top of page

Artificial Intelligence portfolio

Live Lightning Detection with Deep Learning and Tensorflow on Android

The Internet is full of tutorials about detecting objects in images. But can you apply the same concepts to a live video stream?
Yes, you can! In this article series, we’ll demonstrate how to use AI to determine what’s going on in a live video stream by building a lightning detector that runs in realtime on an Android device.
The title of the series – "Live Lightning Detection ..." – sounds too emotionally charged. Poetry? Cinema? Or a marketing trick? Let’s parse the "emotion" into technical terms. Live = in a set of consecutive frames that constitute video; Lightning Detection = nothing but an object detection, using Deep Learning (DL) and Tensorflow (TF) implemented in an app running on the Android OS. Sounds more up your alley?

Tags: DeepLearning, AndroidStudio, LiveDetection, ObjectDetection, LightningDetection, Android, Tensorflow, TensorflowLite, MLkit, TensorflowModel, MachineLearning

An Improved Approach of Intention Discovery with Machine Learning for POMDP-based Dialogue Management

An Embodied Conversational Agent (ECA) is an intelligent agent that works as the front end of software applications to interact with users through verbal/nonverbal expressions and to provide online assistance without the limits of time, location, and language. To help to improve the experience of human-computer interaction, there is an increasing need to empower ECA with not only the realistic look of its human counterparts but also a higher level of intelligence. This thesis first highlights the main topics related to the construction of ECA, including different approaches of dialogue management, and then discusses existing techniques of trend analysis for its application in user classification. As a further refinement and enhancement to prior work on ECA, this thesis research proposes a cohesive framework to integrate emotion-based facial animation with improved intention discovery. In addition, a machine learning technique is introduced to support sentiment analysis for the adjustment of policy design in POMDP-based dialogue management. It is anticipated that the proposed research work is going to improve the accuracy of intention discovery while reducing the length of dialogues.

Tags: Human-Computer Interaction, Q-Learning, POMDP, Sentiment Analysis, Reinforcement Learning, Machine Learning, Artificial Intelligence, 3D model, ECA, Decision-making Process, Interaction

Google AI open images – visual relationship track

In this work, we have tried to solve the Visual Relationship Detection Track competition launched by Kaggle. The aim of the competition is to check if computers can detect the relationship between objects presented in images. Not only it is a very state-of-the-art research area, but it is also a very challenging task to accomplish compared to existing computer vision tasks. It is a combination of two prominent tasks – object detection and image caption generation. Although deep learning models are able to produce highly accurate results in these individual tasks, it still struggles to perform with acceptable accuracy in the visual relationship detection task. In this paper, we have attempted a different approach to solve this problem and compared the result with the state-of-the-art baseline result. We have explored two attention-based caption generation models and modified them to solve the visual relation detection task.

Tags: Machine Learning, Computer Vision, Natural Language Processing, Deep Learning, Image Processing, Object Detection, Auto caption

Machine learning and Pattern recognition approaches on Prostate Cancer data

A machine learning approach to identify meaningful biomarkers by recognizing the possible patterns in the genes data given from the cBioPortal platform for the prostate cancer data. By having 495 different samples (patients) with 60K+ genes combinations, classification and feature selection, solving multi-class problems, dimensionality reduction, etc. techniques will be applied using clinical dataset along with the genes dataset. The visualization will be applied to follow up using seaborn and matplotlib libraries to bifurcate distinguished features using a different kind of maps and graphs. Right estimator algorithms are applied to test the accuracy to avoid the overfitting. Different classifiers are applied with the standard procedures and the accuracy is measured using cross-validation. The classifiers used in this scenario are Stochastic Gradient Descent (SGD), Support Vector Machine (SVM), Nearest Neighbours (NN), Naïve Bayes, Forest and tree methods. Hence, previously selected features using feature selection methods will be compared with the complete set of datasets. Each estimator has a different set of hyperparameters and using a different set of hyperparameters combinations using grid search approach over best performing classifier model. 

Tags: Machine Learning, Data Science, Pattern Recognition, Data Analysis, Classification, Feature Selection

Spdybot: Authenticity Analysis of Opinion and Claim in Online

This project, called Spdybot, presented an end-to-end learning model without any human participation. The team’s proposed solution is capable of mining all opinions available on the web on a certain topic along with relevant evidences so that users can decide the authenticity of each opinion before taking them into account. The user is presented with visual statistics of a list of opinions with supporting and contradicting evidence and a credibility score that they can use to decide if an opinion deserves sufficient authenticity

Tags: Opinion Mining, Artificial Intelligence, Authenticity, Natural Language Processing

Anti-Glossophobia ARVR (MR) application

Enhance personal, emotional, social, presentation skills as a speaker to avoid the fear of stage and speech while facing the audience and crowd with virtual reality enabled approach using Augmented Reality as a platform to overcome Glossophobia (speech anxiety) with voice and gesture activated virtual audience to be able to stand up & speak with the assistance of VR headset and/or mobile device and/or computer.

Tags: Virtual Reality, Augmented Reality, Mixed Reality, Anti-Glossophobia, Speech Anxiety, Google VR, VR box, VR Headset

Using collaborative filtering with frequent sequential patterns mining for product recommendation systems in multiple E-commerce Databases

The research was conducted to represent how improvements can be made for product recommendations using the lifetime-user prediction technique.

Tags: Collaborative Filtering, Sequential Pattern Mining, E-commerce, Database mining 

bottom of page