• Stars
    star
    208
  • Rank 182,751 (Top 4 %)
  • Language
    Python
  • License
    MIT License
  • Created over 6 years ago
  • Updated about 1 year ago

Reviews

There are no reviews yet. Be the first to send feedback to the community and the maintainers!

Repository Details

Dermatologist AI

Introduction

In this mini project, you will design an algorithm that can visually diagnose melanoma, the deadliest form of skin cancer. In particular, your algorithm will distinguish this malignant skin tumor from two types of benign lesions (nevi and seborrheic keratoses).

The data and objective are pulled from the 2017 ISIC Challenge on Skin Lesion Analysis Towards Melanoma Detection. As part of the challenge, participants were tasked to design an algorithm to diagnose skin lesion images as one of three different skin diseases (melanoma, nevus, or seborrheic keratosis). In this project, you will create a model to generate your own predictions.

Skin Disease Classes

Getting Started

  1. Clone the repository and create a data/ folder to hold the dataset of skin images.
git clone https://github.com/udacity/dermatologist-ai.git
mkdir data; cd data
  1. Create folders to hold the training, validation, and test images.
mkdir train; mkdir valid; mkdir test
  1. Download and unzip the training data (5.3 GB).

  2. Download and unzip the validation data (824.5 MB).

  3. Download and unzip the test data (5.1 GB).

  4. Place the training, validation, and test images in the data/ folder, at data/train/, data/valid/, and data/test/, respectively. Each folder should contain three sub-folders (melanoma/, nevus/, seborrheic_keratosis/), each containing representative images from one of the three image classes.

You are free to use any coding environment of your choice to solve this mini project! In order to rank your results, you need only use a pipeline that culminates in a CSV file containing your test predictions.

Create a Model

Use the training and validation data to train a model that can distinguish between the three different image classes. (After training, you will use the test images to gauge the performance of your model.)

If you would like to read more about some of the algorithms that were successful in this competition, please read this article that discusses some of the best approaches. A few of the corresponding research papers appear below.

While the original challenge provided additional data (such as the gender and age of the patients), we only provide the image data to you. If you would like to download this additional patient data, you may do so at the competition website.

All three of the above teams increased the number of images in the training set with additional data sources. If you'd like to expand your training set, you are encouraged to begin with the ISIC Archive.

Evaluation

Inspired by the ISIC challenge, your algorithm will be ranked according to three separate categories.

Category 1: ROC AUC for Melanoma Classification

In the first category, we will gauge the ability of your CNN to distinguish between malignant melanoma and the benign skin lesions (nevus, seborrheic keratosis) by calculating the area under the receiver operating characteristic curve (ROC AUC) corresponding to this binary classification task.

If you are unfamiliar with ROC (Receiver Operating Characteristic) curves and would like to learn more, you can check out the documentation in scikit-learn or read this Wikipedia article.

The top scores (from the ISIC competition) in this category can be found in the image below.

Category 1 Rankings

Category 2: ROC AUC for Melanocytic Classification

All of the skin lesions that we will examine are caused by abnormal growth of either melanocytes or keratinocytes, which are two different types of epidermal skin cells. Melanomas and nevi are derived from melanocytes, whereas seborrheic keratoses are derived from keratinocytes.

In the second category, we will test the ability of your CNN to distinuish between melanocytic and keratinocytic skin lesions by calculating the area under the receiver operating characteristic curve (ROC AUC) corresponding to this binary classification task.

The top scores in this category (from the ISIC competition) can be found in the image below.

Category 2 Rankings

Category 3: Mean ROC AUC

In the third category, we will take the average of the ROC AUC values from the first two categories.

The top scores in this category (from the ISIC competition) can be found in the image below.

Category 3 Rankings

Getting your Results

Once you have trained your model, create a CSV file to store your test predictions. Your file should have exactly 600 rows, each corresponding to a different test image, plus a header row. You can find an example submission file (sample_submission.csv) in the repository.

Your file should have exactly 3 columns:

  • Id - the file names of the test images (in the same order as the sample submission file)
  • task_1 - the model's predicted probability that the image (at the path in Id) depicts melanoma
  • task_2 - the model's predicted probability that the image (at the path in Id) depicts seborrheic keratosis

Once the CSV file is obtained, you will use the get_results.py file to score your submission. To set up the environment to run this file, you need to create (and activate) an environment with Python 3.5 and a few pip-installable packages:

conda create --name derm-ai python=3.6
source activate derm-ai
pip install -r requirements.txt

Once you have set up the environment, run the following command to see how the sample submission performed:

python get_results.py sample_predictions.csv

Check the terminal output for the scores obtained in the three categories:

Category 1 Score: 0.526
Category 2 Score: 0.606
Category 3 Score: 0.566

The corresponding ROC curves appear in a pop-up window, along with the confusion matrix corresponding to melanoma classification.

Sample ROC curve Sample confusion matrix

The code for generating the confusion matrix assumes that the threshold for classifying melanoma is set to 0.5. To change this threshold, you need only supply an additional command-line argument when calling the get_results.py file. For instance, to set the threshold at 0.4, you need only run:

python get_results.py sample_predictions.csv 0.4

To test your own submission, change the code to instead include the path to your CSV file.

