https://github.com/sbatururimi/convstar
https://github.com/sbatururimi/convstar
Last synced: about 2 months ago
JSON representation
- Host: GitHub
- URL: https://github.com/sbatururimi/convstar
- Owner: sbatururimi
- Created: 2024-07-27T08:41:57.000Z (9 months ago)
- Default Branch: master
- Last Pushed: 2024-09-12T08:29:36.000Z (8 months ago)
- Last Synced: 2024-09-12T18:58:36.894Z (8 months ago)
- Language: Python
- Size: 188 KB
- Stars: 1
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Description
This repository is a PyTorch + PyTorch Lightning implementation of [Crop mapping from image time series: deep learning with multi-scale label hierarchies](https://arxiv.org/pdf/2102.08820). The initial repository unfortunately contains inaccuracies that have been copied many times.
Thank you [Mehmet Ozgur Turkoglu](https://github.com/0zgur0), one of the authors of the paper, for guidance and support.
This current implementation is used in a production solution of an AgriTech company I'm helping with in R&D.
# How to
## Sources:
* https://github.com/0zgur0/multi-stage-convSTAR-network/tree/master?tab=readme-ov-file
* https://github.com/0zgur0/STAckable-Recurrent-network
## Setup
Create a virtual environment from the `requirements.txt` file.
## Starting
A simpler approach with PyTorch Lightning and some StarCell adjustments. We don't split explicitly into folds but this can be done. Check the [original repository](https://github.com/0zgur0/multi-stage-convSTAR-network/tree/master?tab=readme-ov-file) for such info.
NB: The original repository implements a GRU cell instead of a ConvStar cell. Check the [old](https://github.com/0zgur0/STAckable-Recurrent-network) TensorFlow code for comparison.
---
- Download the ZueriCrop Dataset:
1) Download the dataset via https://polybox.ethz.ch/index.php/s/uXfdr2AcXE3QNB6
2) Create a `storage` folder here and rename the downloaded file to `data.h5`
3) Adjust `config.py` per your structure.
- Run the `train.py`