TFNet

Official Pytorch Code base for "Integrating Segment Anything Model derived boundary prior and high-level semantics for cropland extraction from high-resolution remote sensing images"

Project

Introduction

We propose a two-flow network based on multitask VFM, named TFNet, to extract croplands with well-delineated boundaries from high-resolution remote sensing images. TFNet consists of a mask flow and a boundary flow. It first uses a VFM as visual encoder to obtain universal semantic features regarding croplands, and then aggregates them into the two flows.

Using the code:

The code is stable while using Python 3.9.0, CUDA >=11.0

• Clone this repository:

git clone https://github.com/long123524/TFNet
cd TFNet

Installation

Install FastSAM following the instructions. Modify the Ultralytics source files following the instructions at: 'TFNet/models/FastSAM/README.md'.

Preprocessing

Using the code preprocess.py to obtain boundary maps.

Data Format

Make sure to put the files as the following structure:

inputs
└── <train>
    ├── image
    |   ├── 001.tif
    │   ├── 002.tif
    │   ├── 003.tif
    │   ├── ...
    |
    └── mask
    |   ├── 001.tif
    |   ├── 002.tif
    |   ├── 003.tif
    |   ├── ...
    └── contour
    |   ├── 001.tif
    |   ├── 002.tif
    |   ├── 003.tif
    |   ├── ...
    └── ...

For testing and validation datasets, the same structure as the above.

Training and testing

python train.py

testing

python pred.py

Evaluation

python eval.py

A pretrained weight

A pretrained weight of FastSAM is provided: https://drive.google.com/file/d/1uzeVfA4gEQ772vzLntnkqvWePSw84F6y/view?usp=sharing.

A GF-2 dataset

Shandong GF-2 image:https://drive.google.com/file/d/1JZtRSxX5PaT3JCzvCLq2Jrt0CBXqZj7c/view?usp=drive_link A corresponding partial cropland label can be accessible at https://drive.google.com/file/d/19OrVPkb0MkoaUvaax_9uvnJgSr_dcSSW/view?usp=sharing.

Citation:

If you find this work useful or interesting, please consider citing the following references.

[1] Long J, Zhao H, Li M, et.al. Integrating Segment Anything Model derived boundary prior and high-level semantics for cropland extraction from high-resolution remote sensing images. IEEE GEOSCIENCE AND REMOTE SENSING LETTERS.
[2] Long J, Li M, Wang X, et.al. Delineation of agricultural fields using multi-task BsiNet from high-resolution satellite images. International Journal of Applied Earth Observation and Geoinformation, 2022, 112:102871.
[3] Li M, Long J, Stein A, et.al. Using a semantic edge-aware multi-task neural network to delineate agricultural parcels from remote sensing images. ISPRS Journal of Photogrammetry and Remote Sensing, 2023, 200:24-40.

A large cropland dataset collected from VHR images:

Will be accessible at https://github.com/NanNanmei/HBGNet, more details can be found at a recent collaborative paper "A large-scale VHR parcel dataset and a novel hierarchical semantic boundary-guided network for agricultural parcel delineation (https://www.sciencedirect.com/science/article/pii/S0924271625000395)"

A parcel vectorization model:

More details can be found at a recent collaborative paper "Extracting vectorized agricultural parcels from high-resolution satellite images using a Point-Line-Region interactive multitask model" published in the journal of Computers and Electronics in Agriculture. Code is available at https://github.com/mengmengli01/PLR-Net-demo/tree/main.