/finger_digit_identification

Design of a finger digit pain point images using deep learning and classical segmentation techniques.

Primary LanguageJupyter Notebook

Identifying the landing position of the image : Rheumatoid Arthritis application

Steps to train U-net

  1. save all images and labels in two separate folders (image and corresponding label should have the same name, i.e.: 'img00001.png')
  2. define in src/data/segmentation_dataset.py the path where the images and labels are stored
  3. create the .txt files in get_img_names.ipynb. For this, define the path of the images in this file
  4. open the autoencoder.ipynb folder and define all necesary paths

Dataset

Characteristics and definition - what the dataset has and what it does not have.

  1. Preprocessing dataset: Data augmentation, also understanding more on hand to understand results, edge cases
    1. Expanding dataset (via data augmentation techniques with skin color)
  2. Discuss processing architecture :
  3. Train the model - divide training, validation etc.
  4. Fine tune the parameters and iterate

Define knowledge into the system. Also we need to come up with Get an image -> Is the image good -> Preprocess the image for inputs to the algorithm -> Process with learned model -> Compute landing position -> Check if the answer is correct