Shijunqiang/weather-classification-swin-coatnet

Weather Classification Models

简体中文

This repository contains only three trained PyTorch model checkpoints.
The complete source code, evaluation scripts, and experiment results are available on GitHub:

GitHub Project:
https://github.com/shijunqiang27/swin-coatnet-weather-classification

Model Files

Model File	Architecture	Dataset	Classes
model_CloudRainSnowSun_swin_v2_s.pth	Swin Transformer V2 Small	Dataset v4.1	Cloud, Rain, Snow, Sun
model_HazeRainSnowSun_swin_v2_s.pth	Swin Transformer V2 Small	MWI Extend DID v2	Haze, Rain, Snow, Sun
model_HazeRainSnowSun_CoAtNet.pth	CoAtNet	MWI Extend DID v2	Haze, Rain, Snow, Sun

Datasets

• Dataset v4.1
  https://www.kaggle.com/datasets/sjq247/dataset-v4-1

• MWI Extend DID v2
  https://www.kaggle.com/datasets/sjq247/dataset-mwi-extend-did-v2

Usage Notes

These .pth files are PyTorch model weights. When using them, make sure that:

• the model architecture and number of classes match the training configuration;
• the class order matches the corresponding dataset;
• image preprocessing is consistent with the training and evaluation code;
• dataset and model paths are updated to match the actual local environment.

For loading and evaluation examples, see the src/evaluate/ directory in the GitHub project.

These models are intended for research and educational use only. They should not be used as professional meteorological instruments or as the sole basis for safety-critical decisions.