RESEARCH & dEVELOPMENT

Published in a Q1 journal (Artificial Intelligence in Agriculture with an Impact Factor 10.26). Isaac’s contributions in the fields of Food, Water, Energy, and Healthcare have also been presented at International IEEE conferences Intelligent Computing Instrumentation and Control Technologies (ICICICT), Advanced Computing and Communication Systems (ICACCS), Circuit Power and Computing Technologies (ICCPCT), Electronics and Renewable Systems (ICEARS), Intelligent Technologies (CONIT) which all have now been indexed under the Scopus database.

AWS DeepRacer

Racing to the Top 100 Racers of the AWS DeepRacer Student Championship 2022 - India

Given a limited amount of training time (10 hrs) during the qualifiers, trying out multiple reward functions being a terrible idea, the choice of utilizing a complex reward function dampened the model's capability to learn the environment but in the end, it was a simpler solution that did the trick (to find the optimal policy). During times, at the brink of giving up, I received immense support from Dr.Vinodh Ewards, for which I am truly grateful.


If you would like to know more about the reward functions that I used, contact me below !.

Top 100 Winner of AWS Deep Racer Student Championship (#48 INDIA )

Powered by NVIDIA Jetson Nano, RTX GPU

HACKATHON TEAM EVENTS

GLOBAL ACADEMIC EXCHANGE (Singapore)

Ranked No.1 in Asia and currently 8th position in the world ( QS ), Isaac was among the few individuals to have qualified for the global academic exchange program, a joint program held by the National University of Singapore (NUS) and HPE (Hewlett Packard Enterprise). The program is entitled ‘Data Analytics using Deep Learning’.

Presenting his team’s final product ‘VIDA - developed to support Accurate waste segregation’ under the theme of global sustainability, with student members from different universities across the world.

The goal of the project was to ease the process through automated waste segregation when passed through a conveyor belt through deep learning solutions. They developed a real-time vision-based system that utilizes image processing to identify different waste materials, A fine-tuned Custom Convolutional Neural Network was trained and deployed in the Cloud (Azure Kubernetes) so that real-time predictions can be made.

Building an Assistive Navigation Hardware for the Hard-to-See using Haptic Feedback - Winning Team at Equipathon 2023 Sponsored by Intel



Driven by the Nvidia Jetson, inside is a DenseNet model actively determining the position of the individual on the road, based on whose feedback (model) the Servo motors assists the individual in re-directing towards the point of interest. In addition, to tackle dynamic transitory objects like pedestrians, bikes & cycles the distance of these objects during approach is calculated using functions determining the depth using the point from the image of reference achieving its goal of obstacle detection with its proximity precisely, Nonetheless, GPS integrations were made to track the individual in the global scene.

The model was trained on the One API platform utilising Tensorflow optimizations from Intel which reduced the training period by half. Along with Fusion 360 aiding in designing an ingenious 3D model of the glass (frame) that can hold the 2 servo motors.

BUILDING ASSISTIVE TECHNOLOGY

  • FUSION 3D PRINTING

    Fusion 3D played a key role in refining the assistive navigation hardware design, facilitated iterative prototyping for testing component fit, allowing for adjustments to the 3D model in for optimal functionality and comfort, used to create specific custom components for the device.

  • HARDWARE DESIGN & DEVELOPMENT

    Uniquely designed frame holds servo motors for haptic feedback based on user's position, aiding anyone in navigating obstacles and staying safe while moving from point A to point B. Potential further uniqueness in custom components, sensor integration, or lightweight comfort design.

  • ARTIFICIAL INTELLIGENCE and hardware integration

    The project integrates AI with hardware: Convolutional model guides user, detects obstacles & measures distance, all using computer vision and sensor data for a safe and dynamic navigation experience.

  • FUSION 360

    Fusion 360 powered the assistive navigation device. Not just design, it likely facilitated 3D printed prototypes for crucial testing and user feedback. This synergy of virtual tweaks and physical iterations refined the frame, even enabling custom-designed components for a personalized solution.

COLLABORATE