Sidrah Liaqat

Machine learning research engineer — deep learning, computer vision, and video understanding.

I am a machine learning research engineer with a PhD in Electrical Engineering from the University of Kentucky (September 2026), advised by Dr. Samson Cheung. I build deep learning and computer vision pipelines for action recognition and behavior interpretation from video — most recently for autism spectrum disorder behavior detection from infant videos. I work in Python with PyTorch, TensorFlow, and OpenCV. Check out my resume.

Latest release — ASD-FEAT: a multi-modal infant video dataset and benchmark for early ASD risk prediction, with open-source code. Code on GitHub · Dataset DOI: 10.5281/zenodo.22261227

Featured Projects

ASD-FEAT: Multi-Modal Infant Video Dataset & Benchmark RELEASED

Released dataset + open-source code (Zenodo DOI, MIT-licensed GitHub repo) for early ASD risk prediction from infant video. Transformer-based frame-level behavior detection over fused face, gaze, pose, I3D, and audio features, followed by session-level risk classification.

ASD Behavioral Analysis from Video

Deep learning + computer vision pipelines for autism-related behavior detection from infant video. PyTorch/TensorFlow, OpenCV.

Predicting Autism in Children from Eye-Gaze Patterns

Predicting autism in children from how they visually explore images — combining gaze-fixation heatmaps with synthetically generated saccades to augment scarce clinical data. Published in Signal Processing: Image Communication (2021).

Bird Audio Detection (DCASE 2018)

A CNN ensemble that detects bird calls in field recordings — UKYSpeechLab's submission to the DCASE 2018 Bird Audio Detection Challenge, built to generalize across mismatched acoustic domains.

Earlier Research

Radar Ground-Target Recognition & Micro-Doppler Analysis

Recognizing ground-surveillance-radar targets (pedestrians, vehicles) from their micro-Doppler signatures — a fast real-time feature-based classifier paired with high-resolution time-frequency signal representations. EuRAD 2013 / INISTA 2011.