readme.md

Vision transformer - RVSA for object detection

Table of Contents

• Features
• Getting Started
  • Prerequisites
  • Installation

Features

Load object detection model - remote sensing image.

• Visualize

Getting Started

Below is the list of dependencies and steps to use the repository

Prerequisites

List any software or libraries that are required to run your project.

• Python 3
• mmcv
• mmrotate
• torch
• mmdetection

Installation

Step-by-step instructions on how to install your project.

1. Install environment:

• Install torch

• CPU:

pip install -q torch==2.0.1 torchvision==0.15.2 torchaudio==2.0.2 --index-url https://download.pytorch.org/whl/cpu
pip install -q mmcv==2.0.1 -f https://download.openmmlab.com/mmcv/dist/cpu/torch2.0/index.html

• GPU:

pip install -q torch==2.0.1 torchvision==0.15.2 torchaudio==2.0.2
pip install -q mmcv==2.0.1 -f https://download.openmmlab.com/mmcv/dist/cu117/torch2.0/index.html

• Install other dependencies:

pip install -q "mmdet>=3.0.0rc6,<3.1.0"
git clone https://github.com/open-mmlab/mmrotate -b 1.x
cd mmrotate
pip install -q -e .
cd ..
git clone https://github.com/ViTAE-Transformer/MTP
cp -r ./MTP/RS_Tasks_Finetune/Rotated_Detection/mmrotate1.x/mmrotate/* ./mmrotate/mmrotate
git clone https://huggingface.co/datasets/KhangTruong/NWPU-Caption
tar -xf NWPU-Caption/02_NWPU_RESISC45.tar
tar -xf NWPU-Caption/02_NWPU_caption.tar
mv 02_NWPU_RESISC45 data
mv 02_NWPU_caption data
python install.py --root .

2. Change the directory to mmrotate, which is inside the current working directory

cd mmrotate

3. Run the inference

python ../extract_feature.py -n ../data/airplane/airplane_002.jpg

4. It will produce an image in vis_data, which is the labelled image.