torch--SP-RAN-Self-paced-Residual-Aggregated-Network

SP-RAN

Source code for "**SP-RAN: Self-Paced Residual Aggregated Network for Solar Panel Mapping in Weakly Labeled Aerial Images **", accepted in TGRS. The paper's PDF can be found in Here.

Jue Zhang; Xiuping Jia; Jiankun Hu

School of Engineering and Information Technology, University of New South Wales, Canberra

Prerequisites

environment

Windows 10
Torch 1.12.0
CUDA 11.6.0
Python 3.7.13
Opencv 3.4.2

data sets

GoogleEarth Static Map API

Parameters

Please refer to act_config.json

Training

1st training stage

GradCAM

2st training stage

Setting the training data to the proper root as follows:

-- data -- train -- fore

                 -- back
                 
                 -- Pseudo labels
                
         -- test -- cls -- fore
         
                        -- back
                        
                 -- seg -- img
                 
                        -- gt

set self-pace to False in act_config.json.

Run python train.py --config_path act_config.json.

3nd training stage

Set self-pace to True in act_config.json. Set the label update dir in act_config.json. Run python train.py --config_path act_config.json.

Testing

python predict.py --config_path act_config.json

Noting that the results in our paper do not adopt any post-process including CRF.

The evaluation code can be found in here.

Contact me

If you have any questions, pleas feel free to contact me: jue.zhang@adfa.edu.au.

Citation

We really hope this repo can contribute the conmunity, and if you find this work useful, please use the following citation:

@ARTICLE{9585690,
  author={Zhang, Jue and Jia, Xiuping and Hu, Jiankun},
  journal={IEEE Transactions on Geoscience and Remote Sensing}, 
  title={SP-RAN: Self-Paced Residual Aggregated Network for Solar Panel Mapping in Weakly Labeled Aerial Images}, 
  year={2022},
  volume={60},
  number={},
  pages={1-15},
  doi={10.1109/TGRS.2021.3123268}}

zhangjue1993/torch--SP-RAN-Self-paced-Residual-Aggregated-Network