Pain Management / Computer Vision Lab - Berk Dikmen

Semester Goals: https://drive.google.com/open?id=1Dy1OaNB9PYbdbOZnqJI8r4FCsOnz5yBE

Week 1

Date: 01/21/2019 - 01/27/2019

Pain Management

 -Software for survey is completed in first semester. I need to a learning algorithm, a database and permissions from local hospitals to gather data.
 -I am planing to start gathering data at the beginning of April end at the end of the month I will evaluate the data I have and compare it to result of original paper. 

GAIC Lab

 -I'm working on a simple GAN (Generative adversarial network) that is highly modifiable for our use in future projects.
 -We're planing to create  a poster for one of the computer vision seminars.

Week 2

Date: 01/28/2019 - 02/03/2019

Pain Management

 -I couldn't meet with Prof. Yin this week so presentation of my work and permission for data gathering delayed.

GAIC Lab

 -Problems with GAN caused some delay but early steps are completed.
 -I need do create a dictionary for my data and increase the accuracy of the predictions my machine makes.

Week 3

Date: 02/04/2019 - 02/10/2019

Pain Management

 -I started to work on implementing keras machine learning algorithm to my facade data I will use to polish my program before starting to test it in hospital.

GAIC Lab

 -I'm working on a GAN (pix2pix) that helps you to create new image based on previous images fed into to system.

Week 4

Date: 02/11/2019 - 02/17/2019

Pain Management

 -My prediction system needs more data (around 5000 to 10000) to predict result more accurately.
 -Thus I'm trying to implement better algorithm that will require less data to function properly.

GAIC Lab

 -I decided to use a different version of pix2pix that is implemented via keras which is more efficient and easier to understand.
 -I also trained acgan (Auxiliary Classifier Generative Adversarial Network) to better understand the inner workings of GAN.

Week 5

Date: 02/18/2019 - 02/24/2019

Pain Management

 -I'm still working on my algorithm.

GAIC Lab

 -We experienced a hardware failure in the lab so I spend this week trying to recover our data and rebuild the system.
 -Data is recovered and system is mostly working with an ongoing problem with tensorflow-keras_contrib compatibility.

Week 6

Date: 02/25/2019 - 03/03/2019

GAIC Lab

 -I completed a full run of pix2pix to have a better understanding of algorithm's inner workings.
 -I started to create first dataset of houseplans that we will use after I modify the algorithm.

Week 7

Date: 03/04/2019 - 03/10/2019

 -

Week 8

Date: 03/11/2019 - 03/17/2019

 -

Week 9

Date: 03/18/2019 - 03/24/2019

 -

Week 10

Date: 03/25/2019 - 03/31/2019

Pain Management

 -I decided to change my machine learning algorithm from supervised to unsupervised which will help me to produce accurate results with less data however I need to find a way to separate unrelated diseases.

GAIC Lab

 -I completed first prototype of our data set.
 -Depending of the behavior of pix2pix I might need to adjust creation method of our data set or find ways to improve it.

Week 11-12

Date: 04/01/2019 - 04/14/2019

Pain Management

 -I decided to scrap my database and use a file based format to save user data.
 -This will help me to analyze data without needing internet connection.
 -Since we don't need to prioritize security this will allow me to code faster without downsides. 

GAIC Lab

 -Tests for data set is still ongoing.
 -Each test takes couple days before producing result I can analyze to configure the data set.

Week 13

Date: 04/15/2019 - 04/21/2019

Pain Management

 -My plan for storing data on a file based system failed due to my survey being web based and modern technology doesn't allow web based applications to make changes on data located on hardware due to security issues that caused a lot of problems in the past.
 -After my meeting with Prof. Foreman I decided to go for mail based data storage can I can gather via a script I will code to scrap data from it.
 -I need to find a way to give feedback to web application to complete system and finish polishing machine learning algorithm.