Fadime Sener

Guodong Ding, Fadime Sener, Angela Yao

Temporal action segmentation (TAS) in videos aims at densely identifying video frames in minutes-long videos with multiple action classes. As a long-range video understanding task, researchers have developed an extended collection of methods and examined their performance using various benchmarks. Despite the rapid growth of TAS techniques in recent years, no systematic survey has been conducted in these sectors. This survey analyzes and summarizes the most significant contributions and trends. In particular, we first examine the task definition, common benchmarks, types of supervision, and prevalent evaluation measures. In addition, we systematically investigate two essential techniques of this topic, i.e., frame representation and temporal modeling, which have been studied extensively in the literature. We then conduct a thorough review of existing TAS works categorized by their levels of supervision and conclude our survey by identifying and emphasizing several research gaps. In addition, we have curated a list of TAS resources, which is available at https://github.com/nus-cvml/awesome-temporal-action-segmentation.

Fadime Sener, Dibyadip Chatterjee, Daniel Shelepov et al.

Assembly101 is a new procedural activity dataset featuring 4321 videos of people assembling and disassembling 101 "take-apart" toy vehicles. Participants work without fixed instructions, and the sequences feature rich and natural variations in action ordering, mistakes, and corrections. Assembly101 is the first multi-view action dataset, with simultaneous static (8) and egocentric (4) recordings. Sequences are annotated with more than 100K coarse and 1M fine-grained action segments, and 18M 3D hand poses. We benchmark on three action understanding tasks: recognition, anticipation and temporal segmentation. Additionally, we propose a novel task of detecting mistakes. The unique recording format and rich set of annotations allow us to investigate generalization to new toys, cross-view transfer, long-tailed distributions, and pose vs. appearance. We envision that Assembly101 will serve as a new challenge to investigate various activity understanding problems.

Takehiko Ohkawa, Kun He, Fadime Sener et al.

We present AssemblyHands, a large-scale benchmark dataset with accurate 3D hand pose annotations, to facilitate the study of egocentric activities with challenging hand-object interactions. The dataset includes synchronized egocentric and exocentric images sampled from the recent Assembly101 dataset, in which participants assemble and disassemble take-apart toys. To obtain high-quality 3D hand pose annotations for the egocentric images, we develop an efficient pipeline, where we use an initial set of manual annotations to train a model to automatically annotate a much larger dataset. Our annotation model uses multi-view feature fusion and an iterative refinement scheme, and achieves an average keypoint error of 4.20 mm, which is 85% lower than the error of the original annotations in Assembly101. AssemblyHands provides 3.0M annotated images, including 490K egocentric images, making it the largest existing benchmark dataset for egocentric 3D hand pose estimation. Using this data, we develop a strong single-view baseline of 3D hand pose estimation from egocentric images. Furthermore, we design a novel action classification task to evaluate predicted 3D hand poses. Our study shows that having higher-quality hand poses directly improves the ability to recognize actions.

Dibyadip Chatterjee, Fadime Sener, Shugao Ma et al.

Human actions in egocentric videos are often hand-object interactions composed from a verb (performed by the hand) applied to an object. Despite their extensive scaling up, egocentric datasets still face two limitations - sparsity of action compositions and a closed set of interacting objects. This paper proposes a novel open vocabulary action recognition task. Given a set of verbs and objects observed during training, the goal is to generalize the verbs to an open vocabulary of actions with seen and novel objects. To this end, we decouple the verb and object predictions via an object-agnostic verb encoder and a prompt-based object encoder. The prompting leverages CLIP representations to predict an open vocabulary of interacting objects. We create open vocabulary benchmarks on the EPIC-KITCHENS-100 and Assembly101 datasets; whereas closed-action methods fail to generalize, our proposed method is effective. In addition, our object encoder significantly outperforms existing open-vocabulary visual recognition methods in recognizing novel interacting objects.

Guodong Ding, Fadime Sener, Shugao Ma et al.

One promising use case of AI assistants is to help with complex procedures like cooking, home repair, and assembly tasks. Can we teach the assistant to interject after the user makes a mistake? This paper targets the problem of identifying ordering mistakes in assembly procedures. We propose a system that can detect ordering mistakes by utilizing a learned knowledge base. Our framework constructs a knowledge base with spatial and temporal beliefs based on observed mistakes. Spatial beliefs depict the topological relationship of the assembling components, while temporal beliefs aggregate prerequisite actions as ordering constraints. With an episodic memory design, our algorithm can dynamically update and construct the belief sets as more actions are observed, all in an online fashion. We demonstrate experimentally that our inferred spatial and temporal beliefs are capable of identifying incorrect orderings in real-world action sequences. To construct the spatial beliefs, we collect a new set of coarse-level action annotations for Assembly101 based on the positioning of the toy parts. Finally, we demonstrate the superior performance of our belief inference algorithm in detecting ordering mistakes on the Assembly101 dataset.

