g. gif)

Additional information

Resources : https://opencv.org/
Original video clip : Pharcyde, Drop (by Spike Jonze)