deepest3d

3D coordinate estimation of surgical instruments in laparoscopic surgery using deep neural network

Prerequisites

1. To build a Docker image from Dockerfile, run this command on the main directory.

docker build -t <image name> .

2. Create and start a Docker container

docker run -it -p <host port>:<container port> --name <container name> -v <host directory>:<container directory> etarho/dl-gpu:<version> /bin/bash

3. Connect to the workstation server

ssh -L <port(client)>:localhost:<port(host)> <user>@<host IP address>

4. Run JupyterLab

jupyter lab --allow-root

5. Access to JupyterLab

Access to localhost:<port(client)>/lab with your browser.

1. Visualize training results

mlflow server -p <port> -h 0.0.0.0

Then, access to localhost:<port> with your browser.

Example

docker run -it -p 8888:8888 --gpus 0 --name dl -v "/home/rock/workspace:/home/workspace" etarho/dl-gpu:1.2 /bin/bash

2. Hyperparameter Optimization

optuna dashboard --storage 'sqlite:///<database>.db' --study-name '<study name>'