Fulin-Gao / OpenGCD

OpenGCD: Assisting Open World Recognition with Generalized Category Discovery

A desirable open world recognition (OWR) system requires performing three tasks: (1) Open set recognition (OSR), i.e., classifying the known (classes seen during training) and rejecting the unknown (unseen/novel classes) online; (2) Grouping and labeling these unknown as novel known classes; (3) Incremental learning (IL), i.e., incrementally learning these novel classes and retaining the memory of old classes.

🔥 Preparation

Dependencies

All dependencies are included in environment.yml. To install, run

conda env create -f environment.yml

Data

You can find the required data from the link below and place them according to the path shown in OpenGCD/config.py.

• CIFAR-10, CIFAR-100
• CUB
• ImangeNet

Pretrained weights

You can download the ViT weights (dino_vitbase16_pretrain.pth) trained on ImageNet with DINO self-supervision at ViT-B/16.

Features

You can run extract_features.py to get the feature embedding for each dataset.

🚀 Code

We provide code and models for our experiments on CIFAR10, CIFAR100, and CUB in OpenGCD:

• Code for exemplar selection in OpenGCD/methods/exemplars_selection
• Code for closed set recognition in OpenGCD/methods/closed_set_recognition
• Code for open set recognition in OpenGCD/methods/open_set_recognition
• Code for generalized category discovery in OpenGCD/methods/novel_category_discover
• Code for evaluation metrics (Harmonic normalized accuracy and Harmonic clustering accuracy) in OpenGCD/project_utils/metrics.py
• Code for our experiments in OpenGCD/exp

🔖 Procedure

The schematic and pseudocode of the formulated OpenGCD are provided in Appendix.pdf.

💡 Acknowledgement

Our code is partially based on the following repositories:

• A Tool for Software Defect Prediction
• Generalized Category Discovery
• AutoNovel: Automatically Discovering and Learning Novel Visual Categories