karimmamer / cropdetectiondl

Crop Detection from Satellite Imagery using Deep Learning

First place solution for Crop Detection from Satellite Imagery competition organized by CV4A workshop at ICLR 2020.

Getting Started

A summarized description of the approach can be found here.

Prerequisites

Firstly, you need to have

• Ubuntu 18.04
• Python3
• 20 GB RAM
• 11 GB GPU RAM

Secondly, you need to install the challenge data and sample submission file by the following the instructions here.

Thirdly, you need to install the dependencies by running:

pip3 install -r requirements.txt

Running

Dataset Preparation

python3 prepare_data.py --data_path ...

This step generates patches around each crop field in the data and saves all of them in a numpy matrix along side their ground truth labels.

Generating a Submission File

python3 main.py --data_path ...

This step trains an ensemble of 10 instances of the same DL model on different train/valid splits then generate a submission file with results on test set.

All augmentations are used except for Mixup augmentation. In order to use it, run

python3 main.py --data_path ... --mixup_augmentation True

However it uses a lot of RAM (~50 GB) so I wouldn't recommend using it.

cropdetectiondl's Issues

Error while running requirements.txt on Python 3.9.12

When running requirements.txt after cloning the repo, I am getting the following error: ERROR: Could not find a version that satisfies the requirement torch==1.4.0 (from versions: 1.7.1, 1.8.0, 1.8.1, 1.9.0, 1.9.1, 1.10.0, 1.10.1, 1.10.2, 1.11.0, 1.12.0, 1.12.1, 1.13.0, 1.13.1, 2.0.0, 2.0.1)
ERROR: No matching distribution found for torch==1.4.0