seg_mlflow

Image Segmentation tracked/ deployed with mlflow

In this file the same task from the parent folder is performed using the same model with the addition of mlflow workflow.

mlflow is an open source platform to manage the machine learning lifecycles. Using mlflow one can track the model accuracies/ parameters during training and also save the model. The mlfow projects allows the machine learning code to be reused in any other system just by running the mlflow command. The model saved by the mlflow package can be easily used as an independent entity across different languages, without the need of installing any of the paskages used to train the model.

Packages used

keras, sklearn, tensorlfow, tenosrboard, numpy, pandas, cv2, matplotlib, mlflow

Dataset

For showing comparison study in the task of Image Segmentation, a dataset where we need to segment lesions from the CT scans of lungs.

There are total 267 CT scans of lungs corresponding with the manually labelled segmented masks.

Here is the link to the dataset.

Usage

Make sure you have the above mentioned packages installed in your environment and you have dowloaded and extracted the dataset at Image_Segmentation folder.

Download the reposritory, and in a cmd prompt navigate to the folder Image_Segmentation folder and run:

mlflow run seg_mlflow --no-conda

--no-conda option is given if you want to run the project in your existing environment. If you want to create a new environment omit this option.

For viewing the mlflow ui open another cmd prompt and navigate to the folder semantic-segmentation and run:

mlflow ui

*Note For running Mlflow ui in windows go to the last comment in the link-> mlflow/mlflow#154

For viewing Tensorboard ui open a cmd prompt and run:

tensorboard --logdir="PATH\TO\LOGS"

For passing parameters to the programs run cmd:

mlflow run seg_mlflow --no-conda -P PARAM_NAME_1=PARAM_VALUE_1 -P PARAM_NAME_2=PARAM_VALUE_2

Here are the available parameters:

--image_path TEXT       Path to images folder
--annotation_path TEXT  Path to annotations folder
--weights_path TEXT     Path to base model weights file
--log_dir TEXT          Path to store log files
--initial_lr FLOAT      Initial learning rate
--batch_size INTEGER    Batch size for training
--seed INTEGER          numpy random seed

Folder Structure

Image_Segmentation
                  finding-lungs-in-ct-data 
                                          finding-lungs-in-ct-data
                                                                  2d_images
                                                                  2d_masks
                  seg_mlflow
                  weights
                  logs