Zhanzhong Pang, Fadime Sener, Shrinivas Ramasubramanian et al.

Procedural activity videos often exhibit a long-tailed action distribution due to varying action frequencies and durations. However, state-of-the-art temporal action segmentation methods overlook the long tail and fail to recognize tail actions. Existing long-tail methods make class-independent assumptions and struggle to identify tail classes when applied to temporal segmentation frameworks. This work proposes a novel group-wise temporal logit adjustment~(G-TLA) framework that combines a group-wise softmax formulation while leveraging activity information and action ordering for logit adjustment. The proposed framework significantly improves in segmenting tail actions without any performance loss on head actions.

Zhanzhong Pang, Dibyadip Chatterjee, Fadime Sener et al.

Multimodal Large Language Models (MLLMs) have advanced open-world action understanding and can be adapted as generative classifiers for closed-set settings by autoregressively generating action labels as text. However, this approach is inefficient, and shared subwords across action labels introduce semantic overlap, leading to ambiguity in generation. In contrast, discriminative classifiers learn task-specific representations with clear decision boundaries, enabling efficient one-step classification without autoregressive decoding. We first compare generative and discriminative classifiers with MLLMs for closed-set action understanding, revealing the superior accuracy and efficiency of the latter. To bridge the performance gap, we design strategies that elevate generative classifiers toward performance comparable with discriminative ones. Furthermore, we show that generative modeling can complement discriminative classifiers, leading to better performance while preserving efficiency. To this end, we propose Generation-Assisted Discriminative~(GAD) classifier for closed-set action understanding. GAD operates only during fine-tuning, preserving full compatibility with MLLM pretraining. Extensive experiments on temporal action understanding benchmarks demonstrate that GAD improves both accuracy and efficiency over generative methods, achieving state-of-the-art results on four tasks across five datasets, including an average 2.5% accuracy gain and 3x faster inference on our largest COIN benchmark.

Zicong Fan, Edoardo Remelli, David Dimond et al.

The ability to grasp objects, signal with gestures, and share emotion through touch all stem from the unique capabilities of human hands. Yet creating high-quality personalized hand avatars from images remains challenging due to complex geometry, appearance, and articulation, particularly under unconstrained lighting and limited views. Progress has also been limited by the lack of datasets that jointly provide accurate 3D geometry, high-resolution multiview imagery, and a diverse population of subjects. To address this, we present PALM, a large-scale dataset comprising 13k high-quality hand scans from 263 subjects and 90k multi-view images, capturing rich variation in skin tone, age, and geometry. To show its utility, we present a baseline PALM-Net, a multi-subject prior over hand geometry and material properties learned via physically based inverse rendering, enabling realistic, relightable single-image hand avatar personalization. PALM's scale and diversity make it a valuable real-world resource for hand modeling and related research.

Cheeun Hong, German Barquero, Fadime Sener et al.

Instructional video generation is an emerging task that aims to synthesize coherent demonstrations of procedural activities from textual descriptions. Such capability has broad implications for content creation, education, and human-AI interaction, yet existing video diffusion models struggle to maintain temporal consistency and controllability across long sequences of multiple action steps. We introduce a pipeline for future-driven streaming instructional video generation, dubbed SneakPeek, a diffusion-based autoregressive framework designed to generate precise, stepwise instructional videos conditioned on an initial image and structured textual prompts. Our approach introduces three key innovations to enhance consistency and controllability: (1) predictive causal adaptation, where a causal model learns to perform next-frame prediction and anticipate future keyframes; (2) future-guided self-forcing with a dual-region KV caching scheme to address the exposure bias issue at inference time; (3) multi-prompt conditioning, which provides fine-grained and procedural control over multi-step instructions. Together, these components mitigate temporal drift, preserve motion consistency, and enable interactive video generation where future prompt updates dynamically influence ongoing streaming video generation. Experimental results demonstrate that our method produces temporally coherent and semantically faithful instructional videos that accurately follow complex, multi-step task descriptions.

Anna Kukleva, Fadime Sener, Edoardo Remelli et al.

Lately, there has been growing interest in adapting vision-language models (VLMs) to image and third-person video classification due to their success in zero-shot recognition. However, the adaptation of these models to egocentric videos has been largely unexplored. To address this gap, we propose a simple yet effective cross-modal adaptation framework, which we call X-MIC. Using a video adapter, our pipeline learns to align frozen text embeddings to each egocentric video directly in the shared embedding space. Our novel adapter architecture retains and improves generalization of the pre-trained VLMs by disentangling learnable temporal modeling and frozen visual encoder. This results in an enhanced alignment of text embeddings to each egocentric video, leading to a significant improvement in cross-dataset generalization. We evaluate our approach on the Epic-Kitchens, Ego4D, and EGTEA datasets for fine-grained cross-dataset action generalization, demonstrating the effectiveness of our method. Code is available at https://github.com/annusha/xmic

Sammy Christen, Shreyas Hampali, Fadime Sener et al.

