# PFGF

**Repository Path**: 910024445/PFGF

## Basic Information

- **Project Name**: PFGF
- **Description**: No description available
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: main
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2025-06-26
- **Last Updated**: 2025-06-26

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# Pseudo Visible Feature Fine-Grained Fusion for Thermal Object Detection (CVPR-25)
**Note:** 🔥 This repository currently provides **inference code**. Full training code is almost there! 😎

## Overview
![](README.assets/overview.png)

## Environment Configuration
1. Create Virtual Environment
```
conda create -n PFGF_Inference python=3.9
conda activate PFGF_Inference
```
2. Install PyTorch
```
pip install torch==1.12.1+cu116 torchvision==0.13.1+cu116 torchaudio==0.12.1 --extra-index-url https://download.pytorch.org/whl/cu116
```
3. Install MMCV using MIM
```
pip install -U openmim
mim install mmcv-full==1.7.2
```
4. Install MMDetection
```
git clone https://github.com/liting1018/PFGF.git
cd PFGF_Inference
pip install -r requirements/build.txt
pip install -v -e .
```
## Quick Inference
For FLIR dataset, please run
```
python tools/analysis_tools/eval_metric.py configs/yolox_l_tirgraphmamba_1x8_200e_FLIR_r_test.py ./pkl/FLIR.pkl --eval mAP
```

For LLVIP dataset, please run
```
python tools/analysis_tools/eval_metric.py configs/yolox_l_tirgraphmamba_1x8_200e_LLVIP_r_test.py ./pkl/LLVIP.pkl --eval mAP
```

For Autonomous Vehicles dataset, please run
```
python tools/analysis_tools/eval_metric.py configs/yolox_l_tirgraphmamba_1x8_200e_AV_r_test.py ./pkl/AV.pkl --eval mAP
```