FGDCC: Fine-Grained Deep Cluster Categorization - An Architecture for Intra-Class Variability in Problems in FGVC tasks
FGDCC is an architecture developed to tackle intra-class variability problems in FGVC tasks. It operates by performing hierarchical classification of class-wise cluster-assignments conditioned on parent labels (original dataset targets).
- Python 3.8 (or newer)
- PyTorch 2.0
- torchvision
- Faiss
- Other dependencies: pyyaml, numpy, opencv, submitit, timm
This repository is built upon the I-JEPA repository.
See the LICENSE file for details about the license under which this code is made available.
If you find this repository useful in your research, please consider giving a star ⭐ and a citation