Separates files that have annotations and those which don't - files with annotations are loaded as validation files - all others

2D Convolutional LSTM cell with (optional) normalization and recurrent dropout. The implementation is based on: tf.contrib.rnn.LayerNormBasicLSTM

2D Convolutional GRU cell with (optional) normalization. Modified from these: https://github.com/carlthome/tensorflow-convlstm-cell/blob/mast

Use files from one held-out robot for testing and files from all other robots for training

Evaluate a fixed TensorFlow graph repeatedly, safely, efficiently. Extend this class to create a particular kind of op evaluator, like an imag

Separates files that have annotations and those which don't - files with annotations are loaded as validation files - all others