Russell L. Agrusa (BS '76), founder and CEO of ICONICS, Inc., a Massachusetts-based manufacturing and building automation firm, understands that global demand for innovative automation and connectivity technology is growing rapidly. As such, he sponsors this annual competition to encourage and reward outstanding student research work in these fields.
Eligible projects may include, but are not limited to, Internet of Things (IoT), mobile/wireless systems, artificial intelligence (AI), and cloud computing.
Russ and the selection committee select the three winning projects based on the extent to which they:
The selection committee awarded a total of $13,000 in the 2021 competition.
For the 2021 competition, 17 teams have submitted projects. Below is the list of 16 projects submitted by a total of 33 individuals (30 from CSE), and 18 faculty mentors (16 from CSE), in no particular order, which are eligible for the cash prizes.
When the only East bound lane on an East-West street is closed for work, what to do with the East bound traffic? In this project, CSE PhD student candidate Foad Hajiaghajani Memar designed and prototyped a low-cost system consisting of two intelligent traffic lights equipped with cameras, image processing units, and wireless connectivity. The system can be used to regulate two-way traffic sharing the remaining West-bound lane without human flaggers, thus reducing worker injuries and costs in work zones.
To accelerate, decelerate or cruise along? – that is the question whether it is a human-driven car or an autonomous vehicle when it approaches a signaled intersection. In this project, a team of CSE students led by Foad Hajiaghajani Memar designed and implemented an intelligent intersection crossing algorithm and system wherein autonomous vehicles can optimize its acceleration and deceleration for improved safety and efficiency (in terms of both travel time and energy) based on the Signal Phase & Timing (SPaT) information received from a traffic signal controller.
Pressure ulcers occur in up to 23% of patients in long-term and rehabilitation facilities and at an incidence of 10% to 41% in ICU patients. The AHRQ reported nearly 2.5 million individuals are affected by pressure ulcers, which cost over 11 billion dollars per year in the US. In this project, an interdisciplinary team led by CSE student Huining Liu proposed a transformative solution to reduce pressure ulcer using a low-cost light sensor and commodity smartphone. This smartphone-based solution can enable “see- through” the skin for evaluating subdermal vasculature of the wound region. The system can be used for high-precision self-assessment of complex wound regions in daily life or nursing homes.
The AI-synthesized fake media (e.g., GAN-synthesized faces) are eroding our trust in online information and have already caused real damage. In this project, CSE PhD student Shu Hu developed a tool that automatically identifies AI-synthesized fake face photos by analyzing the inconsistency of corneal specular highlights between two eyes. The tool can proactively defend individuals from becoming victims of AI-synthesized face attacks and help our society to contend with the pandemic of fake digital media.
Are you tired of handling countless receipts and forgetting about them eventually? Are you frustrated from having to input all your expenses into apps and then "tracking" them? Well fret no more, our project helps you manage all your expenses across all the stores at any place in the world so that you don't have to carry those foot long receipts and also don't have to input all your expenses to track those because "your expenses should track themselves". With an integrated payment system and an expense report categorized by product type, store, and time, you get all you need at your fingertips and no more carrying too many cards. If you are a store owner, get detailed insights on your customers and stock well for the right age groups according to what they buy and what the actual "demand" is.
How does a self-driving car recognize pedestrians and other vehicles on the street? In this project, CSE PhD candidate Jialian Wu designed an online real-time object tracker that can recognize and track multiple categories of objects by cameras. This system is the core part of a self-driving car for perceiving and detecting various objects around the car, such that the car can control itself and plan its driving routes. This system can not only predict object 3D locations in the world but also precisely generate object boundaries and shapes.
Home security has been a major issue due to the increase in crime rates. Automatic door locking systems have become a popular solution to provide security effectively. In this project, a team of students led by CSE student Saj Maru have developed a product that provides a safe and efficient solution for controlling home automation. The system is designed to control the door lock wirelessly through commands over the internet or Bluetooth, thereby eliminating the traditional method of using keys. The system is more reliable, convenient, and safer than the conventional way of physically unlocking the door.
Waste Management in the United States and other countries globally has emerged as a major concern over the past few years. The rise in urban population and economic growth in the absence of an effective management mechanism has manifested in the current state of solid waste management in the US which is far from perfect. As a solution to this problem, a team of students led by CSE student Saj Maru have developed a solution that includes an automated recognition system using a Deep learning Algorithm in Artificial Intelligence to classify objects as dry, wet, E-waste, or biohazard waste, Based on this classification, the nearby community will be notified whether the dumped waste from their locality has been segregated or not as well as inform them about the presence of banned plastic in the classified waste.
