We address the task of supervised action segmentation which aims to partition a video into non-overlapping segments, each representing a different action. Recent works apply transformers to perform temporal modeling at the frame-level, which suffer from high computational cost and cannot well capture action dependencies over long temporal horizons. To address these issues, we propose an efficient BI-level Temporal modeling (BIT) framework that learns explicit action tokens to represent action segments, in parallel performs temporal modeling on frame and action levels, while maintaining a low computational cost. Our model contains (i) a frame branch that uses convolution to learn frame-level relationships, (ii) an action branch that uses transformer to learn action-level dependencies with a small set of action tokens and (iii) cross-attentions to allow communication between the two branches. We apply and extend a set-prediction objective to allow each action token to represent one or multiple action segments, thus can avoid learning a large number of tokens over long videos with many segments. Thanks to the design of our action branch, we can also seamlessly leverage textual transcripts of videos (when available) to help action segmentation by using them to initialize the action tokens. We evaluate our model on four video datasets (two egocentric and two third-person) for action segmentation with and without transcripts, showing that BIT significantly improves the state-of-the-art accuracy with much lower computational cost (30 times faster) compared to existing transformer-based methods.
Skeleton-based action segmentation requires recognizing composable actions in untrimmed videos. Current approaches decouple this problem by first extracting local visual features from skeleton sequences and then processing them by a temporal model to classify frame-wise actions. However, their performances remain limited as the visual features cannot sufficiently express composable actions. In this context, we propose Latent Action Composition (LAC), a novel self-supervised framework aiming at learning from synthesized composable motions for skeleton-based action segmentation. LAC is composed of a novel generation module towards synthesizing new sequences. Specifically, we design a linear latent space in the generator to represent primitive motion. New composed motions can be synthesized by simply performing arithmetic operations on latent representations of multiple input skeleton sequences. LAC leverages such synthesized sequences, which have large diversity and complexity, for learning visual representations of skeletons in both sequence and frame spaces via contrastive learning. The resulting visual encoder has a high expressive power and can be effectively transferred onto action segmentation tasks by end-to-end fine-tuning without the need for additional temporal models. We conduct a study focusing on transfer-learning and we show that representations learned from pre-trained LAC outperform the state-of-the-art by a large margin on TSU, Charades, PKU-MMD datasets.
While large-scale image-text pretrained models such as CLIP have been used for multiple video-level tasks on trimmed videos, their use for temporal localization in untrimmed videos is still a relatively unexplored task. We design a new approach for this called UnLoc, which uses pretrained image and text towers, and feeds tokens to a video-text fusion model. The output of the fusion module are then used to construct a feature pyramid in which each level connects to a head to predict a per-frame relevancy score and start/end time displacements. Unlike previous works, our architecture enables Moment Retrieval, Temporal Localization, and Action Segmentation with a single stage model, without the need for action proposals, motion based pretrained features or representation masking. Unlike specialized models, we achieve state of the art results on all three different localization tasks with a unified approach. Code will be available at: \url{https://github.com/google-research/scenic}.
Modeling long-term context in videos is crucial for many fine-grained tasks including temporal action segmentation. An interesting question that is still open is how much long-term temporal context is needed for optimal performance. While transformers can model the long-term context of a video, this becomes computationally prohibitive for long videos. Recent works on temporal action segmentation thus combine temporal convolutional networks with self-attentions that are computed only for a local temporal window. While these approaches show good results, their performance is limited by their inability to capture the full context of a video. In this work, we try to answer how much long-term temporal context is required for temporal action segmentation by introducing a transformer-based model that leverages sparse attention to capture the full context of a video. We compare our model with the current state of the art on three datasets for temporal action segmentation, namely 50Salads, Breakfast, and Assembly101. Our experiments show that modeling the full context of a video is necessary to obtain the best performance for temporal action segmentation.
In this technical report, we present our findings from the research conducted on the Human-Object Interaction 4D (HOI4D) dataset for egocentric action segmentation task. As a relatively novel research area, point cloud video methods might not be good at temporal modeling, especially for long point cloud videos (\eg, 150 frames). In contrast, traditional video understanding methods have been well developed. Their effectiveness on temporal modeling has been widely verified on many large scale video datasets. Therefore, we convert point cloud videos into depth videos and employ traditional video modeling methods to improve 4D action segmentation. By ensembling depth and point cloud video methods, the accuracy is significantly improved. The proposed method, named Mixture of Depth and Point cloud video experts (DPMix), achieved the first place in the 4D Action Segmentation Track of the HOI4D Challenge 2023.
