Train a custom YOLOv4-tiny object detector

Original Video by cottonbro from Pexels

Mask detection using YOLOv4-tiny

The yolov4-tiny folder in this repository contains the 4 custom files needed. (i.e. yolov4-tiny-custom.cfg, obj.data, obj.names and process.py) except obj.zip(labeled images). I have shared my labeled custom mask dataset obj.zip file on the following link. You can download it from here.

https://www.kaggle.com/techzizou/labeled-mask-dataset-yolo-darknet

The obj.zip file contains 1510 images along with their YOLO labeled text files. I have labeled around 1350 of these and downloaded 149 labeled images from roboflow. I have given the links for my dataset sources at the bottom.

This dataset has mostly close-up images (around 1300) and very few long-shot images (around 200). If you want to download more long-shot images, you can search for datasets online. There are many sites where you can find more datasets. Check out my Medium article below for links to these sites. I have also given a few links for mask datasets. Some of them have more than 10,000 images. You can add your own images and their YOLO labeled text files to the dataset. Try to find good quality images.

NOTE : The yolov4-tiny-custom.cfg, obj.data, and obj.names files are customized for the 2 classes I am working with. (i.e. "with_mask" & "without_mask"). You can edit these files for your custom objects.