arrow_back All Software
model_training
Deep Learning

PyTorch / TensorFlow

Neural network frameworks for deep learning research and deployment

tag2.0+ / 2.13+ gavelBSD / Apache 2.0 codePython devicesCPU / GPU / TPU

info Overview

PyTorch and TensorFlow/Keras are the primary deep learning frameworks used in DASCLAB research. PyTorch is preferred for research-stage model development and custom architectures, while TensorFlow/Keras is used in teaching (ML301 module) and deployment contexts. Both frameworks are applied to medical imaging, NLP, and crop yield prediction.

checklist Key Features

  • Dynamic computation graphs (PyTorch): flexible model debugging
  • Keras high-level API (TensorFlow): rapid prototyping
  • torchvision and torchtext: domain-specific dataset tools
  • TensorBoard: real-time training visualisation and profiling
  • ONNX export: cross-framework model deployment
  • Pre-trained model hub: ResNet, EfficientNet, BERT, XLM-R

photo_library Screenshots

article Publications

Transfer Learning for Malaria Detection in Blood Smear Images
Mwaniki P., Kamanu T.K.
bookPLOS Computational Biology calendar_today2024 linkDOI: #