Suha Kwak
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Suha Kwak has a total of 69 co-authors. Here are the top co-authors by number of shared publications:
- Minsu Cho
Pohang University of Science and Technology (POSTECH)
11 shared publications - Sungyeon Kim
POSTECH
8 shared publications - Boseung Jeong
Pohang University of Science and Technology (POSTECH) South Korea
5 shared publications - Sehyun Hwang
Pohang University of Science and Technology (POSTECH) South Korea
5 shared publications Dongwon Kim
4 shared publications- Namyup Kim
Microsoft Research Asia
4 shared publications
Latest Publications
- 10 September 2026
- Published by Springer Nature
- p. 416-434
RePL: Pseudo-label Refinement for Semi-supervised LiDAR Semantic Segmentation
- 10 September 2026
- Published by Springer Nature
- p. 416-434
RePL: Pseudo-label Refinement for Semi-supervised LiDAR Semantic Segmentation
- 10 September 2026
- Published by Springer Nature
- p. 653-672
Learning Sample-Wise Rank-Aware Interpolation Weights for Composed Visual Data Retrieval
- 10 September 2026
- Published by Springer Nature
- p. 653-672
Learning Sample-Wise Rank-Aware Interpolation Weights for Composed Visual Data Retrieval
- 6 December 2024
- Published by Springer Nature
- p. 74-91
Online Temporal Action Localization with Memory-Augmented Transformer
- 6 December 2024
- Published by Springer Nature
- p. 74-91
Online Temporal Action Localization with Memory-Augmented Transformer
- Open Access
ActFusion: a Unified Diffusion Model for Action Segmentation and Anticipation
Temporal action segmentation and long-term action anticipation are two popular vision tasks for the temporal analysis of actions in videos. Despite apparent relevance and potential complementarity, these two problems have been investigated as separate and distinct tasks. In this work, we tackle these two problems, action segmentation and action anticipation, jointly using a unified diffusion model dubbed ActFusion. The key idea to unification is to train the model to effectively handle both visible and invisible parts of the sequence in an integrated manner; the visible part is for temporal segmentation, and the invisible part is for future anticipation. To this end, we introduce a new anticipative masking strategy during training in which a late part of the video frames is masked as invisible, and learnable tokens replace these frames to learn to predict the invisible future. Experimental results demonstrate the bi-directional benefits between action segmentation and anticipation. ActFusion achieves the state-of-the-art performance across the standard benchmarks of 50 Salads, Breakfast, and GTEA, outperforming task-specific models in both of the two tasks with a single unified model through joint learning.
Open AccessActFusion: a Unified Diffusion Model for Action Segmentation and Anticipation
Temporal action segmentation and long-term action anticipation are two popular vision tasks for the temporal analysis of actions in videos. Despite apparent relevance and potential complementarity, these two problems have been investigated as separate and distinct tasks. In this work, we tackle these two problems, action segmentation and action anticipation, jointly using a unified diffusion model dubbed ActFusion. The key idea to unification is to train the model to effectively handle both visible and invisible parts of the sequence in an integrated manner; the visible part is for temporal segmentation, and the invisible part is for future anticipation. To this end, we introduce a new anticipative masking strategy during training in which a late part of the video frames is masked as invisible, and learnable tokens replace these frames to learn to predict the invisible future. Experimental results demonstrate the bi-directional benefits between action segmentation and anticipation. ActFusion achieves the state-of-the-art performance across the standard benchmarks of 50 Salads, Breakfast, and GTEA, outperforming task-specific models in both of the two tasks with a single unified model through joint learning.
- 27 November 2024
- Published by Springer Nature
- p. 450-467
Classification Matters: Improving Video Action Detection with Class-Specific Attention
- 27 November 2024
- Published by Springer Nature
- p. 450-467
Classification Matters: Improving Video Action Detection with Class-Specific Attention
- 23 November 2024
- Published by Springer Nature
- p. 240-258
Towards More Practical Group Activity Detection: A New Benchmark and Model
- 23 November 2024
- Published by Springer Nature
- p. 240-258
Towards More Practical Group Activity Detection: A New Benchmark and Model
- 22 November 2024
- Published by Springer Nature
- p. 440-458
Efficient and Versatile Robust Fine-Tuning of Zero-Shot Models
- 22 November 2024
- Published by Springer Nature
- p. 440-458
Efficient and Versatile Robust Fine-Tuning of Zero-Shot Models
- 3 November 2024
- Published by Springer Nature
- p. 1-18
FREST: Feature RESToration for Semantic Segmentation Under Multiple Adverse Conditions
- 3 November 2024
- Published by Springer Nature
- p. 1-18
FREST: Feature RESToration for Semantic Segmentation Under Multiple Adverse Conditions
- 31 October 2024
- Published by Springer Nature
- p. 467-483
MemBN: Robust Test-Time Adaptation via Batch Norm with Statistics Memory
- 31 October 2024
- Published by Springer Nature
- p. 467-483
MemBN: Robust Test-Time Adaptation via Batch Norm with Statistics Memory
- 26 October 2024
- Published by Springer Nature
- p. 474-490
PLOT: Text-Based Person Search with Part Slot Attention for Corresponding Part Discovery
- 26 October 2024
- Published by Springer Nature
- p. 474-490
PLOT: Text-Based Person Search with Part Slot Attention for Corresponding Part Discovery