Nicholas Cheung's Projects
A versatile unsupervised factor-based text tokenizer that streamlines natural-language processing applications
An efficient real-time object detection application powered by lightDenseYOLO. It combines lightDenseNet and YOLO v2 for faster and more accurate object detection, trained on MS COCO and Pascal VOC 07+12 datasets.
An advanced model dedicated for Monocular Depth Estimation and Image Segmentation with custom data set.
A dynamic tool that enables classifying 3D point clouds using various (pretrained) classification models and parameters. It supports lifting certain objects based on predefined criteria e.g. class or neural net response. You can prepare training data for models via 2D projections.
This repository contains the Pytorch implementation for the project titled "SAM-Path: A Segment Anything Model for Semantic Segmentation in Digital Pathology", initially admitted for oral presentation at MICCAI2023. All the credits go to the original authors.
A tool employing active learning for semi-supervised image segmentation which can be interactive. Suitable for applications like lesion/tumor segmentation.