System and Method for Identifying Products in a Shelf Management System
A system and method for automated product identification on retail shelves using deep learning feature extraction and matching against a product gallery library.
M. Savvides
About
I'm a Senior Applied Scientist at Microsoft, building multimodal query recommendation systems for Bing Image Search. I work at the intersection of computer vision and language, turning modern ML into reliable, high-impact product experiences.
I'm interested in VLMs, multimodal retrieval, and recommender systems. If you're building in multimodal AI, I'm happy to chat - especially around 0→1 prototypes and productionizing models.
Education
M.S. in Electrical and Computer Engineering
2019-08-01
2020-12-31
Carnegie Mellon University
B.Eng. in Electronics and Telecommunication Engineering
2015-06-01
2019-05-31
University of Mumbai
Interests
Vision Language Models
Recommendation Systems
Computer Vision
Machine Learning
Internet of Things
Featured Work (Publications & Patents)
•
•
1 min read
Detecting Hate Speech in Multi-modal Memes
A multi-modal approach to detect hate speech in memes using object detection, image captioning, and sentiment analysis for the Facebook Hateful Memes Challenge.
Multi-Image Steganography Using Deep Neural Networks
A deep learning approach for encoding and decoding multiple secret images inside a single cover image of the same resolution.
Siamese Manhattan LSTM Implementation for Predicting Text Similarity and Grading of Student Test Papers
This paper presents a method to grade answer papers written by the students by assessing the semantic similarity between the written answers and the actual answers and grading them …
W. Dalmet
Talks
Research Ideas for the Facebook Hateful Memes Challenge
We present Object Detection based Image Captioning, and Sentiment Analysis as our two research ideas to enhance performance against the adversarial examples introduced in the …
Selected Writings
How to build a Convolutional Neural Network from scratch using tf.keras for Image Classification
Predict survival on the Titanic challenge in MATLAB
All Publications, Patents & Writing
(2022).
System and Method for Identifying Products in a Shelf Management System.
US Patent US20220051179A1.
(2022).
Detecting Hate Speech in Multi-modal Memes.
arXiv 2012.14891.
(2021).
Multi-Image Steganography Using Deep Neural Networks.
arXiv 2101.00350.
(2019).
Siamese Manhattan LSTM Implementation for Predicting Text Similarity and Grading of Student Test Papers.
In ICWiCOM 2019.
(2019).
Study of Performance of an OFDM Transceiver Using SDR Platform.
In ICWiCOM 2019.
(2018).
Smart Car Features Using Embedded Systems and IoT.
Proceedings of DJ Strike 2018. ISBN 978-93-5300-801-7.
(2018).
Smart Refrigerator Using Internet of Things and Android.
Proceedings of DJ Spark 2018. ISBN 978-93-86724-75-5.