Intel® Student Ambassador

AI Poster Presentations

7 Search Results

Toward Autonomous UAV Flights in Cluttered Forestry Environments

Bruna Pearson

Autonomous flight within a forest canopy is a key challenge for generalized scene understanding onboard a future unmanned aerial vehicle (UAV) platform. To address this challenge, see how automatic trail navigation successfully generalizes across differing image resolutions. This allows UAVs with varying sensor payload capabilities to operate equally in such challenging environmental conditions. The optimized deep neural network architecture delivers state-of-the-art performance across varying resolution aerial UAV imagery. Use it to improve forest trail detection for UAV guidance even when using significantly low-resolution images, which are representative of low-cost search and rescue capable UAV platforms.

Self-Driving Vehicles & Risk Analysis

Edwin Williams

Guidance, navigation, and control (GNC) systems take data from the environment to create a mathematical representation of the system state. This presentation demonstrates a method to create a guidance solution based on system risk for a system’s goal. The methodology uses a goal to create a situation, which creates a situation model whose risk can be determined. This risk assessment can then be learned by pattern recognition systems to make decisions on how to control a vehicle.

Medical Image Processing for Brain Tumor Recognition

Moloti Tebogo Nakampe

Brain tumors are the second most dangerous cancer in children, accounting for almost as much as all solid tumors in the body combined. Knowledge about this cancer is insufficient, and the general public and clinicians tend to know much less about brain cancer in children than leukemia. Despite some progress being made over the last two decades, overall patient outcomes remain poor when compared with other childhood cancers.

Use AI to Enhance Electronic Health Records

Pallab Paul

This project attempts to solve the problem of accurately transferring individual patient records when a patient changes doctors by creating a blockchain that is accessible to any medical office as long as they are part of the network. The solution uses various machine learning methods, including deduplication and record linkage, to help simplify the data being entered into these records. It also uses a mixture of machine learning and natural language processing algorithms to better understand the data and to make future predictions.

Invasive Water Hyacinth Monitoring on Lake Victoria

Ngesa Naphtally Marvin

Recently, in Lake Victoria, Africa, water hyacinth has become a major invasive plant species and the plant has negatively affected local ecosystems. The weeds create excellent breeding sites for mosquitoes and other insects, which results in an increase of diseases such as skin rash, cough, malaria, encephalitis, gastro-intestinal disorders, and bilharzia (schistosomiasis). Despite several efforts, water hyacinth in Lake Victoria continues to expand. This project uses drone technology to monitor the source of water hyacinth, how it spreads, and track its movement in the lake. The data can assist organizations to anticipate its growth and work to stop it.

Cash Recognition for the Visually Impaired Using Deep Learning

Kshitiz Rimal

In Nepal, local currency has no special features to aid visually impaired persons who are then dependent on other people, such as friends, family, or shopkeepers, to perform transactions. This project aims to assist the visually impaired person by using a smartphone application and an offline, in-device deep learning model trained on MobileNet. When the application hovers over the cash item, it classifies it and plays a sound to indicate its value. This offline mobile-based application can be used in shops, public transportation, on the street, and many other diverse locations.

Visual Serving of Thin & Small Objects Using Learning Methods

Hanz Cuevas Velasquez

Automating rose pruning is an unexplored area that involves different computer vision problems because of the complex nature of the rose bushes. This research identifies three main steps that a robot should follow to perform this task and uses a convolutional neural network architecture and reinforcement learning:

  • Plant modelling to segment the branches
  • Object detection to detect the eye-buds on the bush
  • Visual surveying to make a robotic arm approach the branches and eye-buds