nearlyphd/masters-thesis---melanoma-analysis-with-fnn

Melanoma analysis with Fractal neural networks

Melanoma analysis with Fractal neural networks

The purpose of this research is to compare the performance of a convolutional neural network, MobileNet V2, Inception ResNet V2, and our own fractal neural network.

Melanoma

Melanoma, also redundantly known as malignant melanoma, is a type of skin cancer that develops from the pigment-producing cells known as melanocytes. Melanomas typically occur in the skin, but may rarely occur in the mouth, intestines, or eye (uveal melanoma). In women, they most commonly occur on the legs, while in men, they most commonly occur on the back. About 25% of melanomas develop from moles. Changes in a mole that can indicate melanoma include an increase in size, irregular edges, change in color, itchiness, or skin breakdown.

Pic.1. A melanoma of approximately 2.5 cm (1 in) by 1.5 cm (0.6 in)

Getting started

Prerequisites

• docker
• docker-compose >= 1.28
• NVIDIA Container Toolkit

Steps

1. Clone or download the repo.
2. In the terminal, go to /docker.
3. Run docker-compose up -d.
4. Run docker logs jp.
5. Open a browser and go to the URL, you got after step 4.

masters-thesis---melanoma-analysis-with-fnn's Issues

remove transformations, it may make out model less performant

https://github.com/amyshenin-tech/masters-thesis---melanoma-analysis-with-fnn/blob/facecf3c52e9eef896e332e6929050f1468cc9e3/simple_cnn.py#L8

try using PLGrind on the .py file

use fit_generator, instead of fit

https://github.com/amyshenin-tech/masters-thesis---melanoma-analysis-with-fnn/blob/facecf3c52e9eef896e332e6929050f1468cc9e3/simple_cnn.py#L48

https://datascience.stackexchange.com/questions/34444/what-is-the-difference-between-fit-and-fit-generator-in-keras

place all downloaded files into one folder and use validation split during training

use new 40K dataset

example of model

https://github.com/meldev00/skin_lesion_cnn/blob/master/cnn_models/model_7_final.ipynb

ROC curve

ROC for multiclass classification: https://stackoverflow.com/questions/37017400/sklearn-metrics-roc-curve-for-multiclass-classification
use categorical mode for image generator

• write to tensorboard https://groups.google.com/a/tensorflow.org/g/discuss/c/MkDnT-T1HgA?pli=1
• add breakdown by classes + average

add weights to classes, when val_los is calculated

https://datascience.stackexchange.com/questions/13490/how-to-set-class-weights-for-imbalanced-classes-in-keras

use check points

https://github.com/amyshenin-tech/masters-thesis---melanoma-analysis-with-fnn/blob/facecf3c52e9eef896e332e6929050f1468cc9e3/simple_cnn.py#L48

https://medium.com/swlh/a-high-level-overview-of-keras-modelcheckpoint-callback-deae8099d786