Action segmentation plays an important role in enabling robots to automatically understand human activities. To train the action recognition model, while obtaining action labels for all frames is costly, annotating timestamp labels for weak supervision is cost-effective. However, existing methods may not fully utilize timestamp labels, which leads to insufficient performance. To alleviate this issue, we proposed a novel learning pattern in our training stage, which maximizes the probability of action union of surrounding timestamps for unlabeled frames. In our inference stage, we provided a new refinement solution to generate better hard-assigned action classes from soft-assigned predictions. Importantly, our methods are model-agnostic and can be applied to existing frameworks. On three commonly used action-segmentation data, our method outperforms previous timestamp-supervision methods and achieves new state-of-the-art performance. Moreover , our method uses less than 1% of fully-supervised labels to obtain comparable or even better results.
Understanding comprehensive assembly knowledge from videos is critical for futuristic ultra-intelligent industry. To enable technological breakthrough, we present HA-ViD - the first human assembly video dataset that features representative industrial assembly scenarios, natural procedural knowledge acquisition process, and consistent human-robot shared annotations. Specifically, HA-ViD captures diverse collaboration patterns of real-world assembly, natural human behaviors and learning progression during assembly, and granulate action annotations to subject, action verb, manipulated object, target object, and tool. We provide 3222 multi-view, multi-modality videos (each video contains one assembly task), 1.5M frames, 96K temporal labels and 2M spatial labels. We benchmark four foundational video understanding tasks: action recognition, action segmentation, object detection and multi-object tracking. Importantly, we analyze their performance for comprehending knowledge in assembly progress, process efficiency, task collaboration, skill parameters and human intention. Details of HA-ViD is available at: https://iai-hrc.github.io/ha-vid.
Automatic surgical phase recognition is one of the key technologies to support Video-Based Assessment (VBA) systems for surgical education. Utilizing temporal information is crucial for surgical phase recognition, hence various recent approaches extract frame-level features to conduct full video temporal modeling. For better temporal modeling, we propose SlowFast Temporal Modeling Network (SF-TMN) for surgical phase recognition that can not only achieve frame-level full video temporal modeling but also achieve segment-level full video temporal modeling. We employ a feature extraction network, pre-trained on the target dataset, to extract features from video frames as the training data for SF-TMN. The Slow Path in SF-TMN utilizes all frame features for frame temporal modeling. The Fast Path in SF-TMN utilizes segment-level features summarized from frame features for segment temporal modeling. The proposed paradigm is flexible regarding the choice of temporal modeling networks. We explore MS-TCN and ASFormer models as temporal modeling networks and experiment with multiple combination strategies for Slow and Fast Paths. We evaluate SF-TMN on Cholec80 surgical phase recognition task and demonstrate that SF-TMN can achieve state-of-the-art results on all considered metrics. SF-TMN with ASFormer backbone outperforms the state-of-the-art Not End-to-End(TCN) method by 2.6% in accuracy and 7.4% in the Jaccard score. We also evaluate SF-TMN on action segmentation datasets including 50salads, GTEA, and Breakfast, and achieve state-of-the-art results. The improvement in the results shows that combining temporal information from both frame level and segment level by refining outputs with temporal refinement stages is beneficial for the temporal modeling of surgical phases.
This paper presents an unsupervised transformer-based framework for temporal activity segmentation which leverages not only frame-level cues but also segment-level cues. This is in contrast with previous methods which often rely on frame-level information only. Our approach begins with a frame-level prediction module which estimates framewise action classes via a transformer encoder. The frame-level prediction module is trained in an unsupervised manner via temporal optimal transport. To exploit segment-level information, we utilize a segment-level prediction module and a frame-to-segment alignment module. The former includes a transformer decoder for estimating video transcripts, while the latter matches frame-level features with segment-level features, yielding permutation-aware segmentation results. Moreover, inspired by temporal optimal transport, we introduce simple-yet-effective pseudo labels for unsupervised training of the above modules. Our experiments on four public datasets, i.e., 50 Salads, YouTube Instructions, Breakfast, and Desktop Assembly show that our approach achieves comparable or better performance than previous methods in unsupervised activity segmentation.
Egocentric temporal action segmentation in videos is a crucial task in computer vision with applications in various fields such as mixed reality, human behavior analysis, and robotics. Although recent research has utilized advanced visual-language frameworks, transformers remain the backbone of action segmentation models. Therefore, it is necessary to improve transformers to enhance the robustness of action segmentation models. In this work, we propose two novel ideas to enhance the state-of-the-art transformer for action segmentation. First, we introduce a dual dilated attention mechanism to adaptively capture hierarchical representations in both local-to-global and global-to-local contexts. Second, we incorporate cross-connections between the encoder and decoder blocks to prevent the loss of local context by the decoder. We also utilize state-of-the-art visual-language representation learning techniques to extract richer and more compact features for our transformer. Our proposed approach outperforms other state-of-the-art methods on the Georgia Tech Egocentric Activities (GTEA) and HOI4D Office Tools datasets, and we validate our introduced components with ablation studies. The source code and supplementary materials are publicly available on https://www.sail-nu.com/dxformer.
