People Counting Yolo 3

This is a simple yolo application that is intended to count people on images and videos. The code with minimal modifications originates from:

https://blog.paperspace.com/tag/series-yolo/

https://github.com/ayooshkathuria/YOLO_v3_tutorial_from_scratch?ref=blog.paperspace.com

Some example outputs:

Using the original multi-class object detection:

• processing images: python detect.py --images dog-cycle-car.png --det det

• processing videos: python video.py --video metro.gif

Changes relative to original project:

• migrating to up-to-date packages
• adding people counting mode
• saving sample gif

Counting people mode:

python video.py --count_people --video metro.gif

Saving sample:

python video.py --count_people --save_sample --video metro.gif

Setup

Current system -known to be working on :

• MacOS 12.5.1

• pytorch 2.3.1

• python 3.12.4

• numpy 2.0.0

• opencv 4.10.0

( for saving video:)

• imageio 2.34.2

Setup

• git clone https://github.com/kppeterkiss/YOLO_v3_tutorial_from_scratch.git
• cd YOLO_v3_tutorial_from_scratch
• conda install python conda-forge::opencv pytorch::pytorch numpy pandas conda-forge::imageio
• wget https://pjreddie.com/media/files/yolov3.weights

YOLO_v3_tutorial_from_scratch

Accompanying code for Paperspace tutorial series "How to Implement YOLO v3 Object Detector from Scratch"

Here's what a typical output of the detector will look like ;)

About the training Code

This code is only mean't as a companion to the tutorial series and won't be updated. If you want to have a look at the ever updating YOLO v3 code, go to my other repo at https://github.com/ayooshkathuria/pytorch-yolo-v3

Also, the other repo offers a lot of customisation options, which are not present in this repo for making tutorial easier to follow. (Don't wanna confuse the shit out readers, do we?)

About when is the training code coming? I have my undergraduate thesis this May, and will be busy. So, you might have to wait for a till the second part of May.

Cheers