代码拉取完成，页面将自动刷新
The official implementation of "A Comprehensive Deep-Learning Framework for Fine-Grained Farmland Mapping from High-Resolution Images" Paper Link.
We are delighted to share that our paper has been successfully accepted by the IEEE Transactions on Geoscience and Remote Sensing (TGRS 2024).
This repository contains the full implementation of our model, including training and testing.
We have developed a groundbreaking dataset encompassing diverse types of farmland, taking into account the varying terrain across China.
You can download the whole dataset via Baidu Disk:
Code: FGFD
It is worth noting that when annotating the data, we labeled farmland without crops (in red) and farmland with crops (in green). However, in practical use, we are only concerned with distinguishing between farmland and non-farmland.
We have provided a series of compared methods to estabish the benchmark.
| Method Name | Description |
|---|---|
| UNet | UNet with ResNet-50 encoder for segmentation. |
| DeeplabV3+ | DeepLabV3+ with ResNet-50 encoder for segmentation. |
| PSPNet | Pyramid Scene Parsing Network for semantic segmentation. |
| HRNet | High-Resolution Network for fine-grained segmentation. |
| ABCNet | Attentive Bilateral Contextual Network for Efficient Semantic Segmentation. |
| CMTFNet | CNN and Multiscale Transformer Fusion Network for semantic segmentation. |
| MCCANet | Boundary Supervision-Aided Multi scale Channelwise Cross Attention Network for semantic segmentation. |
| CGNet | Context-Guided Network for efficient segmentation. |
| DenseASPP | Densely connected Atrous Spatial Pyramid Pooling network. |
| ENet | Efficient Neural Network for real-time semantic segmentation. |
| SegNet | Encoder-decoder network for pixel-wise classification. |
| BuildFormer | Specialized model for building segmentation tasks. |
| UANet | Uncertainty-Aware Network with ResNet-50 for segmentation. |
| DSNet | A Local–Global Dual-Stream Network for segmentation. |
| UNetFormer | UNet-based model incorporating transformer layers. |
| DBBANet | Dual-Branch Boundary-Aware Network for segmentation. |
To train the provoided models, follow these steps:
Set the hyperparameters for training.
Run the following command:
python train.py --batchsize 32 --model_name DBBANet --gpu_id 0
To evaluate the trained model, follow these steps:
Ensure the model is properly trained and paths are set.
Run the following command:
python test.py --model_name DBBANet --batchsize 32
If you use our work in your research, please cite:
@ARTICLE{10793088,
author={Li, Jiepan and Wei, Yipan and Wei, Tiangao and He, Wei},
journal={IEEE Transactions on Geoscience and Remote Sensing},
title={A Comprehensive Deep-Learning Framework for Fine-Grained Farmland Mapping From High-Resolution Images},
year={2025},
volume={63},
number={},
pages={1-15},
keywords={Feature extraction;Data mining;Image segmentation;Remote sensing;Benchmark testing;Accuracy;Vectors;Semantics;Annotations;Production;Dual-branch;farmland extraction;remote sensing (RS);semantic segmentation},
doi={10.1109/TGRS.2024.3515157}}此处可能存在不合适展示的内容，页面不予展示。您可通过相关编辑功能自查并修改。
如您确认内容无涉及 不当用语 / 纯广告导流 / 暴力 / 低俗色情 / 侵权 / 盗版 / 虚假 / 无价值内容或违法国家有关法律法规的内容，可点击提交进行申诉，我们将尽快为您处理。
马建仓 AI 助手