DermWise

Empowering Skin Health: Predict, Prevent, Prevail

Comment

Inspiration

The problem at hand revolves around the need for a reliable and accessible platform that empowers individuals to identify and understand common skin diseases promptly. With the rise of digital health solutions, there is a gap in the market for a user-friendly and accurate skin disease prediction app.

What it does

Key Components of the Project Early Detection Timely Intervention Accessibility Education and Awareness

How we built it

The system comprises a React Native UI for seamless mobile app interaction, a Python backend handling data processing and server-side logic, and a trained model specialized in detecting skin conditions. Users interact with the React Native UI on their mobile devices, which communicates with the Python backend. The backend orchestrates requests, processes data, and interfaces with the trained model for skin condition detection. We used a type of model called a Sequential model. This means we can stack layers one after the other, creating a linear flow of data. Our model consists of two main parts: a feature extractor layer and a dense layer. Before we can train our model, we need to configure some settings. We chose the 'adam' optimizer, which is like a guide that helps the model adjust and improve its predictions over time. We defined the 'sparse_categorical_crossentropy' as our loss function. Together, these components form an integrated system providing users with a convenient, efficient, and reliable tool for skin condition detection via their mobile devices.

Challenges we ran into

Model Adaptability: Constant pivoting due to the inefficiency of trained models to effectively address real-world scenarios highlights the challenge of adapting models to diverse and dynamic conditions.

Data Adequacy: The limited corpus for skin disease data underscores the challenge of acquiring and curating a sufficiently diverse and comprehensive dataset for robust model training and validation.

Technical Integration: Ensuring seamless integration between the React Native UI, Python backend, and trained model presents a technical challenge, requiring compatibility and efficient communication between disparate components.

Sustainability: Sustaining system accuracy amidst evolving user needs and technological advancements is an ongoing challenge, necessitating continuous model refinement, data augmentation, and system updates to maintain relevance and effectiveness over time.

Accomplishments that we're proud of

We built the react-native app from scratch. Found a novel way to deploy and test the app using ngrok. Our model had a very high rate of accuracy 97.01%

What we learned

Iterative Problem-Solving: The hackathon taught us the importance of embracing iterative problem-solving approaches, especially when faced with challenges like model inefficiencies and limited data. Each pivot served as a learning opportunity, driving us to refine our strategies and adapt our solutions to better address the underlying issues.

Data Accessibility and Quality: The experience highlighted the critical significance of data accessibility and quality in machine learning projects. We learned the importance of meticulously sourcing, curating, and augmenting datasets to ensure the robustness and generalizability of our models, particularly in domains like healthcare where data availability can be limited.

Interdisciplinary Collaboration: Collaborating across disciplines – from UI/UX design to backend development and machine learning – underscored the value of interdisciplinary collaboration in tackling complex problems. Leveraging diverse skill sets and perspectives enabled us to approach challenges from multiple angles, leading to more comprehensive and innovative solutions.

What's next for DermWise

On-boarding doctors for real time feedback and consultation on the app. Adding disease detection for more diseases. Making the app HIPPA compliant.

Built With

Share this project:

Updates