When you need to communicate with Deaf and hard-of-hearing people, what to do if you have no knowledge of sign language, and do you want a convenient tool to help you recognize and translate sign language? In this project, a team of UB CSE students led by Yincheng Jin designed and prototyped a low-cost, easy-to-use, and highly accessible earphone system to break the communication barriers between the deaf and hearing people. The system can recognize sign language gestures, show the transcription on the smartphone, and potentially play the virtual voice in the earphone through the text-to-speech function.
This project focuses on developing user-friendly translation devices to bridge the communication gap between hearing and Deaf and Hard-of-Hearing (DHH) community. In this project, first-year PhD student – Lipisha Chaudhary, mentored by Dr. Ifeoma Nwogu and Dr. Karthik Dantu, implemented a video-to-text sign language translation system, SignNet – a two-way framework built using Neural Machine Translation techniques which facilitates end-to-end sign language interpretation. This project, after being deployed on a personal programmable device, will act as a 2-way interpreting agent to simplify open communication between non-signer hearing and Deaf and Hard-of-Hearing (DHH) individuals, whose communication is otherwise restricted to hand-operated devices.
Whether you’re in a car or wheelchair, on a bicycle or scooter, having accurate information on road geometry features such as the grade (slope), elevation, curvature and transverse slope for every road segment along the route you plan to travel can be very useful. Unfortunately, neither Google nor any other commercial digital map provides good enough information on road geometry with adequate coverage and frequent data updates. In this project, CSE PhD student Abhishek Gupta designed and implemented a crowd-sourced system to obtain accurate road geometry features using commodity smartphones. Leveraging the ubiquity of smartphones and the power of crowd-sourcing, the proposed system can be crucial in developing a high accuracy, cost-effective and scalable solution for mapping road geometry features.
Aarogya is the one-stop solution that leverages a culmination of AI, NLP, cloud-based services, and medical diagnostics for depression detection and prevention. This is leveraged through a technical and medical analysis of the social media, questionnaire, and therapist notes to provide the user with the analysis, in addition to future directions and suggestions. With this solution, we look forward to a well-balanced community that is built on innovation and technology for collective growth and well-being.
Trusted Execution Environments (TEE) protect sensitive code and data on commodity devices thanks to new hardware security features, such as Intel SGX and Arm TrustZone. Even though the existing TEEs bring many benefits, they have some crucial drawbacks, such as increased TCB size, context switching overhead, single TEE for the system. CSE Ph.D. student, MD Armanuzzaman designed and implemented BYOTEE (Build Your Own Trusted Execution Environments) framework using FPGA SoC devices to overcome those drawbacks. Using BYOTEE, developers can build multiple equally secure TEEs on-demand with configurable hardware and software TCBs to execute their security-sensitive applications.
This COVID pandemic has made us rethink how we interact with biometric devices. Traditional contact-based fingerprint scanners elevate the risk of spread of contagious viruses. In this project, CSE PhD student Bhavin Jawade developed a contactless fingerprint acquisition and matching solution that works with a user's smartphone. By acquiring a user's fingerprint using their smartphone camera and then matching them against legacy contact-based fingerprints remotely on a secure server, we can achieve a hygienic, secure and a modern solution for fingerprinting. As part of this project, we have built a robust contactless-to-contact based fingerprint matching algorithm (Accepted in WIFS 2021), developed an application for acquisition of contactless fingerprints, and collected a large scale multi-finger cross-sensor fingerprint dataset that will further help boost research in this domain.
When in the midst of an epidemic how should the governmental authorities decide on the most efficient mitigation strategy? In this project, CSE Ph.D. student Nitin Kulkarni developed a framework based on a reinforcement learning algorithm to optimize governmental responses to the state of the epidemic at each time-step. The framework is able to take into consideration the government’s priorities in a specific region and then adapt to the rapidly changing characteristics of the epidemic spread given the advances in disease treatment methods and public health interventions. Our framework is validated based on the COVID-19 data collected from New York State, USA.
Video streaming on mobile devices requires high energy consumption and becomes easily a burden for mobile users due to bitrate greedy approaches. CSE Ph.D. students Bekir Turkkan and Adithya Raman designed an energy-aware algorithm that can reduce the energy consumption by up to 48% without sacrificing users' quality of experience.