The official pytorch implementation of IGFNet: Illumination-Guided Fusion Network for Semantic Scene Understanding using RGB-Thermal Images.

We test our code in Python 3.6.9, CUDA 11.1, cuDNN 8, and PyTorch 1.10.1.

IGFNet introduces an interpretable RGB-Thermal fusion network and utilize illumination to guide the fusion of multi-modal features, which includes an Illumination Estimation Module (IEM) and an Illumination-Guided-Cross-Modal Rectification Module (IGCM-RM).

The dataset can be downloaded from the MFNet project page.

The pretrained weight of IGFNet can be downloaded from here

• Clone this repo

$ git clone https://github.com/lab-sun/IGFNet.git

• Build docker image

$ cd ~/IGFNet
$ docker build -t docker_image_igfnet .

• Download the dataset

$ (You should be in the IGFNet folder)
$ mkdir ./datasets
$ cd ./datasets
$ (download our preprocessed dataset.zip in this folder)
$ unzip -d .. dataset.zip

• To reproduce our results, you need to download our pretrained weights.

$ (You should be in the IGFNet folder)
$ mkdir ./pretrained
$ cd ./pretrained
$ docker run -it --shm-size 8G -p 1234:6006 --name docker_container_igfnet --gpus all -v ~/IGFNet:/workspace docker_image_igfnet
$ cd /workspace
$ python3 train.py

• To see the training process

$ (fire up another terminal)
$ docker exec -it docker_container_igfnet /bin/bash
$ cd /workspace
$ tensorboard --bind_all --logdir=./runs/tensorboard_log/
$ (fire up your favorite browser with http://localhost:1234, you will see the tensorboard)

The results will be saved in the ./runs folder. Note: Please change the smoothing factor in the Tensorboard webpage to 0.999, otherwise, you may not find the patterns from the noisy plots. If you have the error docker: Error response from daemon: could not select device driver, please first install NVIDIA Container Toolkit on your computer!

Some of the codes are borrowed from CMX