Detecting Seatbelt and Phone violations using AI..How it works, let's see>>
The first model's task is to detect where is the windshield in the image, as an example see this image:
The first model was built using object detection API with pre-trained model named my_ssd_resnet50_v1_fpn
.
After detecting the windshield, we take the coordinates of the windshield and cut the image and fed it into the second model.
source code: https://tensorflow-object-detection-api-tutorial.readthedocs.io/en/latest/training.html
The second model was built using image classification pre-trained model named VGG19
.
You will see the training source code and details in the file Classification.ipynp
.