Class Similarity Transition: Decoupling Class Similarities and Imbalance from Generalized Few-shot Segmentation

This project is used to train the base class model for ClassTrans. We suggest to skip this step and directly use this checkpoint to reimplement our results.

🎬 Getting Started

1️⃣ Download data

Pre-processed data from drive

We use a adapted version of OpenEarthMap datasets. You can download the full .zip and directly extract it in the data/ folder.

From scratch

Alternatively, you can prepare the datasets yourself. Here is the structure of the data folder for you to reproduce:

data
├── trainset
│   ├── images
│   └── labels
|
└── train.txt

2️⃣ Download pre-trained models

Pre-trained backbone and models

We use ConvNext_large pre-trained using CLIP as backbone. You can download the weight here and move it to pretrain/.

🗺 Overview of the repo

Data are located in data/ contains the train dataset. All the codes are provided in src/. Testing script is located at the root of the repo.

⚙ Training

python train.py 

📊 Output

The weights of the model are saved in the weight/ and you can use them in ClassTrans.