MediDermis

• GIF

  A walk-through for the web app

Inspiration

Due to the shortages of doctors and clinics in rural areas, early diagnosis of skin diseases that may seem harmless on the outside but can become life-threatening is a real problem. MediDermis uses Computer Vision to predict skin diseases and provide the underlying symptoms associated which are prominent in rural India. The lockdown has not helped either, with the increasing shortage of doctors due to many of them going on COVID duties. Keeping the goal of helping out our community in any way we can, Bhuvnesh Nagpal and Mehul Srivastava decided to create this AI-enabled project to help the underprivileged with one slogan in mind – “Prevention is better than Cure”

What it does

MediDermis uses Computer Vision to predict skin diseases and provide the underlying symptoms associated which are prominent in rural India.

How we built it

The image classification model is integrated with a web app. There is an option to either click a picture or upload a saved one. The model based on resnet32 architecture then classifies the image of the skin disease into 1 of the 29 classes and shows the predicted disease and its common symptoms. We trained it on a custom dataset using the fastai library in python.

Challenges we ran into

Collecting the dataset was a big problem as medical datasets are not freely available. We collected the data from various sources including google images, various sites, etc.

Accomplishments that we're proud of

We were able to make an innovative solution to solve a real-world problem. This solution might help a lot of people in the rural parts of India. We are really proud of what we have built. The app aims to provide a simple and accurate diagnosis of skin disease in rural parts of India where medical facilities are scarce.

What we learned

We brainstormed a lot of ideas during the ideation part of this project and realized that there was a dire need for this app. While developing the project, we learned about the Streamlit framework which allows us to easily deploy ML projects. We also learned about the various sources from where we can collect image data.

What's next for MediDermis

We plan to try and improve this model to a level where it can be certified and deployed into the real-world setting. We can do this by collecting and feeding more data to the model. We also plan to increase the number of diseases that this app can detect.

Built With

• fastai
• ml
• python
• streamlit