Exercise Learn By Example: LSTM for Video Activity Recogntion

...with Keras using TesnorBoard and deploying to Android for runtime

 

Follow - Exercise: Learn by Example: LSTM for Video Activity Recognition:--BUT you MUST change
the CNN to a TF's MobileNet (rather than Inception) as it will run MUCH faster on a Mobile device--
So you must change the training code to do this.

 

• Like previous exercise, deploy the model in an Android App (this code will be more involved than previous exercise as you
  must take the input and pass it to a MobileNet model and then package the output for submission to your LSTM based network.

• Like previous exercise, use TensorBoard to display accuracy metrics.

• Make a Youtube video (one per group)showing the Jupyeter notebook running at different stages (cut, and paste the video --I don't want to see it training for too long--but, show me the results after each step of the code to the end)AND add to video running the application. Post to Canvas-Assignments-Ex:Learn LSTM