We design a solution for camera-based blind spot detection.
In this project, I implemented the combination of separable depthwise convolution, residual learning and squeeze-and-excitation module.
Here is a demo gif to show the result:
All the codes are here, we can't upload data since the limitation of the github. If you want, we can email you with data and a longer demo video.