Unified Machine Learning Tool

This paper is a written illustration of a GUI based tool for Machine Learning. This paper, analyses the need and working of the Unified Machine Learning Tool Project. The central idea of this project is to reduce the efforts involved in machine learning model training and testing by developing a central tool, which provides a GUI based platform for the developers to easily achieve their task. The project provides options for, in which way model has to be trained. This project also aims to reduce the hardware toll, so that the enthusiasts can also use this project to start off with machine learning. This project improves upon the most common machine learning neural network implementation, the Convolutional Neural Network using face detection, which then produces a face only image of the person which in turns produces high accuracy models to recognize a particular person, and also using random image manipulations to increase datasets and increase the model’s ability to recognize any object in a non-ideal condition such as the object being in an angle with the frame and object being small in the frame. This project also allows users to train model for mobile devices, as the mobile devices do not have the same amount of computational power as compared to computers, they require a mobile model which is very light as compared to their full-fledged counterparts. The project also features functionality of batch testing, in which the directory of the testing data is selected and the project tests if the model is correctly detecting the data in accordance to user defined minimum confidence/probability threshold, this batch testing mode is available for both a mobile model and a full-fledged computer model. The project also provides the functionality of testing the model with a single image file, with the Single File Testing mode and Mobile Single File Testing mode. The project features a Pro mode that enables the user to customize the training method and have greater control over how the model is being trained. This allows the user to get a more customized machine learning model. The Pro mode may include setting the number of rounds, settings custom layer names etc. This Pro Mode is also available for testing of a model, for if a customized model has to be tested. The project on boot up checks for python and its version as the fast retraining of a pretrained model, as this project does, requires at least the python version 3, and as this project aims to automate many functions that the user had to do manually it requires a lot of pre-installed modules, the project on boot up also checks if all the required modules are installed.