Nailed it!

predicted fingernail segments masked over test image
DRF hosts the server which is used to deploy the model
Evaluation of performance using Intersection of Union vs Number of Epochs

Inspiration

The initial inspiration came from the prize category "Best Women's Health & Well-being Hack". Anemia is very prevalent in women, and often goes undiagnosed until it becomes very severe.
So we wanted to make an application that could use a segmentation model to recognize fingernails in images, and analyze their pixel color values to calculate Haemoglobin.
From there the project condensed down to where it now is.

What it does

It uses a U-net CNN segmentation model to segment out portions of images which contain fingernails. The model was trained on Jupyter Notebook and was deployed using Django REST Framework.

How we built it

The model itself is a U-net segmentation model trained on an open-sourced dataset of fingernail images. We deployed it using DRF's api to submit images using the post method in order to get predictions.

Challenges we ran into

We all live in different time zones so maintaining communication while collaborating was a challenge. Other than that, we also found that the project to be very technically advanced and the time too little to incorporate all what we had planned for it.

Accomplishments that we're proud of

Learning how to use DRF to deploy a trained ML model and working out how to collaborate together while being on the opposite sides of the globe. Just bringing this project up to where it now is is a huge accomplishment in itself.

What we learned

Tons of stuff!

Making a U-net model for fingernail segmentation.
Using Django's Rest Framework API to send post requests to a trained model in order to make predictions.

What's next for Fingernail Segmentation

A lot of what we had in mind for the project initially, but were not able to do because of the time limit.

Disease detection via color deviation: We could see how the average color across the segmented section deviates from the normal average color of a fingernail.
High deviations towards some colors could indicate the presence of diseases.
For example, a heavy deviation towards yellow could indicate Bleeding, diabetes, digestive problems, liver disease and more.
Disease detection via pattern abnormailities: Different patterns on the fingernails could be indicative of different types of diseases.
For example, Vertical Ridges and Split Nails could indicate Vitamin A deficiency or nervous problems.
Anemia detection: It can be analysed how Heamoglobin scales with fingernail color, and further use this knowledge to detect anemia in those with low heamoglobin.