iML: Interactive Deep Machine Learning Data Apps (webpage-based) 

 

​

​

​

For my work so far, see:

​

​

https://harry-muzart.github.io/

https://github.com/Harry-Muzart/harry-muzart.github.io

​

This is a project I am currently working on (Jan - Aug 2018, more TBC). I build:

 

#1 --- Deep Convolutional Neural Networks, train those with large datasets, and use them for Objection Recognition in scenes of people and other items. The visual field is screened with sets of 2-dim matrices. First, simple edges are detected, then these are combined into more complex shapes with deeper layers. The objects are classified into their respective labels. The positional x,y information for the labels, as well as the %age confidence, is calculated and outputted. The packaged libraries, module dependencies, code, and datasets, have been adapted from:https://github.com/thtrieu/darkflow   https://github.com/pjreddie/darknet (using Python 3.6 ; AnacondaCmd/Cloud ; TensorFlow1.py ; OpenCVis3 ; numpy ; cython/darknet ; YOLO cfg weights ; ImageNet ; .json ) .  This was all done from my MS DOS Windows system on my personal machine. Here there will then be a user-based interactive web-based Machine Learning application which will use deep convolutional neural networks to classify fMRI & dtMRI hippocampus-neocortex data in connectivity-strength groups with inference, as a computational model for clinical decisions, the source code will be made open-source on GitHub with push/pull commits welcome.   I will eventually then relate those to NeuroImaging data - based on research literature. I will also be using ML on neuroimaging data itself - with nilearn (Python) and 3D Slicer (Python/C++), using openfmri.com nii files. This can then be used to generate models of Biological Neural Networks. Also see https://www.bioneurotech.com/c-cogn-s for cloud-supported Matlab DCNNs use.

​

​

​

​

​

​

​

​

​

​

http://harrymuzart02.pythonanywhere.com/

​

​

#2 --- Neural Networks for Sentiment Analysis Classification of Linguistic Textual Info. The general-purpose AI can be extended horizontally in its functions: eg. I will also be setting up a internet-browser-based user-input-based NLP (natural language processing) system - that is the emotionality of speech content. (Using Python, PythonAnywhere, Flask, SQLite, PythonML, PHP, MS Access, SKLearn, Html5/Css, G-WForms, Tensorflow.js ). There will be a system for linguistic sentiment analysis similar to  https://github.com/rasbt/python-machine-learning-book/tree/master/code/ch09 . My PythonAnywhere scripts, will be linked to Plesk, PHP/SQL scripts and the Flask micro-framework system. Google Cloud Platform TPUs (tensor processing units) will be used to train and test neuroimaging data.

​

This will be implemented as an interactive app directly in this site's webpage, which users can use to input data and extract data.

 

​

​

​
 

​

​

https://sites.google.com/site/harrymuzart1/hci-data-collection

​

#3 - Human Psychometric Data Collection

​

​

​

​

​

​

​

This rest of this webpage is in the process of being set-up...

​

​

​

Also see Sections C-NhaNp-S and C-Cogn-S for now.

​