Many changes

• Removing legacy code
• Port to the new ci
• Estimate the size of the tfrecords file that will be created
• bug-fixes
• Add a new architecture for patch-based CNN
• Refactored the scripts to use the new click tools
• Changed the compute_statistics script completely
• Removed the unused predict_generic script
• Removed the unused train script and renamed train_generic to train

Fixes #50 (closed) #52 (closed) #48 (closed)

bob/learn/tensorflow/script/compute_statistics.py

@@ -73,6 +73,7 @@ def compute_statistics(database, biofiles, load_data, multiple_samples,
     logger.debug('kwargs: %s', kwargs)
 
     assert len(biofiles), "biofiles are empty!"
+    logger.info('Calculating the mean for %d files', len(biofiles))
 
     generator = BioGenerator(
         database,