Dae Lim C
- Research Program Mentor
MS at Columbia University
Expertise
Machine Learning/Computer Science
Bio
I enjoy understanding ML systems that has the promise to augment our inferences made on the world. Applying these powerful models to domains/realities that you already enjoy observing yourself is truly special! If you have a facet of the world that you consider precious, try allowing neural nets to see it as well. (It will gift you with fascinating perspectives) I enjoy going for very long runs. I also love clothes: I enjoy learning about designers and learning to style clothes. I love traveling for a long time at a single place.Project ideas
AI Stylist
We can experiment with vision models (e,g ViT, OpenAI CLIP, Inception) to allow computers to aid us in discerning whether a pair of outfit is compatible or incompatible. Potential directions include: -Collecting a dataset of compatible and incompatible outfits, then training a model to predict if two garments go well together. The final deliverable could be an app that acts as a virtual stylist, letting users upload photos of clothing to get pairing recommendations. -Implementing an outfit generator that creates novel stylish looks from a wardrobe. This could combine garment compatibility prediction with other techniques like style transfer. The generator could be showcased in an interactive demo.