Improved StyleGAN Embedding: Where are the Good Latents?

Simple script to compute CLIP-based scores given a DALL-e trained model.

Simple script to re-rank images using OpenAI's CLIP https://github.com/openai/CLIP.

Using fourier feature mapping Strengthen GON (《Gradient Origin Networks》)'s performance

A dynamic method for converting tabular data to images to apply CNNs on tabular data.

This project exhibits an aesthetically pleasing design with vibrant colors and a responsive layout. The use of 'Hanken Grotesk' Google Font enhances readability, contributing to an engaging and polished user experience.

Using Graph Attention NN for image embedding and classification

A method for steganography using a modified cycle GANs that's trained on texture dataset and analysed by steganalysis tool.

Check out this image similarty model live @ https://dinomatch.drismir.ca/

A multimodal similarity search tool for textile patterns. Embeds 10,000+ objects from The Met and Cooper Hewitt using CLIP to capture image and text features. Stores results in ChromaDB for fast semantic search. Built to demonstrate end-to-end multimodal ML, data, and retrieval pipeline skills.

It contains a few useful notebooks from "Shopee - Price Match Guarantee" competition hosted on kaggle. It covers topics ranging from EDA, Efficientnet for image embeddings, ArcFace loss for metric learning, LR scheduler, etc.

Semantic image clustering using CLIP and GPT-powered metadata in a Next.js + MongoDB stack

Convert an image to Base64 data URL in your browser

This is an implementation of Least Significant Bit Embedding.