- Machine Learning • Deep Learning • Computer Vision
- Image Processing • Applied AI Research • Video Processing • Object Detection
- Cloud & MLOps (AWS, Docker, Kubernetes)
Occlusion Resilient Human Pose Estimation
This is an ongoing projoct, still implementing and comparing models for occlusion-robust pose estimation using Mean Teacher, ResNet101, and regularization with human pose priors.
Focused on unsupervised domain adaptation and visibility‑guided curriculum learning.
MRI Brain Image Generation
Designed a Deep Convolutional GAN for synthetic MRI brain data generation.
Optimized generator–discriminator structure for stable training and realistic medical image synthesis.
Deep Learning Face Detection Model at MODERERING
Built and trained a deep learning face detection model achieving 0.02 cross‑entropy loss and <4% mean absolute error, fine‑tuned on the MRCD dataset to handle face variation, with experiments deployed on AWS SageMaker.
Programming: Python, MATLAB, C++, Bash
Frameworks & Libraries: PyTorch, OpenCV, scikit‑learn, NumPy, pandas, Matplotlib
ML Skills: CNNs, GANs, ResNet, Transformers, Transfer Learning
platforms / Cloud: AWS SageMaker, Linux
Methodologies: Agile, Git‑based Development
📄 View my résumé (Google Drive)
📫 Email : shams.ben.mefteh5@gmail.com
💼 LinkedIn
