A Novel Implementation of LiDAR Mesh Classification and Image Classifiers in Assistive Technology for the

Overview

Current assistive technologies for the 253 million visually impaired are unsuitable for numerous reasons: unaffordability, poor ergonomic design, and computational inefficiency. Oracle is an application that efficiently classifies obstacles and relays their relative proximity using auditory and haptic feedback. Oracle utilizes the LiDAR sensor on an iPhone 12 Pro to classify objects by using convolutional neural networks to identify depth shapes. To increase classification variety, Oracle employs supplemental image classifiers, capable of classifying objects that are too geometrically complex to be identified through LiDAR depth data. To relay pertinent audio-haptic feedback to the user, Oracle implements a sectioning algorithm, dividing the camera’s view into 40 subsections, sending raycasts to the center of each section. Both the sectioning algorithm and a hierarchy algorithm calculate which object is most pertinent based on distance and obstacle type. Classification accuracy, distance accuracy, and navigational time were measured to test the device’s efficacy. The models (LiDAR and Image ML) had a theoretical accuracy of 99% and empirical accuracy of 97.4% in their classifications. Additionally, Oracle had an average overall absolute error of only 3.09% in its measurement of distances and yielded an average navigational time that was 28.8% lower than the White Cane across 10 obstacle courses. Statistical tests display statistical significance between Oracle’s navigational time when compared to that of a White Cane, confirming the alternative hypothesis. The device can be used for affordable, non-intrusive navigational assistance, replacing technologies such as electronic canes and furthering intelligent assistive technology research. Development of a Targeted Drug Delivery System for the Treatment of Covid-19 Sahil Sood Lambert High School, Suwanee, GA SARS-CoV-2 has triggered a public health outbreak across the world, resulting in almost 5 million deaths as of January 2022. The arrival of vaccines has provided temporary relief, but these vaccines target the spike protein, which is highly prone to mutation, making it impossible to develop a long term cure to the coronavirus. As such, there is an urgent need for site-specific inhibition of the virus in the respiratory tract, as well as targeting the internal proteins of the virus itself. Past literature has identified 3CLpro and PLpro as enzymes essential to replication of the virus, as they assemble almost the entirety of the viral genome; as such, inhibiting the activity of these enzymes can stymie the spread of the virus. This project proposes the use of inhaled drug delivery to inhibit Covid-19 by synthesizing a formulation that can travel directly to the lungs via inhalation. In order to streamline synthesis, existing FDA-approved drugs were analyzed using computational docking software and in vitro assays for inhibitory activity against these two enzymes. High-performing drugs were then encapsulated in PLGA nanoparticles to synthesize a drug delivery system, which was tested and characterized in vitro. Furthermore, in an effort to improve this drug delivery system relative to other drug delivery systems, the use of enzyme nanomotors was explored as a way to increase the accuracy of delivery by using computational simulations that mimicked conditions in the human body to model the velocity and trajectory of the nanomotors. GREATER WASHINGTON, D.C.

Competition history

• JSHS 2022 Category not listed

Resources

• National JSHS abstract book ↗

Related projects

Closest projects by meaning, across every fair and year in the corpus.