SculptML
Building and Deploying Neural Nets from Your Pocket
SculptML is an iOS app designed to make deep learning model-building, deployment, and experimentation possible without writing any code.
Using the click-and-swipe model-building screen, the user can assemble a Tensorflow deep learning architecture from Keras-level layer abstractions, send the model to a server for training (with a selected dataset), and test trained image-input models by taking pictures or drawing on a notepad.
This project was done through Virtual Reality @ Berkeley’s Mobile AR team, where I co-lead and developed the project with classmate and Berkeley student Noah Gundotra.
Category
Education & Deep Learning
Organizations
Virtual Reality @ Berkeley (Mobile AR Team, with co-lead Noah Gundotra)
Project Dates
Fall ‘18 - Spring ‘19 Semesters
Click-and-Swipe Model Building
Testing a Trained MNIST Model