Vehicle Accident Detection Using YOLOV3
It detects an occurrence of vehicle accidents using three classes namely- collision, flipping, and fire in images and videos using the YOLOV3 algorithm which is running on Darknet-53.
Created on 16th October 2020
•
Vehicle Accident Detection Using YOLOV3
It detects an occurrence of vehicle accidents using three classes namely- collision, flipping, and fire in images and videos using the YOLOV3 algorithm which is running on Darknet-53.
The problem Vehicle Accident Detection Using YOLOV3 solves
The goal is to provide instant help to the crash site on highways by raising alarm to the concerned authority. It is a completely software-based solution that automates the detection of an accident on highways using an Image Processing algorithm which is YOLOV3. Its competitor was RCNN. We used YOLOV3 so as to get results in realtime without any delay. The whole setup was done on Google Colab. The accuracy of the model was 85%. We have three classes for detection namely: Collision, fire, and flipping of the vehicle.
Challenges I ran into
The project was done during the lockdown period in the month of April-May' 20. We knew that the accuracy of YOLOV3 is little less than RCNN, so we wanted to model this project on HPC (High-Performance Computing) by applying Big Data Analytics. However, we were obligated to work with Google Colab which gave us pretty decent results.
Technologies used