Students who registered for the synchronous version of the course formed teams and worked on their own deep learning-powered products.
Whether you're looking for your next startup idea or deciding how to improve your portfolio, we hope these projects inspire you to build something real with DNNs!
Many of these projects were made possible thanks to a generous donation of GPU-accelerated compute infrastructure by LambdaLabs. Check them out if you're looking for on-prem or cloud GPU machines!
If you're interested in working on full stack projects, join us on Discord and post/ask around about group project work.
An ML powered application for streamlining the process of creating chapter markers and lesson summaries for course content creators.
A full-stack ML-powered website that utilizes users’ webcam feeds to answer open-ended questions requiring outside knowledge.
Green-Screen Image Composition-Transfer
An ML-powered app for adding (optionally Stable Diffusion-generated) virtual backgrounds to images that uses style transfer to match lighting anad composition.
Weak Supervision and Active Learning with Text Data
An approach to minimise human labelling for text classification tasks.
X-Ray Diagnosis AI Assistant
An interface to support medical practitioners in diagnosing and interpreting x-ray images.
Team: Arun Hegde, Samarth Keshari, Amulya Badal, Ross Cheung, Seyhan Karakulak GitHub Repo.
Mom's AI Food Logger
An app for my mom that automatically identifies and tracks the food she eats.
Team: Prince Javier Live Demo.
Archaeological Feature Detector
A prototype web app to help archaeologists interpret automatically detected objects as part of a machine-learning-powered survey workflow.
Semantic Search Engine for Images
A semantic text search engine over images, along with monitoring.
An image to recipe food classifier.
A pragmatic approach to identifying illustrated pages in digitised historic books.
Full Stack Stable Diffusion
A deployment of Stable Diffusion Text-to-Image and Image-to-Image pipelines with a full stack architecture.
Team: Okan Ulusoy and Omid Abdollahi Aghdam GitHub Repo.
Multimodal Fusion Models for Healthcare
An architecture for using multiple modalities of healthcare data to train deep learning models.
Measure the diameter of nanofibers in microscopy images.
An app that guesses the location of an image, video, or video url.
👻 Image Anonymiser
An ML-powered image anonymisation web app.
An interface for red-teaming open source text generation models from the Hugging Face hub.
Board Game Rules Explainer
A board game question-answering system to save players from having to check the rulebook.
Gesto AI - ASL Word Recognizer
A real-time, word-level American Sign Language translation app.
Choosistant helps you decide which product to buy by summarizing pros and cons from written reviews.
Team: Kimotho, Murad Khalilov, Nam, Omar Ali Sheikh, Sofiane Chami Project Page.
Semantic Search & Sentiment Analysis
Upload a PDF or text document and enable semantic QA and sentiment analysis.
Run modern neural networks directly in your browser from a computer or phone.
Animate a cartoon with facial expressions using only your voice.
OCR SemSearch allows you to perform semantic search on text within images from different types of documents.
Team: Sebastian Gonzalez Aseretto, Paramtap Mewada Project Poster.
Live Art in Context
Draw on the creative power of modern ML models to create art responsive to events in text or video streams.
Team: David Murphy, Angel Carvajal, Theresa Thoraldson, Chris Lonsberry Slide Deck.
A plant species identifier available as a web app and as a cross-platform mobile app.
A data product for multi-class semantic segmentation of earth observation images using a UNet architecture. Team: Suzana, Roland Ritt, Sheebo