When i heard projects regarding surveillance camera using machine learning to detect traffic, theft and other computer vision project i felt inspired and need to create such system to detect accidents for better healthcare
What it does
It detects whether the frames captured by video feeds generated by a dashcam/CCTV are predicted to be classified as 'accidents' or 'non-accidents', as to insure better safety measures for drivers and report that to nearby emergency services
How we built it
I built this project using both the Pytorch Deep Learning framework and trained my model on my laptop on different Convolution neural networks architecture
Challenges we ran into
Our main challenge was to gather accident images and videos and manually categuorize images into accient and non-accident frames
To design a deep convolutional neural networks model for this project.
Limited hardware resorces like GPU's.
Accomplishments that we're proud of
As a deep learning fresher, I have build this project with limited hardware and manually categorizing Data to yield better results and accuracy
What we learned
I am pround that i have learnt about computer vision, Pytorch framework and i will continue to learn more advances deep learning concepts
What's next for Dash cam accident detection
To improve model further more to increase its accuracy and to create a responsive web app with google map api to share location during accident to nearby emergency services