├── README.md
├── requirements.txt
├── datasets
│ └── kaist-rgbt/ (see below explanation)
├── data
│ ├── ...
│ └── kaist-rgbt.yaml
├── models
│ ├── ...
│ ├── common.py
│ ├── yolo.py
│ ├── yolo5n_kaist-rgbt.yaml
│ └── yolo5s_kaist-rgbt.yaml
├── utils
│ ├── ...
│ ├── eval
│ │ ├── coco.py
│ │ ├── cocoeval.py
│ │ └── kaisteval.py
│ ├── dataloaders.py
│ └── loss.py
├── detect.py
├── debug_kaist.ipynb
├── val.py
└── train_simple.py-
(Recommended) Set W&B environment variables Plase find API Keys from https://wandb.ai/settings
$ echo "export WANDB_API_KEY={YOUR_WANDB_API_KEY}" >> ~/.bashrc $ echo "export WANDB_ENTITY={USER_NAME}" >> ~/.bashrc $ echo "export WANDB_DIR={YOUR_WANDB_LOG_DIR}" >> ~/.bashrc $ source ~/.bashrc
Please refer this webpage for more details: https://docs.wandb.ai/ko/guides/track/environment-variables/
-
Prepare dataset (5.8GB, multispectral(visible + lwir) images with bbox labels)
$ wget http://ircv-nas.hanyang.ac.kr:9088/downloads/kaist-rgbt-aue8088.tar $ tar xvf kaist-rgbt-aue8088.tar
-
Clone base code repository (replace
ircvlabtoYOUR_GITHUB_ACCOUNTif you forked the repository)$ git clone https://github.com/ircvlab/aue8088 $ cd aue8088 -
Create python virtual environment
$ python3 -m venv venv/aue8088 $ source venv/aue8088/bin/activate -
Check whether the virtual environment set properly : The result should end with
venv/aue8088/bin/python.$ which python
-
[!] Create a symbolic link for kaist-rgbt dataset
-
Assume the below folder structure
├── kaist-rgbt ├── aue8088 │ ├── data/ │ ├── models/ │ ├── train_simple.py │ ├── ... │ └── README.md (this file)
-
Use below commands
$ mkdir datasets $ ln -s $(realpath ../kaist-rgbt) datasets/kaist-rgbt $
-
-
Install required packages
$ pip install -r requirements.txt
- Command
$ python train_simple.py \ --img 640 \ --batch-size 16 \ --epochs 20 \ --data data/kaist-rgbt.yaml \ --cfg models/yolov5n_kaist-rgbt.yaml \ --weights yolov5n.pt \ --workers 16 \ --name yolov5n-rgbt \ --entity $WANDB_ENTITY \ --rgbt \ --single-cls
- On your labtop, go to the website:
http://166.104.168.170:8888/(Open: 5/23)- Only available in Hanyang internal network
- If you're not in campus, please use VPN (https://vpn.hanyang.ac.kr)
- It takes a day (or two) to get the permission from IT department.
- Sign up
- Send a message to me via LMS (then, I'll manually verify your account.)
- Go to
All Challenges-Multispectral Pedestrian Detection Challenge-Submit - Upload your predictions on
test-all-20.txt- If you run
train_simple.pywith the default setting, predictions ontest-all-20.txtwill be generated:runs/train/*/epoch*_predictions.json - You can download this file onto your computer.
- Note: if size of the prediction file is too large (about > 30MB), evaluation on the server could be failed.
- If you run