Skip to content

An object detection model to identify and count helmets from video.

Notifications You must be signed in to change notification settings

haiderzm/Helmet-Detection

Repository files navigation

Helmet-Detection and Tracking

Task :- Train object detection and tracking model to identify and count helmets from video.

Steps Involved :-

1. Data gathering

2. Data Labelling

3. Model/Algorithm Selection

4. Fine-Tunning the model for helmet detection

5. Testing on the given videos

Data gathering

The images for training the deeplearning model are gathered from github repository (0816-Radhu/helmet_detection_using_machine_learning) and some random images of traffic containing bikers from websites like wired.com, bbc.com and youtube.com

Data Labelling

To label the images an open-source tool labelImg is used (https://github.com/tzutalin/labelImg) in the format of yolo.

Model/Algorithm Selection(Object-Detection)

Model used for the task is ultralytics implementation of yolo-v5 model in PyTorch (https://github.com/ultralytics/yolov5)

YOLOv5 🚀 is a family of object detection architectures and models pretrained on the COCO dataset, and represents Ultralytics open-source research into future vision AI methods, incorporating lessons learned and best practices evolved over thousands of hours of research and development.

Model/Algorithm Selection(Object-Tracking)

Model used for tracking is DeepSort on yolo v5 base detection, Repository used (https://github.com/mikel-brostrom/Yolov5_DeepSort_Pytorch)

Fine-Tunning the model for helmet detection

Model is trained/fine-tuned using the free-gpu provided by google colaboratory on the custom helmet detection dataset for 350 epochs. The library used is PyTorch.

Testing on videos

The model was tested on 2 youtube videos link to the original videos are :-

https://www.youtube.com/watch?v=9WNzIDEcNP4 (Bike racing | royalty free video | no copyright | use in your project #racing #bikes - YouTube)

https://www.youtube.com/watch?v=iEIk3RpV6RA (CRAZY INDIAN TRAFFIC CONGESTION - YouTube)

Link to the results of detection :-

About

An object detection model to identify and count helmets from video.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published