01 / Graduation project · 2025–2026
FaceVote — Deepfake Detection
My CSC496 graduation project at King Saud University, developed in a four-person team under Prof. Khalil ElHindi. I designed a video-level pipeline to sample frames, detect and crop faces, classify them using transfer-learned CNNs, and combine predictions through majority voting.
- Defined a controlled comparison of ImageNet-pretrained MobileNetV2, ResNet50, and EfficientNetB0, considering accuracy and computational cost.
- Face detection and preprocessing for frontal or near-frontal faces in videos.
- Planned training on FaceForensics++ and cross-dataset evaluation on Celeb-DF.
- Specified consistent preprocessing and evaluation using accuracy, precision, recall, and F1-score.
Scope: Video-based face manipulation detection. The research report defines the comparison and evaluation plan; final experimental results are not presented here.
- Deep learning
- Computer vision
- Transfer learning
02 / Parallel computing
Integral Calculation Using CUDA
A CSC453 team project implementing numerical integration on the GPU using fixed trapezoidal integration and adaptive quadrature.
- Assigned integration intervals to GPU threads and combined partial sums using shared-memory reduction.
- Used adaptive interval splitting and atomic operations to manage parallel work.
- Used the integral of sin(x) from 0 to π, whose exact value is 2, as a correctness reference.
- CUDA
- C / C++
- GPU computing
- Numerical methods
03 / Machine learning
Football Match Prediction
An academic machine learning project exploring the prediction of football match outcomes from historical match data.
- Prepared historical data and selected match-related features.
- Explored classification and model evaluation for match outcome prediction.
- Python
- Pandas
- Scikit-learn
- Machine learning