mwambafancy-tech/Mwamba-Vision-Classifier

Android TensorFlow Lite image classifier built by Mwamba Mutale. # Mwamba Vision Classifier Mwamba Vision Classifier is a computer vision and Android image classification project built by **Mwamba Mutale**. It includes an Android TensorFlow Lite app plus a Python training pipeline for CNNs, transfer learning, preprocessing, dataset handling, model evaluation, and TensorFlow Lite export. ## Features - Capture an image with the device camera. - Select an image from the device gallery. - Run on-device inference with TensorFlow Lite. - Display the predicted class instantly. - Clean mobile interface branded for Mwamba Mutale. - Train a custom CNN from class-named image folders. - Train a MobileNetV2 transfer-learning model. - Evaluate with accuracy, precision, recall, F1-score, and confusion matrix. - Export trained models to Android-ready TensorFlow Lite format. ## Model The bundled model is stored at: ```text app/src/main/ml/model.tflite ``` Current class labels: ```text Apple Banana Orange ``` The app resizes input images to `32 x 32` pixels before passing them into the model. ## Tech Stack - Java - Android SDK - Python - TensorFlow / Keras - TensorFlow Lite - TensorFlow Lite Support Library - scikit-learn - Matplotlib - Gradle ## Machine Learning Pipeline The training workflow lives in: ```text training/ ``` It supports: - Dataset loading from folders such as `training/data/apple`, `training/data/banana`, and `training/data/orange`. - Image preprocessing with resizing, batching, caching, augmentation, and prefetching. - Custom CNN training. - MobileNetV2 transfer learning. - Evaluation reports with classification metrics and confusion matrix. - Export to `app/src/main/ml/model.tflite`. Training docs: ```text training/README.md docs/MODEL_CARD.md docs/TESTING.md ``` ## Run Locally 1. Open the project in Android Studio. 2. Let Android Studio sync Gradle dependencies. 3. Connect an Android device or start an emulator. 4. Run the `app` configuration. Command-line test run: ```bash GRADLE_USER_HOME=.gradle-user-ho …