A5003260120- Advanced Image and Video Retrieval Techniques
- Advanced Neural Network Applications
- Medical Image Segmentation Techniques
- 3D Shape Modeling and Analysis
- Multimodal Machine Learning Applications
- Human Pose and Action Recognition
- 3D Surveying and Cultural Heritage
- Natural Language Processing Techniques
- Remote Sensing and LiDAR Applications
University of Cambridge
2022-2024
Medical image registration and segmentation are critical tasks for several clinical procedures. Manual realisation of those is time-consuming the quality highly dependent on level expertise physician. To mitigate that laborious task, automatic tools have been developed where majority solutions supervised techniques. However, in medical domain, strong assumption having a well-representative ground truth far from being realistic. overcome this challenge, unsupervised techniques investigated....
Current 3D open-vocabulary scene understanding methods mostly utilize well-aligned 2D images as the bridge to learn features with language. However, applying these approaches becomes challenging in scenarios where are absent. In this work, we introduce a new pipeline, namely, OpenIns3D, which requires no image inputs, for at instance level. The OpenIns3D framework employs "Mask-Snap-Lookup" scheme. "Mask" module learns class-agnostic mask proposals point clouds. "Snap" generates synthetic...
This report provides an overview of the challenge hosted at OpenSUN3D Workshop on Open-Vocabulary 3D Scene Understanding held in conjunction with ICCV 2023. The goal this workshop series is to provide a platform for exploration and discussion open-vocabulary scene understanding tasks, including but not limited segmentation, detection mapping. We workshop, present dataset, evaluation methodology, brief descriptions winning methods. For additional details, please see...
Current point cloud segmentation architectures suffer from limited long-range feature modeling, as they mostly rely on aggregating information with local neighborhoods. Furthermore, in order to learn features at multiple scales, most methods utilize a data-agnostic sampling approach decrease the number of points after each stage. Such methods, however, often discard for small objects early stages, leading inadequate learning. We believe these issues are can be mitigated by introducing...