Applied Deep Learning for Specialized Domains

2018 – ongoing

Deep Learning
Computer Vision
Applied Machine Learning
Collaborative Research
Collaborations applying deep learning methods to specific, real-world domain problems — from ecological image classification to medical imaging and code generation.
Published

January 1, 2018

This project groups collaborative work that applies deep learning methods to specific domain problems, developed with researchers and students from other fields rather than as part of a single unifying research line. Each collaboration adapts general-purpose deep learning techniques — convolutional ensembles, synthetic data generation, large language model reasoning — to the practical constraints and data of a particular domain.

Key Research Topics:

  • Ecological Image Classification: Ensembles of convolutional neural networks for fine-grained taxonomic identification from images.
  • Medical Image Synthesis: Combining synthetic and real data to improve the training of models on medical MR images, where real annotated data is scarce.
  • LLM Reasoning for Code Generation: Structural alignment and reasoning distillation techniques to improve the quality of automatically generated code.