Rolling-Unrolling LSTMs for Action Anticipation from First-Person Video

Abstract

In this paper, we tackle the problem of egocentric action anticipation, i.e., predicting what actions the camera wearer will perform in the near future and which objects they will interact with. Specifically, we contribute Rolling-Unrolling LSTM, a learning architecture to anticipate actions from egocentric videos. The method is based on three components: 1) an architecture comprised of two LSTMs to model the sub-tasks of summarizing the past and inferring the future, 2) a Sequence Completion Pre-Training technique which encourages the LSTMs to focus on the different sub-tasks, and 3) a Modality ATTention (MATT) mechanism to efficiently fuse multi-modal predictions performed by processing RGB frames, optical flow fields and object-based features. The proposed approach is validated on EPIC-Kitchens, EGTEA Gaze+ and ActivityNet. The experiments show that the proposed architecture is state-of-the-art in the domain of egocentric videos, achieving top performances in the 2019 EPIC-Kitchens egocentric action anticipation challenge. The approach also achieves competitive performance on ActivityNet with respect to methods not based on unsupervised pre-training and generalizes to the tasks of early action recognition and action recognition. To encourage research on this challenging topic, we made our code, trained models, and pre-extracted features available at our web page: http://iplab.dmi.unict.it/rulstm.

arXiv commentarXiv admin note: substantial text overlap with arXiv:1905.09035

References in corpus (14)

• Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
• PyTorch: An Imperative Style, High-Performance Deep Learning Library
• Neural Machine Translation by Jointly Learning to Align and Translate
• Sequence to Sequence Learning with Neural Networks
• Show, Attend and Tell: Neural Image Caption Generation with Visual Attention
• Two-Stream Convolutional Networks for Action Recognition in Videos
• An Empirical Evaluation of Generic Convolutional and Recurrent Networks for Sequence Modeling
• Spatiotemporal Residual Networks for Video Action Recognition
• Scaling Egocentric Vision: The EPIC-KITCHENS Dataset
• TSM: Temporal Shift Module for Efficient Video Understanding
• Next-Active-Object prediction from Egocentric Videos
• Choosing Smartly: Adaptive Multimodal Fusion for Object Detection in Changing Environments
• TrafficPredict: Trajectory Prediction for Heterogeneous Traffic-Agents
• Attention is All We Need: Nailing Down Object-centric Attention for Egocentric Activity Recognition

Cited by in corpus (15)

• Visual Object Tracking in First Person Vision
• On the Pitfalls of Batch Normalization for End-to-End Video Learning: A Study on Surgical Workflow Analysis
• POV: Prompt-Oriented View-Agnostic Learning for Egocentric Hand-Object Interaction in the Multi-View World
• Semantically Guided Representation Learning For Action Anticipation
• Cross-Domain First Person Audio-Visual Action Recognition through Relative Norm Alignment
• Streaming egocentric action anticipation: An evaluation scheme and approach
• Ego-Exo: Transferring Visual Representations from Third-person to First-person Videos
• Bringing Online Egocentric Action Recognition into the wild
• PEAR: Phrase-Based Hand-Object Interaction Anticipation
• Higher Order Recurrent Space-Time Transformer for Video Action Prediction
• Enhancing Action Recognition by Leveraging the Hierarchical Structure of Actions and Textual Context
• Cognition Transferring and Decoupling for Text-supervised Egocentric Semantic Segmentation
• Hier-EgoPack: Hierarchical Egocentric Video Understanding with Diverse Task Perspectives
• Is First Person Vision Challenging for Object Tracking?
• SlowFast Rolling-Unrolling LSTMs for Action Anticipation in Egocentric Videos