The purpose of this project is to build a converter that can translate models built in Tensorflow, Keras, Pytorch, Mxnet, ONNX to mlpack's model format and vice-versa.
With the advent of extremely deep neural architectures and their high training cost, transfer learning is the only way out. That said, the number of pre-trained models for mlpack is practically zero till now while more popular frameworks like Tensorflow has dozens of them.
Models trained in mlpack can be converted and used in Tensorflow for better benchmarking and feature-testing. Moreover, it can open up the field for mlpack and make it as popular as the other frameworks mentioned.