ForAfric-Agricultural-Fields-Delineation

KO | EN

gitlite — search

#python #ai #typescript #golang #go #android #machine-learning #nodejs #java #awesome #kubernetes #mcp #ios #open-source

ForAfric-Agricultural-Fields-Delineation

(★ 17)

Instance Aware segmentation of Agricultural Fields Using Mask R-CNN (Computer vision project)

ForAfric-Agricultural-Fields-Delineation 최신버젼 다운로드

최종 버전 다운로드 (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# ForAfric-Deep-Learning-For-Agricultural-Fields-Delineation
Instance Aware segmentation  of Agricultural Fields Using Mask R-CNN (Computer vision project)

## Abstract
This article aims to delimit agricultural field parcels from satellite images through deep learning for instance segmentation.
Manual delineation is precise but time-consuming, and several automated approaches using traditional image segmentation techniques are struggling to capture the variety of possible appearances in the field.
Deep learning has proven to be effective in the various computer vision tasks, and could be a good candidate for accurate, efficient and generalizable vision of agricultural fields. 
Here, we use the best neural network model for instance segmentation to date: the Mask R-CNN that is formed on Sentinel-2 images corresponding to agricultural field polygons in Denmark. 
Unlike many other approaches, the model works on raw RGB images without pre- and post-processing. 
The results are generalized across different field sizes, shapes and other properties, but show characteristic problems in some cases, especially when we apply it to portions of images from another geographical area (the Zaer region in our case). 
Overall, the results appear promising and would therefore prove the validity of the deep learning approach. 
In addition, the methodology offers many opportunities for improvement.

## Results

<img width="350" src="images/bigresult.PNG">
 
 
## Instructions
### 1. File and Folder
- The Construct_The_Dataset_ForafriPro folder, contains the notebook code needed to build 
our Dataset as indicated in the preprint article. You can apply it to any RGB image with a shapefile containing the field geometries.
For more details about the preprocessing, follow this link (Chrieke Master thesis - it is the main documentation for this project) : https://github.com/chrieke/InstanceSegmentation_Sentinel2

- The "dataset.py" file contains the python code allowing the model to read our training data.

- The program_Project_Forafric file contains all the documented code lines to train the model, evaluate it and make agricultural fields predictions

### 2. How to execute !
Open the Program_Project_Forafric.ipynb folder (it is better to  use Google Colab) and follow the instructions.
 Also remember to replace the dataset file in the Agri/mrcnn folder with the one in the repo ("dataset.py").

ForAfric-Agricultural-Fields-Delineation

ForAfric-Agricultural-Fields-Delineation 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

ForAfric-Agricultural-Fields-Delineation

ForAfric-Agricultural-Fields-Delineation 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

Download README