More Repositories

1

self-driving-car

The Udacity open source self-driving car project
Jupyter Notebook
6,072
star
2

deep-learning-v2-pytorch

Projects and exercises for the latest Deep Learning ND program https://www.udacity.com/course/deep-learning-nanodegree--nd101
Jupyter Notebook
4,963
star
3

deep-reinforcement-learning

Repo for the Deep Reinforcement Learning Nanodegree program
Jupyter Notebook
4,528
star
4

deep-learning

Repo for the Deep Learning Nanodegree Foundations program.
Jupyter Notebook
3,940
star
5

machine-learning

Content for Udacity's Machine Learning curriculum
Jupyter Notebook
3,825
star
6

self-driving-car-sim

A self-driving car simulator built with Unity
C#
3,810
star
7

Sunshine-Version-2

The official repository for Developing Android Apps
Java
2,886
star
8

ud851-Exercises

Java
2,048
star
9

ud851-Sunshine

Java
2,014
star
10

ud120-projects

Starter project code for students taking Udacity ud120
DIGITAL Command Language
1,586
star
11

cs344

Introduction to Parallel Programming class code
Cuda
1,227
star
12

frontend-nanodegree-resume

This repository is used for one of the projects in Udacity's Front-End Web Developer Nanodegree program. Learn how to become a Front-End Developer today with line-by-line code reviewed projects and get a job with career services!
JavaScript
1,183
star
13

course-collaboration-travel-plans

CSS
1,162
star
14

DSND_Term2

Contains files related to content and project of DSND Term 2
Jupyter Notebook
1,095
star
15

CarND-LaneLines-P1

Lane Finding Project for Self-Driving Car ND
Jupyter Notebook
779
star
16

CVND_Exercises

Exercise notebooks for CVND.
Jupyter Notebook
744
star
17

ud777-writing-readmes

Supplemental material for Udacity's "Writing READMEs" course
698
star
18

ud839_Miwok

Java
696
star
19

ud867

Course code repository for Gradle for Android and Java
Java
664
star
20

artificial-intelligence-for-trading

Content for Udacity's AI in Trading NanoDegree.
Jupyter Notebook
646
star
21

ud862-samples

Java
585
star
22

AIPND

Code and associated files for the AI Programming with Python Nanodegree Program
Jupyter Notebook
541
star
23

create-your-own-adventure

This is example repo illustrates the concept of a "Pull Request", created as part of the course How to Use Git and GitHub
Shell
503
star
24

ud891

Google and Udacity course on Accessibility.
JavaScript
498
star
25

CarND-Term1-Starter-Kit

Python
497
star
26

Just-Java

The official repository for the second Android Development for Beginners App : Just Java
Java
497
star
27

sagemaker-deployment

Code and associated files for the deploying ML models within AWS SageMaker
Jupyter Notebook
463
star
28

CarND-Advanced-Lane-Lines

Shell
454
star
29

artificial-intelligence

Python
441
star
30

private-ai

Repo for Udacity's Secure & Private AI course
Jupyter Notebook
441
star
31

ud839_CustomAdapter_Example

Java
439
star
32

Advanced_Android_Development

Repo for the Advanced Android App Development course
Java
429
star
33

CarND-Behavioral-Cloning-P3

Starting files for the Udacity CarND Behavioral Cloning Project
Python
425
star
34

cn-deep-learning

Jupyter Notebook
422
star
35

Full-Stack-Foundations

Solution Code to Full Stack Foundations (ud088)
Python
400
star
36

fullstack-nanodegree-vm

Python
372
star
37

CarND-Traffic-Sign-Classifier-Project

Classify Traffic Signs.
Jupyter Notebook
362
star
38

DL_PyTorch

Code for the Deep Learning with PyTorch lesson
Jupyter Notebook
358
star
39

ML_SageMaker_Studies

Case studies, examples, and exercises for learning to deploy ML models using AWS SageMaker.
Jupyter Notebook
355
star
40

ud843-QuakeReport

Discover earthquake activity around the world.
Java
350
star
41

frontend-nanodegree-mobile-portfolio

JavaScript
344
star
42

ud615

Course code for Scalable Microservices with Kubernetes
Go
338
star
43

cs291

Interactive 3D Graphics class code
JavaScript
319
star
44

CarND-Extended-Kalman-Filter-Project

Self-Driving Car Nanodegree Program Starter Code for the Extended Kalman Filter Project
C++
315
star
45