In our pursuit of advancing multi-modal AI assistants capable of guiding users to achieve complex multi-step goals, we propose the task of "Visual Planning for Assistance (VPA)". Given a succinct natural language goal, e.g., "make a shelf", and a video of the user's progress so far, the aim of VPA is to devise a plan, i.e., a sequence of actions such as "sand shelf", "paint shelf", etc. to realize the specified goal. This requires assessing the user's progress from the (untrimmed) video, and relating it to the requirements of natural language goal, i.e., which actions to select and in what order? Consequently, this requires handling long video history and arbitrarily complex action dependencies. To address these challenges, we decompose VPA into video action segmentation and forecasting. Importantly, we experiment by formulating the forecasting step as a multi-modal sequence modeling problem, allowing us to leverage the strength of pre-trained LMs (as the sequence model). This novel approach, which we call Visual Language Model based Planner (VLaMP), outperforms baselines across a suite of metrics that gauge the quality of the generated plans. Furthermore, through comprehensive ablations, we also isolate the value of each component--language pre-training, visual observations, and goal information. We have open-sourced all the data, model checkpoints, and training code.
In this paper, we propose a novel fully unsupervised framework that learns action representations suitable for the action segmentation task from the single input video itself, without requiring any training data. Our method is a deep metric learning approach rooted in a shallow network with a triplet loss operating on similarity distributions and a novel triplet selection strategy that effectively models temporal and semantic priors to discover actions in the new representational space. Under these circumstances, we successfully recover temporal boundaries in the learned action representations with higher quality compared with existing unsupervised approaches. The proposed method is evaluated on two widely used benchmark datasets for the action segmentation task and it achieves competitive performance by applying a generic clustering algorithm on the learned representations.
In this paper, we present an end-to-end trainable unified multiscale encoder-decoder transformer that is focused on dense prediction tasks in video. The presented Multiscale Encoder-Decoder Video Transformer (MED-VT) uses multiscale representation throughout and employs an optional input beyond video (e.g., audio), when available, for multimodal processing (MED-VT++). Multiscale representation at both encoder and decoder yields three key benefits: (i) implicit extraction of spatiotemporal features at different levels of abstraction for capturing dynamics without reliance on input optical flow, (ii) temporal consistency at encoding and (iii) coarse-to-fine detection for high-level (e.g., object) semantics to guide precise localization at decoding. Moreover, we present a transductive learning scheme through many-to-many label propagation to provide temporally consistent video predictions. We showcase MED-VT/MED-VT++ on three unimodal video segmentation tasks (Automatic Video Object Segmentation (AVOS), actor-action segmentation and Video Semantic Segmentation (VSS)) as well as a multimodal segmentation task (Audio-Visual Segmentation (AVS)). Results show that the proposed architecture outperforms alternative state-of-the-art approaches on multiple benchmarks using only video (and optional audio) as input, without reliance on optical flow. Finally, to document details of the model's internal learned representations, we present a detailed interpretability study, encompassing both quantitative and qualitative analyses.
In this paper we introduce a rule-based, compositional, and hierarchical modeling of action using Therbligs as our atoms. Introducing these atoms provides us with a consistent, expressive, contact-centered representation of action. Over the atoms we introduce a differentiable method of rule-based reasoning to regularize for logical consistency. Our approach is complementary to other approaches in that the Therblig-based representations produced by our architecture augment rather than replace existing architectures' representations. We release the first Therblig-centered annotations over two popular video datasets - EPIC Kitchens 100 and 50-Salads. We also broadly demonstrate benefits to adopting Therblig representations through evaluation on the following tasks: action segmentation, action anticipation, and action recognition - observing an average 10.5\%/7.53\%/6.5\% relative improvement, respectively, over EPIC Kitchens and an average 8.9\%/6.63\%/4.8\% relative improvement, respectively, over 50 Salads. Code and data will be made publicly available.
Fully supervised action segmentation works on frame-wise action recognition with dense annotations and often suffers from the over-segmentation issue. Existing works have proposed a variety of solutions such as boundary-aware networks, multi-stage refinement, and temporal smoothness losses. However, most of them take advantage of frame-wise supervision, which cannot effectively tackle the evaluation metrics with different granularities. In this paper, for the desirable large receptive field, we first develop a novel local-global attention mechanism with temporal pyramid dilation and temporal pyramid pooling for efficient multi-scale attention. Then we decouple two inherent goals in action segmentation, ie, (1) individual identification solved by frame-wise supervision, and (2) temporal reasoning tackled by action set prediction. Afterward, an action alignment module fuses these different granularity predictions, leading to more accurate and smoother action segmentation. We achieve state-of-the-art accuracy, eg, 82.8% (+2.6%) on GTEA and 74.7% (+1.2%) on Breakfast, which demonstrates the effectiveness of our proposed method, accompanied by extensive ablation studies. The code will be made available later.
