AI & Computer Vision Journey
From Data Science foundations to Deep Learning for Computer Vision
// courses, projects & certificates
How I Learned
Data Science Foundations
CompletedDEPI IBM Data Science Track
Learned Python for Data Science core syntax and foundational programming. Used Jupyter Notebooks, GitHub, and cloud environments. Extracted and queried data using SQL and APIs. Analyzed and visualized data with Pandas, NumPy, Matplotlib, and Seaborn. Built regression, classification, clustering, and recommender systems using Scikit-Learn. Completed a hands-on capstone project applying the full data science pipeline to a real-world dataset.
Machine Learning Specialization
CompletedAndrew Ng — Stanford / DeepLearning.AI
Studied supervised learning (linear/logistic regression, neural networks, decision trees), advanced algorithms (ensembles, clustering, anomaly detection, recommender systems), and practical ML engineering (bias/variance, error analysis, CV pipelines, sklearn workflows). Built a strong theoretical and applied foundation in modern ML.
Deep Learning for Computer Vision
CompletedCoursat.ai — DL for CV Certificate
Studied deep learning fundamentals applied to computer vision — CNNs, backpropagation, activation functions, and training pipelines for image data. Built custom CNN architectures from scratch.
Computer Vision Applications
CompletedCoursat.ai — CV Applications Certificate
Applied CV techniques to real-world problems: image classification with AlexNet and ResNet, facial emotion recognition (FER-2013), and transfer learning using pre-trained models like ResNet50V2.
Attention & Transformers
PausedSelf-study — 'Attention Is All You Need' paper
Studied the Attention mechanism and the Transformer architecture. Worked through the seminal paper, understanding multi-head attention, positional encodings, and how transformers revolutionized sequence modeling and vision tasks.
Object Detection & Mobile CV
PausedAla Mahlak — Driver Monitoring System
Built an on-device real-time Driver Monitoring System fusing Google ML Kit Face Mesh, YOLOv8n (ONNX), head pose estimation via pure Dart solvePnP, and multi-signal fusion. Implemented EAR (eye closure), MAR (yawning), gaze estimation, PERCLOS analysis, and a temporal state machine — entirely offline at ~15 FPS on mid-range Android.
What I've Explored
Ala Mahlak — Driver Monitoring System
On-device real-time DMS for a Flutter ride-sharing app. Fuses Google ML Kit Face Mesh (468 landmarks), YOLOv8n ONNX object detection (phone, food/drink), pure Dart head pose estimation (solvePnP), EAR/MAR gaze analysis, and a temporal state machine with PERCLOS — classifying 7 distraction states at ~15 FPS entirely offline.
OpenCV-Learning
Structured computer vision learning journey with OpenCV and MediaPipe — covers image processing fundamentals, DNN-based object/face detection, pose estimation, hand tracking, and real-time projects.
FER-2013 CNN vs ResNet
Facial emotion recognition on FER-2013 — head-to-head comparison between a custom CNN built from scratch and transfer learning with ResNet50V2 (TensorFlow/Keras). Includes EDA, class-balanced training, and confusion-matrix evaluation.
Intel Image Classification
Image classification on the Intel dataset using two architectures: AlexNet and ResNet18 with PyTorch. Covers the full workflow from data preprocessing through training, evaluation, and inference.
Verified Credentials
Deep Learning for Computer Vision
coursat.ai
Computer Vision Applications
coursat.ai
Machine Learning Specialization
Andrew Ng — Stanford / DeepLearning.AI