ianasasira/BIRYOMUMEISHO_JOSHUA_AML_ALNet_CAPSTONE_PROJECT

Capstone project for Artificial Intelligence and Data Science (UTAMU) Ms. Artificial Intelligence. This repository contains the implementation of ALNet, a deep learning-based decision support system for automatic detection of Acute Myeloblastic Leukemia (AML) , including model training, evaluation, desktop application, and project documentation. # ALNet -- Automatic Detection of Acute Myeloblastic Leukemia **MSc AI & Data Science Capstone Project** **Student:** BIRYOMUMEISHO JOSHUA (JAN26/MAIDS/0819U) **Supervisor:** DR. KASSIM KALINAKI (PhD) **Institution:** Universal Technology and Management University (UTAMU) --- ## Overview ALNet (Acute Leukemia Network) is a lightweight deep learning model for automated AML screening from peripheral blood smear images. The model uses depthwise separable convolutions and localized sparse multi-head self-attention to extract morphological features while maintaining a compact 27,393-parameter footprint -- suitable for deployment on standard laboratory hardware. The project includes a **standalone Windows desktop application** (`ALNet_Screening_Tool.exe`) that bundles the trained model for one-click deployment by lab technicians -- no Python, TensorFlow, or any dependency installation required. ### Key Features - **ALNet Architecture**: Dual-branch hybrid model -- depthwise separable convs + sparse attention - **Weighted Focal Loss**: Engineered for 74:1 class imbalance in AML screening - **Desktop App**: CustomTkinter GUI with drag-drop image input, real-time inference, confidence scores, SQLite audit logging - **Standalone .exe**: PyInstaller-packaged -- runs on any Windows machine with one click - **Metrics**: AUC-ROC 0.9675 | 27,393 params | **Note:** The .exe is ~2.5 GB because it bundles PyTorch, CUDA, and all dependencies. No installations needed. ### Option 2: Run from Source ```powershell # 1. Clone the repo git clone github.com cd BIRYOMUMEISHO_JOSHUA_AML_ALNet_CAPSTONE_PROJECT # 2. Install dependencies pip install torch torchvision --index-url download.pytorch.org pip install -r requirements.txt # 3. Run the desktop app python src/desktop_app.py # 4. (Optional) Build the standalone .exe pip install pyinstaller python src/build_exe.py ``` --- ## Report The c …