PyTorch webinar

I had the opportunity to present a webinar for the PyTorch Foundation about the research I’ve been doing on transfer learning with self-supervised Vision Transformers (DINOv2) for multi-label plant species classification in the PlantCLEF 2025 challenge.

In the webinar, I covered how to efficiently extract feature embeddings from a dataset of 1.4 million images and utilize PyTorch Lightning for training a model using the embeddings.

Learn about our image processing techniques, including transforming images into grids of tiles and aggregating predictions to overcome computational challenges. Discover the significant performance improvements achieved and get insights into multi-label image classification. Perfect for PyTorch developers, this session included a Q&A and access to our complete codebase at github.com/dsgt-arc/plantclef-2025.