mohamed-tsx/Somali-AI-Models

In this repo i'll add ton of somali ai models which i trained with their dataset and training codes stay tuned insha allah # Somali-AI-Models Welcome — this repository collects small, focused Somali-language machine learning projects. The goal is practical research and tooling that helps explore Somali audio, text, and other modalities. Each project is self-contained with dataset pointers, notebooks, and (where available) a saved model for quick inference. ## Projects in this repo - Somali Dialect Detector - Location: `Somali Dialect Detector/` - Purpose: classify short Somali audio clips into dialect categories (e.g., Maay vs Standard Somali). - Contents: audio dataset, training notebooks (`Training Code/`), and a saved model (`Model/somali_dialect_model.pkl`). See the project README at `Somali Dialect Detector/README.md` for full details and quick-start instructions. More projects can be added as separate folders with the same pattern: `Data/`, `Notebooks/`, `Model/`, `scripts/`. ## Quick start 1. Clone the repo and create an environment: ```powershell git clone cd "Somali AI Models" python -m venv .venv .\.venv\Scripts\Activate.ps1 pip install -r requirements.txt ``` 2. Open the project-specific README for details. For the dialect detector: ```powershell jupyter notebook "Somali Dialect Detector/Training Code/Somali_Dialect_Detector_v1.ipynb" ``` 3. To run inference you will typically load a saved pipeline and run the same feature extraction used during training — refer to `Somali Dialect Detector/README.md` for a minimal example. ## Data & models - Datasets live inside each project folder (`Somali Dialect Detector/Audio Dataset/...`). Keep datasets alongside code for reproducibility. - Saved models (if present) are under `Model/` inside each project folder. These are usually serialized with `joblib`/`pickle` or framework-specific formats. ## Development notes - Notebooks are the source-of-truth for preprocessing and experimentation. If you plan to productionize code, extract notebook logic to `scripts/` or a small `src/` packa …