Dishant Padalia
Computer Science Master’s Student at UMass Amherst
University of Massachusetts Amherst
Manning College of Information and Computer Sciences
Biography
I am a graduate student pursuing an MS in Computer Science at the University of Massachusetts (UMass) Amherst. Previously, I’ve interned at the Indian Institute of Technology (IIT) Bombay, working under the guidance of Prof. Ganesh Ramakrishnan and Dr. Venkatapathy Subramanian. During my undergraduate studies at K.J. Somaiya College of Engineering (KJSCE), I worked as a Research Assistant for Prof. Ninad Mehendale. Additionally, I collaborated with Prof. Deepak Sharma (Vice Principal of KJSCE) on two research projects.
My research interests primarily revolve around Computer Vision, with a specific focus on the application of Artificial Intelligence in healthcare and medical diagnosis. I am also intrigued by the intersection of Natural Language Processing and Computer Vision in fields such as Medical Image Report Generation and Document Image Translation.
In my free time, I enjoy playing chess and board games.
- Computer Vision
- AI in Healthcare
- Natural Language Processing
- Information Retrieval
MS in Computer Science
University of Massachusetts Amherst
BTech in Electronics & Telecomm Engg
K.J. Somaiya College of Engineering (Mumbai University)
Experience
- Developed a Transformer-based OCR (TrOCR) model for handwritten Hindi documents, increasing Word Recognition Rate by 41% and Character Recognition Rate by 10%.
- Created an algorithm that generates synthetic Indian language image datasets for improving model generalization and robustness.
- Authored research paper on digitizing scanned documents while preserving layout structure and OCRing content.
- Pitched our software to potential clients, landing contracts with two of them.
- Conducted technical interviews for hiring junior interns and guided them throughout their tenure.
- Conducted research on breast cancer detection from mammography scans using image processing and deep learning techniques.
- Designed an algorithm for pre-processing of mammography scans using CLAHE and K-Means Clustering, boosting the model accuracy by 17%.
- Analyzed various CNN architectures on the benchmark MIAS dataset to inform model selection.
- Developed an Enhanced EfficientNet (EEF-Net) achieving 97.14% accuracy and 98.30% F1-Score, outperforming existing methods in performance and precision.
- Created a Crop Health Monitoring system to classify agricultural fields into five crop health classes.
- Utilized satellite images and implemented a U-Net architecture for crop segmentation, considering color, texture, and density features.
- Led the development of a government project aimed at automating License Plate detection for non-helmeted riders.
- Engineered a 3-stage system using the YOLO model for detecting Motorcycle, Helmet, and License Plate, successfully deployed at multiple signal junctions in Pune city.
Projects
DiagnoSys.AI
Web-based diagnostic system to detect brain tumors from MRI scans and cataracts from fundus images of the eye. Developed a CNN-LSTM model for cataract detection and fine-tuned an EfficientNet-3 model for brain tumor classification.
Image Captioning
Trained an Encoder-Decoder type model using a combination of InceptionV3 and LSTM model on the Flickr-8k dataset for generating sentence-based descriptions for an input image. Built a mobile application with the functionality of uploading an image and generating a caption.
Green Route
Every year, 86,000 people die due to delayed healthcare services. Green Route, a mobile app, helps ambulances move faster through traffic by alerting nearby Google Maps users within 1 km of the ambulance’s path to clear the way for the ambulance.
Hateful Speech on Twitter
Hateful and Offensive speech classification using 30k randomly extracted tweets from Twitter. Applied pre-processing techniques on tweets and designed an LSTM model for binary classification.