I didn’t understand mobilenet, could someone explain the detection process and the architecture please
Sorry, but it doesn’t seem fair to ask a question this general. Prof Ng gives you a total of 24 minutes worth of lectures specifically about MobilNet v1 and v2. He explains how 1 x 1 convolutions and depthwise separable convolutions work and how they are used in MobilNet to make it more efficient and more likely to fit on a mobile device with limited computation resources. He also explains how the “bottleneck” layers are used in the architecture and how that relates to the ideas that we learned about in the earlier section of Week 2 covering Residual Networks.
He also gives you the name of the paper that describes MobilNet and based on that, I was able to find the paper with one google search.
The high level point is that it’s a deep convnet that can be trained as an object classifier. In the assignment, which covers Transfer Learning with MobilNet, you get to see some of the data on which the model was trained and then get to “repurpose” it to a more specific classification task by modifying the final few layers of the network.
My suggestion would be to watch the lectures again and perhaps take a quick look at the paper just to get a sense for what is covered there and then ask some more specific questions. It’s not reasonable to expect one of the mentors to repeat for you all the material about MobilNet that is covered in the lectures.