Mastering Computer Vision with TensorFlow 2.x

Recognizing actions based on an accelerometer

This method involves the following steps:

1. Processing the input accelerometer data: The accelerometer data is sensitive to where it is located. For example, if mounted in the waist region, the hand movements will not see a big change in the accelerometer compared to when mounted in the arm. Also, for a different position, different data needs to be collected and then combined.
2. Preparing data so that it can be input into TensorFlow: Load the data using tf.data.Dataset to develop a simple, highly efficient data pipeline. The tensor_slices command extracts a slice of data from the input.
3. Developing the CNN model and training it: One or two dense layers with the flatten and softmax functions at the ...