ActFusion: a Unified Diffusion Model for Action Segmentation and Anticipation
Preprint
Abstract
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.Keywords
All Related Versions
This publication has 12 references indexed in Scilit:
- Temporal Action Segmentation: An Analysis of Modern TechniquesIEEE Transactions on Pattern Analysis and Machine Intelligence, 2023
- Diffusion Models as Masked Autoencoders2023
- Dilated Transformer with Feature Aggregation Module for Action SegmentationNeural Processing Letters, 2022
- Stacked encoder–decoder transformer with boundary smoothing for action segmentationElectronics Letters, 2022
- Multistage temporal convolution transformer for action segmentationImage and Vision Computing, 2022
- Maximization and restoration: Action segmentation through dilation passing and temporal reconstructionPattern Recognition, 2022
- BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding2019
- Self-Attention with Relative Position Representations2018
- Weakly supervised learning of actions from transcriptsComputer Vision and Image Understanding, 2017
- Learning to recognize objects in egocentric activities2011