This module contains various methods that are generally used for organizing and training and inference.
- Python 2.7
- Ubuntu 16.04
- Pre-installed caffe2.
This module has to be downloaded before using the utility functions. Note: Still working on developing this package, and most of the code is taken from tutorials of caffe2.
Below are the utility functions supported currently:
write_db(db_type, db_name, features, labels)
convertRawDataToNCHWFormat(raw, num_channels, width, height)
AddInputLayer(model, batch_size, db, db_type)
AddAccuracy(model, softmax, label)
AddTrainingParameters(model, softmax, label)
AddBookKeepingOperators(model)
SaveNet(test_model, prefix, tensor_shape_chw)
import caffe2_utils as u
# Dataset conversions.
u.convertRawDataToNCHWFormat(<numpy_raw_features>, num_channels, width, height)
u.write_db(<db_format>, <db_name>, <numpy_features_nchw>, <numpy_labels>)
...
TODO: Need to add various features into this utility package.