Ghazaleh T. Email & Phone Number

Montreal, QC, CA

Canada

Montreal, Quebec, Canada

• Developed and trained action recognition models to recognize key actions of customers in stores (e.g., picking up and putting down items).
• Built an inference pipeline of customer detection, customer tracking, multi-camera re-identification, pose estimation, and action recognition.
• Model optimization for edge deployment including GPU/CPU utilization, runtime memory usage, and model size/storage.
• Video acquisition, anonymization, encoding, streaming, and storage in the cloud.
• Collaborated with Product and Engineering teams to identify problems and come up with creative solutions.
• Involved in selecting hardware and defining system setup for client projects.

Montreal, Quebec, Canada

Montreal, Quebec, Canada

• Studied Vision-based Human Activity Recognition (HAR) methods.
• Used Computer-Vision to recognize atomic activities of construction workers (e.g., hammering, walking) and excavators (e.g., digging, loading).
• Improved spatiotemporal human activity recognition approaches for far-field videos.
• Prepared a video dataset of thousands of frames for activity recognition of workers and excavators using CNNs.
• Combined sequences of construction workers' atomic activities to recognize higher-level micro-tasks (e.g., formwork construction) for productivity analysis of workers and project digital twin creation.
• Used object detection (YOLOV3) to detect completed structures for progress tracking.

• Researched on CNN-based Stereo Vision systems for autonomous driving and their implementation on FPGAs toextract depth information from images
• A Siamese CNN was trained on the KITTI dataset for the task of stereo matching

Master Of Science - Ms, Electrical And Computer Engineering, 4.15

● Worked as a research assistant in the lab. ● Participated in Mitacs Accelerate Fellowship/internship with Hydro-Quebec. ● Improved.

Bachelor Of Science - Bs, Electrical Engineering

Thesis: Power Line Detection for Autonomous UAV Navigation using CNNs (2018 – 2019) UAV’s are used to eliminate the need for manual.