uxhao/deepcv repository overview

⁠deepcv

Author: uxhao ([email protected]⁠)

⁠Description

The deepcv is the docker image which is a deep learning environment for computer vision. It builds on ubuntu18.04 or 22.04, cuda10.2 or 11.8, and cudnn7 or 8. There are different tags in the deepcv. The variety of the tag shows the corresponding environment. The deepcv uses miniconda3 as the default environment manager. A default environment called cv which pre-installs parts of necessary packages used for computer vision includes in the deepcv.

⁠Usage

You can use the ssh tool like Xshell connected it via username(root) and password(root). If you customized the username and password when you created the instance, log in according to your settings. Perform the command conda activate cv to activate the default environment.

Tensorboard starts automatically after root login the instance for the first time, and the default working directory is under /gemini/logs. You can see the following three directories, please try to follow the directory instructions as much as possible to ensure that you can use this docker image correctly.

/gemini/code   # code workplace
/gemini/logs  # tensorboard logdir
/gemini/output  # checkpoint save path

Note: After creating or restarting the instance, please login to the terminal with the root user once, the ssh service will start automatically, you can check ~/.bashrc for details.

⁠Packages

• tag: uxhao/deepcv:pytorch2.0-cuda11.8-ubuntu22.04⁠
• tag: uxhao/deepcv:pytorch1.12.1-cuda10.2-ubuntu18.04⁠

These tags includes the following packages.

pytorch 2.0.0
pydantic
pandas
pillow
matplotlib
comet_ml
opencv-python==4.5.3.56
python-dotenv
torchmetrics
scikit-learn
lightning
albumentations
segmentation-models-pytorch
pytorch-lightning 2.0
dgl
dglgo
tensorboard