Hacking Dermatology 2018

High-throughput dermatological data collection, labeling, and redness detection

Labeling Tool

GUI interface for rapid labeling of dermatologically relevant images.

Usage:

• Download the full contents of the LabelingTool folder and run "python DemoLabeler.py"
• The image file name and the remaining number of unlabeled images are indicated at the top
• Accept or reject the image
• Use the buttons below to provide reasons for the decision
• Click Submit to save your rating to a text file.
• Click Skip to skip irrelevant images
• (Optional) Include additional information about the image in the Notes form field section

(Module dependencies: tkinter, google_images_download, PIL)

Redness Detection

Identifies afflicted red regions of a person's face and provides an overall redness score

Usage:

• Download the full contents of the RednessDetection folder
• Download pretrained Keras weights Keras_FCN8s_face_seg_YuvalNirkin.h5 into this folder (https://drive.google.com/uc?id=1alyR6uv4CHt1WhykiQIiK5MZir7HSOUU&export=download)
• Run the python script with an image of a person's face as input (for example, "python Red.py 4.jpg")
• Outputs image of face with afflicted region mask, and prints redness score

(Module dependencies: opencv-python, keras, tensorflow)

Video Tracking

Collect face images/video adhering to strict lighting and resolution metrics

Usage:

• Download the full contents of the VideoTracking folder
• Run "python webcam.py"
• The webcam will record photo/video of the face once lighting and face resolution criteria are met

(Module dependencies: webcam.py)