n
noorusaba_23

Noor u Saba

@noorusaba_23

Python Developer, Machine Learning, Deep Learning, Cpp, Data Structures Projects

Pakistan
Anglais, Ourdou
Certaines informations sont présentées en anglais.
À propos de moi
Hi, I'm Noor, an AI/ML & Python developer helping clients turn raw data and ideas into working, reliable AI solutions. Stuck with a messy dataset or need a model that actually performs? I build custom ML/DL models, image classification & object detection (YOLO), computer vision solutions (OpenCV), and end to end Python/AI pipelines clean, tested, and well documented. Tech:Python | TensorFlow/Keras | OpenCV | YOLO | NumPy | Pandas | Scikit learn Why work with me: clear explanations, tested deliverables, fast communication, real dataset experience. Send your requirements let's build it together!... Plus d’infos

Compétences

n
noorusaba_23
Noor u Saba
hors ligne • 

Voir mes services

Vision par ordinateur
I will build computer vision and image processing solutions using python opencv
De90 $US / projetPlus de détails

Expérience professionnelle

Freelance

Freelance

Indépendant • 8 mos

Fingerprint Biometrics Research Intern

Jun 2026 - Present • 4 mos

Worked on a fingerprint biometrics pipeline using the DB100 dataset and NBIS tools (mindtct, Bozorth3). Implemented Gabor filter based image enhancement with overlap add processing and Hanning window blending to eliminate block grid seam artifacts and reduce false matches. Analyzed multiple preprocessing pipelines and achieved ~98% Rank 1 matching accuracy with the best performing combination. Also built a pose estimation pipeline using Generalized Hough Transform and the Kabsch algorithm to estimate rotation and translation between fingerprint impressions.

Deep Learning Image Classification Projects

Jun 2026 - Present • 4 mos

Built CNN-based image classification models from scratch in Python using TensorFlow/Keras. Developed a facial expression recognition model on the FER2013 dataset with custom per class rebalancing (oversampling underrepresented classes, undersampling overrepresented ones) to handle class imbalance. Also built a cat/dog image classifier, handling real world data issues such as corrupt images and dataset synchronization.