Inspiration
We wanted to bring research results and medical applications closer together to minimize the effort and increase the confidence in advanced topics of medicine like cell counting/detection and tissue classification.
What it does
We provide a REST API allowing to send images of the two different problems (cell detection & tissue classification) and returning the classifcation/segmentation results. The Application is not limited as the trained model got deployed on a server so users do not need to worry about technical issues.
How we built it
Trained CNNs in Keras on a DeepLearning VM from azure, deployed the trained models on an azure server and implemented the API around it. Further, we created an Apeer module calling this API, which allow further processings by Apeer users. Last but not least, we implemented a web application also calling the API but in a very user friendly way, by uploading an image on the application.
Challenges we ran into
Setting up the whole infrastructure (Azure, Keras, Apeer, Webserver, Websites, ...).
Accomplishments that we're proud of
We finally managed to run every goal on each endpoint.
What we learned
A lot about Azure, Server Structures, Apeer, Containerization, and of course Data Science ;). Working together as a team from different backgrounds
What's next for Project Cellda
Making the infstructure more stable, include andanced visualizations like heatmaps, confidence interfals, ...
Built With
- apeer
- azure
- depthwise-separable-convolution-module
- html
- javascript
- keras
- python
- scalable-image-recognition-module
- tensorflow
Log in or sign up for Devpost to join the conversation.