AlibekovMurad5202 / DLOps-practice

Репозиторий для сдачи практических заданий по курсу Разработка систем глубокого обучения

DAMO-YOLO Demo

Source of original project: DAMO-YOLO

PapersWithCode: DAMO-YOLO: A Report on Real-Time Object Detection Design

NOTE: If you don't have Docker, install it using documentation!

NOTE: Available only for Windows and Linux!

NOTE: If the inference of the model causes an error for you, it may be worth updating the paths in download_model.py!

1. Clone repo

git clone https://github.com/AlibekovMurad5202/DLOps-practice.git && cd DLOps-practice

2. Build docker image

docker build -t damoyolo .

3. Run docker image

Ubuntu:

docker run -it -v "$(pwd)":/DLOps damoyolo

Windows:

docker run -it -v <path_to_DLOps-practice>:/DLOps damoyolo

Example (for Windows): docker run -it -v "C:\Users\murad\Desktop\ITMM\tmp\DLOps-practice":/DLOps damoyolo

4. Activate environment

conda activate yolo_env && source DAMO-YOLO-env/bin/activate

5. Download pretrained model and reference data from Google Drive

cd DLOps && python3 download_model.py -m "damoyolo_tinynasL25_S.pth"

6. Prepare and run inference and test

sed -i -e 's/\r$//' run.sh
chmod +x run.sh
./run.sh

7. Close docker

exit
docker ps -a
docker stop CONTAINER_ID
docker rm CONTAINER_ID