This is a repo for YOLO based models for TrashAI. See ABOUT.md for description of models. For Detectron 2 based models, see https://github.com/running-man-01/trashai_nbs_d2. (Dec 2022 update: Detectron 2 related models deprecated).

STEP 0. Boot up machine

Start a local or remote linux machine with one or more GPU's, preferably with Ubuntu 20.04 LTS.

Go to: https://www.nvidia.com/Download/index.aspx and find the appropriate GPU driver for your machine.

You MUST have Nvidia GPU and Nvidia Driver>=455.23 installed to enable GPU training. Yolo needs GPU and Detectron 2 MUST have GPU.

STEP 1. install necessary dependencies

sudo apt update

sudo apt install git wget curl python3-pip -y

STEP 2. get necessary files

sudo wget https://raw.githubusercontent.com/running-man-01/trashai_nbs/main/starter.sh

STEP 3. deploy docker container

bash starter.sh

sudo docker run --ipc=host -it -v "$(pwd)"/workdir:/usr/src/ -p 8888:8888 ultralytics/yolov5:latest

Note: add --gpus all for multi GPU access.

STEP 4. start a jupyter lab

git clone https://github.com/running-man-01/trashai_nbs && \ cd trashai_nbs && \ jupyter-lab --generate-config && \ echo 'c.NotebookApp.allow_origin = "*"' >> /root/.jupyter/jupyter_notebook_config.py && \ echo 'c.NotebookApp.ip = "0.0.0.0"'>> /root/.jupyter/jupyter_notebook_config.py && \ jupyter-lab --ip=0.0.0.0 --port=8888 --no-browser --allow-root

So far, the environment has been set up. You can go to the Jupyter Lab link pops up in the terminal.

Open Jupyter Notebooks you want to review in Jupyter Lab.