SFND_Lidar_Obstacle_Detection

C++
307
star
46

CarND-Path-Planning-Project

Create a path planner that is able to navigate a car safely around a virtual highway
C++
306
star
47

and-nd-firebase

Course code repository for Firebase in a Weekend by Google: Android
Java
299
star
48

robot_pose_ekf

The robot_pose_ekf ROS package applies sensor fusion on the robot IMU and odometry values to estimate its 3D pose.
C++
293
star
49

frontend-nanodegree-arcade-game

JavaScript
284
star
50

fend

General Front End Nanodegree Content Resources
CSS
275
star
51

FSND

Public repository for the Full-Stack Nanodegree program.
Python
272
star
52

devops-intro-project

Project files for Intro to DevOps class
Shell
269
star
53

rl-cheatsheet

RL Notation and Pseudocode for Udacity's MLND program
TeX
269
star
54

Court-Counter

The official repository for the third Android Development for Beginners App : CourtCounter
Java
264
star
55

Android_Me

App that lets you style your own Android! This uses Fragments to create a flexible and responsive UI.
Java
260
star
56

APIs

Code Repo for API course in Fullstack ND
Python
257
star
57

CarND-Vehicle-Detection

Vehicle Detection Project
Shell
252
star
58

CarND-MPC-Project

CarND Term 2 Model Predictive Control (MPC) Project
C++
251
star
59

nd064_course_1

Python
249
star
60

reactnd-project-myreads-starter

Starter Code for the React MyReads Project
JavaScript
247
star
61

DSND_Term1

Contains files related to content and project of DSND
Jupyter Notebook
247
star
62

ud845-Pets

Java
243
star
63

data-analyst

Content for Udacity's Data Analyst curriculum
HTML
240
star
64

AIND-NLP

Coding exercises for the Natural Language Processing concentration, part of Udacity's AIND program.
Jupyter Notebook
239
star
65

P1_Facial_Keypoints

First project for CVND: facial keypoint detection.
Jupyter Notebook
220
star
66

frontend-nanodegree-styleguide

Official FEND style guides
HTML
220
star
67

NLP-Exercises

Jupyter Notebook
198
star
68

nd1309-work-code

Include for each Part in the ND Program the applications code.
JavaScript
193
star
69

nd9991-c2-Infrastructure-as-Code-v1

Repository for starter code and supporting material
Shell
192
star
70

frontend-nanodegree-styleguide-zh

优达学城(Udacity)前端样式指南
HTML
192
star
71

CarND-Camera-Calibration

Images and notebook for camera calibration
Jupyter Notebook
192
star
72

andfun-kotlin-android-trivia

Kotlin
189
star
73

ud864

Public repository for code examples used in Udacity's Google Maps APIs course (https://www.udacity.com/course/google-maps-apis--ud864).
HTML
188
star
74

CarND-Capstone

CMake
187
star
75

CarND-Semantic-Segmentation

Python
186
star
76

dog-project

Jupyter Notebook
182
star
77

DevOps_Microservices

Supporting material and projects for a course on Cloud DevOps: Microservices.
Jupyter Notebook
181
star
78

ud811

Course materials for Udacity's Intro to Progressive Web Apps course at https://www.udacity.com/course/intro-to-progressive-web-apps--ud811
JavaScript
178
star
79

ShoppingListPlusPlus

ShoppingListPlusPlus is the companion Android app for the Udacity course Firebase Essentials : Build a Collaborative Shopping List App. https://www.udacity.com/course/firebase-essentials-for-android--ud009
Java
176
star
80

CppND-Route-Planning-Project

C++
173
star
81

reactnd-contacts-complete

Code-along project for the Contacts app
JavaScript
171
star
82

ud405

Code supporting the free Udacity class 2D Game Development with LibGDX.
Java
171
star
83

ud859

Course code for Building Scalable Apps with Google App Engine class
Java
170
star
84

ios-nd-networking

Resources for Udacity's iOS Networking with Swift course.
Swift
168
star
85

course-git-blog-project

Sample repo of a blog for the Git course
HTML
168
star
86

intro-to-ml-tensorflow

Projects and exercises for the Udacity Intro to Machine Learning with TensorFlow course
Jupyter Notebook
159
star
87

CarND-Mercedes-SF-Utilities

Tools for Sensor Fusion processing.
Jupyter Notebook
158
star
88

AIND-Sudoku

Python
156
star
89

course-JS-and-the-DOM

JavaScript
155
star
90

didi-competition

Resources for the Udacity/Didi $100k competition
Python
153
star
91

exoplanet-explorer

Learn more about Exoplanets! (built for Promises course)
HTML
146
star
92

ud989-cat-clicker-premium-vanilla

ud989-cat-clicker-premium-vanilla
JavaScript
146
star
93

ud989-retain

JavaScript
146
star
94

CVND---Image-Captioning-Project

Jupyter Notebook
145
star
95

nd9991-c2-Infrastructure-as-Code-v1-Exercises_Solution

This repository contains the solution to the exercises given in the last concept of each Lesson 1 to Lesson 5.
142
star
96

AdvancedAndroid_Emojify

Java
141
star
97

JDND

Public repository for the Java Developer Nanodegree program.
Java
141
star
98

nd027-c3-data-lakes-with-spark

Python
141
star
99

CppND-Capstone-Snake-Game

A 2D Snake game using C++ and SDL
C++
139
star
100

cloudflare-typescript-workers

Types and mocks for building a tested Typescript Cloudflare Worker, generates three NPM packages
TypeScript
139
star