Abstract
tfaip - a Generic and Powerful Research Framework for Deep Learning based on Tensorflow
Highlights
Summary tfaip is a Python-based research framework for developing, structuring, and deploying Deep Learning projects powered by Tensorflow (Abadi et al, 2015) and is intended for scientists of universities or organizations who research, develop, and optionally deploy Deep Learning models. tfaip enables both simple and complex implementation scenarios, such as image classification, object detection, text recognition, natural language processing, or speech recognition
Each scenario is highly configurable by parameters that can directly be modified by the command line or the API
The implementation of a scenario during research and development typically comprises several tasks, for example setting up the graph, the training, and the data pipeline
Summary
The implementation of a scenario during research and development typically comprises several tasks, for example setting up the graph (e.g., the network architecture), the training (e.g., the optimizer or learning rate schedule), and the data pipeline (e.g., the data sources). Tfaip tackles this issue by providing a sophisticated pipeline setup based on “data processors” which apply simple transformation operations in pure Python code and are automatically executed in parallel. Another important step which is simplified by tfaip is the deployment of a scenario. Tfaip will automatically log the training process using the Tensorboard and provides utility scripts to resume a crashed or stopped training, or to set up an array of training configurations via an Excel sheet
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