Temporal action segmentation is crucial for understanding long-form videos. Previous works on this task commonly adopt an iterative refinement paradigm by using multi-stage models. We propose a novel framework via denoising diffusion models, which nonetheless shares the same inherent spirit of such iterative refinement. In this framework, action predictions are iteratively generated from random noise with input video features as conditions. To enhance the modeling of three striking characteristics of human actions, including the position prior, the boundary ambiguity, and the relational dependency, we devise a unified masking strategy for the conditioning inputs in our framework. Extensive experiments on three benchmark datasets, i.e., GTEA, 50Salads, and Breakfast, are performed and the proposed method achieves superior or comparable results to state-of-the-art methods, showing the effectiveness of a generative approach for action segmentation.
Action segmentation is a challenging task in high-level process analysis, typically performed on video or kinematic data obtained from various sensors. This work presents two contributions related to action segmentation on kinematic data. Firstly, we introduce two versions of Multi-Stage Temporal Convolutional Recurrent Networks (MS-TCRNet), specifically designed for kinematic data. The architectures consist of a prediction generator with intra-stage regularization and Bidirectional LSTM or GRU-based refinement stages. Secondly, we propose two new data augmentation techniques, World Frame Rotation and Hand Inversion, which utilize the strong geometric structure of kinematic data to improve algorithm performance and robustness. We evaluate our models on three datasets of surgical suturing tasks: the Variable Tissue Simulation (VTS) Dataset and the newly introduced Bowel Repair Simulation (BRS) Dataset, both of which are open surgery simulation datasets collected by us, as well as the JHU-ISI Gesture and Skill Assessment Working Set (JIGSAWS), a well-known benchmark in robotic surgery. Our methods achieved state-of-the-art performance.
Temporal action segmentation in untrimmed videos has gained increased attention recently. However, annotating action classes and frame-wise boundaries is extremely time consuming and cost intensive, especially on large-scale datasets. To address this issue, we propose an unsupervised approach for learning action classes from untrimmed video sequences. In particular, we propose a temporal embedding network that combines relative time prediction, feature reconstruction, and sequence-to-sequence learning, to preserve the spatial layout and sequential nature of the video features. A two-step clustering pipeline on these embedded feature representations then allows us to enforce temporal consistency within, as well as across videos. Based on the identified clusters, we decode the video into coherent temporal segments that correspond to semantically meaningful action classes. Our evaluation on three challenging datasets shows the impact of each component and, furthermore, demonstrates our state-of-the-art unsupervised action segmentation results.
Recognizing human actions from untrimmed videos is an important task in activity understanding, and poses unique challenges in modeling long-range temporal relations. Recent works adopt a predict-and-refine strategy which converts an initial prediction to action segments for global context modeling. However, the generated segment representations are often noisy and exhibit inaccurate segment boundaries, over-segmentation and other problems. To deal with these issues, we propose an attention based approach which we call \textit{temporal segment transformer}, for joint segment relation modeling and denoising. The main idea is to denoise segment representations using attention between segment and frame representations, and also use inter-segment attention to capture temporal correlations between segments. The refined segment representations are used to predict action labels and adjust segment boundaries, and a final action segmentation is produced based on voting from segment masks. We show that this novel architecture achieves state-of-the-art accuracy on the popular 50Salads, GTEA and Breakfast benchmarks. We also conduct extensive ablations to demonstrate the effectiveness of different components of our design.
Fine-grained temporal action segmentation in long, untrimmed RGB videos is a key topic in visual human- machine interaction. Recent temporal convolution based approaches either use encoder-decoder(ED) architecture or dilations with doubling factor in consecutive convolution layers to segment actions in videos. However ED networks operate on low temporal resolution and the dilations in suc- cessive layers cause gridding artifacts problem. We propose depthwise separable temporal convolution network (DS- TCN) that operates on full temporal resolution and with re- duced gridding effects. The basic component of DS-TCN is residual depthwise dilated block (RDDB). We explore the trade-off between large kernels and small dilation rates us- ing RDDB. We show that our DS-TCN is capable of captur- ing long-term dependencies as well as local temporal cues efficiently. Our evaluation on three benchmark datasets, GTEA, 50Salads, and Breakfast demonstrates that DS-TCN outperforms the existing ED-TCN and dilation based TCN baselines even with comparatively fewer parameters.