Stanford University
Kubar.io
Cowlar Design Studio
Built a reusable canvas annotation component for machine learning workflows with rectangle, polygon, freeform, inspect, click, and delete modes.
Built a reusable canvas annotation component for machine learning workflows with rectangle, polygon, freeform, inspect, click, and delete modes.
View repositoryAdapted transformer-based global encoding with U-Net localization for preprocessing-optimized breast tumor segmentation.
View repositoryBuilt a semantic attention segmentation model on the Montgomery X-ray dataset and reached an 89% Dice score.
View repositoryIntegrated channel and spatial attention into Xception to improve classification from radiography features.
View repositoryDesigned a U-Net variant with wider convolutions, dropout, GridSearchCV tuning, and morphological refinement.
View repositoryUsed grayscale conversion, thresholding, and edge detection to isolate fingerprints and enhance ridge visibility.
View repositoryMask-transformer-based networks for teeth segmentation in panoramic radiographsMDPI Bioengineering, Impact Factor 3.7 / First co-author
Master's student, Kyung Hee University, Republic of Korea.Research interests: diffusion models, LLMs, and VLMs.
BS in Computer Engineering, National University of Science and Technology (NUST), Islamabad, Pakistan.Featured coursework: Computer Vision and Digital Image Processing.