SkinLogic is a mobile application built with Flutter that provides immediate, on-device analysis of various skin conditions using a trained Machine Learning model. This project demonstrates the capability of integrating deep learning models directly into mobile applications for real-time inference, offering quick insights into potential skin issues.
Disclaimer: This application is for informational and educational purposes only and is not intended to be a substitute for professional medical advice, diagnosis, or treatment. Always consult with a qualified healthcare professional for any health concerns.
- On-Device ML Inference: Utilizes a lightweight TensorFlow Lite (TFLite) model for fast, offline skin condition analysis.
- Image Input: Users can capture photos using their device's camera or select images from their gallery.
- Prediction Results: Displays the primary predicted skin condition along with a confidence indicator (High, Moderate, Low).
- Detailed Probabilities: Offers an option to view the full list of probabilities for all detected classes.
- User-Friendly Interface: Clean and intuitive design for ease of use.
- Cross-Platform: Built with Flutter, supporting both Android and iOS (tested primarily on Android).
The current model is trained to identify 9 common skin conditions:
- Actinic Keratosis
- Atopic Dermatitis
- Benign Keratosis
- Dermatofibroma
- Melanocytic Nevus
- Melanoma
- Squamous Cell Carcinoma
- Tinea Ringworm Candidiasis
- Vascular Lesion
(Note: Accuracy may vary. The model's performance can be significantly improved with larger and more diverse training datasets.)
- Flutter: Mobile application development framework.
- Dart: Programming language for Flutter.
- TensorFlow Lite (TFLite): For deploying and running the machine learning model on device.
tflite_flutter: Flutter plugin for TFLite integration.image_picker: For selecting images from gallery or camera.image: Dart package for image processing (resizing, pixel manipulation).permission_handler: For managing camera and storage permissions.- Python (for Model Training):
- TensorFlow / Keras: For building and training the deep learning model (MobileNetV2 Transfer Learning).
- Google Colaboratory: For cloud-based GPU training.
- Dataset: Skin Disease Classification Image Dataset from Kaggle
Follow these instructions to set up and run the project locally.
- Flutter SDK installed (stable channel recommended).
- Android Studio / VS Code with Flutter and Dart plugins.
- An Android or iOS device/emulator.
- Clone the Repository:
git clone https://github.com/MarsadMaqsood/skinlogic.git cd skinlogic - Install Dependencies:
flutter pub get
- Run the Application:
(Ensure a device/emulator is connected.)
flutter run
The pre-trained .tflite model and labels.txt are included in the assets/model/ directory.
- Model Architecture: MobileNetV2 (fine-tuned using transfer learning).
- Input Image Size: 224x224 pixels.
- Normalization: Pixel values scaled from 0-255 to 0-1.
- Training Dataset: Skin Disease Classification Image Dataset from Kaggle
- Train on significantly larger and more diverse datasets (e.g., ISIC, HAM10000) for improved accuracy and robustness.
- Implement advanced preprocessing techniques.
- Explore more sophisticated model architectures or ensemble methods.
- Add features like user history, information about diseases, or integration with external APIs for consulting.
- Improve UI/UX, especially for handling uncertain predictions.
Feel free to fork this repository, open issues, or submit pull requests.
This project is licensed under the MIT License. See the LICENSE file for details.
- The Flutter and TensorFlow communities for their invaluable resources and tools.
- Riya Eliza Shaju for providing the Skin Disease Classification Image Dataset on Kaggle.