Generating natural hand-object interactions in 3D is challenging as the resulting hand and object motions are expected to be physically plausible and semantically meaningful. Furthermore, generalization to unseen objects is hindered by the limited scale of available hand-object interaction datasets. In this paper, we propose a novel method, dubbed DiffH2O, which can synthesize realistic, one or two-handed object interactions from provided text prompts and geometry of the object. The method introduces three techniques that enable effective learning from limited data. First, we decompose the task into a grasping stage and an text-based manipulation stage and use separate diffusion models for each. In the grasping stage, the model only generates hand motions, whereas in the manipulation phase both hand and object poses are synthesized. Second, we propose a compact representation that tightly couples hand and object poses and helps in generating realistic hand-object interactions. Third, we propose two different guidance schemes to allow more control of the generated motions: grasp guidance and detailed textual guidance. Grasp guidance takes a single target grasping pose and guides the diffusion model to reach this grasp at the end of the grasping stage, which provides control over the grasping pose. Given a grasping motion from this stage, multiple different actions can be prompted in the manipulation phase. For the textual guidance, we contribute comprehensive text descriptions to the GRAB dataset and show that they enable our method to have more fine-grained control over hand-object interactions. Our quantitative and qualitative evaluation demonstrates that the proposed method outperforms baseline methods and leads to natural hand-object motions.

Zhanzhong Pang, Dibyadip Chatterjee, Fadime Sener et al.

Streaming video understanding requires processing unbounded video streams with limited memory and computation, posing two key challenges. First, continuously constructing new and evicting old key-value(KV) caches is required for unbounded streams. Secondly, due to the high cost of collecting and training on unbounded streams, models must learn from short sequences while generalizing to long streams. Existing streaming VideoVLLMs fail to scale to unbounded video streams or focus on cache reuse strategies, leaving the impact of cache construction underexplored. In this paper, we propose Decoupled Streaming Cache(DSCache), a training-free cache construction mechanism that adapts pretrained offline models to streaming settings. DSCache maintains a cumulative past KV cache while constructing a separate instant cache on-demand, decoupled from past caches to preserve the informativeness of recent inputs. To enable position extrapolation beyond the training length, DSCache further incorporates a position-agnostic encoding strategy, ensuring KV caches to support unseen positions and preventing position overflow. Experiments on Streaming Video QA benchmarks demonstrate DSCache's state-of-the-art performance, with an average 2.5% accuracy gains over prior methods.

Dibyadip Chatterjee, Zhanzhong Pang, Fadime Sener et al.

Streaming video models should respond the moment an event unfolds, not after the moment has passed. Yet existing online VideoQA benchmarks remain largely retrospective. They pause the video at fixed timestamps, pose questions about current or past events, and score models only at those moments. This protocol leaves streaming predictions untested. To close this gap, we introduce SPOT-Bench, featuring multi-turn proactive queries that evaluate general streaming perception and assistive capabilities required by an always-on, real-time assistant. SPOT-Bench comes with Timeliness-F1, a consolidated metric that measures streaming predictions by their temporal precision and balanced coverage across the entire video. Our benchmark reveals: (i) offline models detect events reliably but spam predictions unprompted; (ii) post-training for silence reduces spamming but induces unresponsiveness; (iii) half of the streaming video expects no response, which we term dead-time - compute spent here does not affect response latency. These findings motivate AsynKV, a training-free streaming adaptation of offline models, that retains their event perception while improving their streaming behavior. AsynKV features a long-short term memory, utilized efficiently by scaling compute during dead-time. It serves as a strong baseline on SPOT-Bench, outperforming existing streaming models, and achieves state-of-the-art on retrospective benchmarks.

Zhanzhong Pang, Fadime Sener, Angela Yao

Online Action Detection (OAD) detects actions in streaming videos using past observations. State-of-the-art OAD approaches model past observations and their interactions with an anticipated future. The past is encoded using short- and long-term memories to capture immediate and long-range dependencies, while anticipation compensates for missing future context. We identify a training-inference discrepancy in existing OAD methods that hinders learning effectiveness. The training uses varying lengths of short-term memory, while inference relies on a full-length short-term memory. As a remedy, we propose a Context-enhanced Memory-Refined Transformer (CMeRT). CMeRT introduces a context-enhanced encoder to improve frame representations using additional near-past context. It also features a memory-refined decoder to leverage near-future generation to enhance performance. CMeRT achieves state-of-the-art in online detection and anticipation on THUMOS'14, CrossTask, and EPIC-Kitchens-100.

Dibyadip Chatterjee, Edoardo Remelli, Yale Song et al.

We introduce ProVideLLM, an end-to-end framework for real-time procedural video understanding. ProVideLLM integrates a multimodal cache configured to store two types of tokens - verbalized text tokens, which provide compressed textual summaries of long-term observations, and visual tokens, encoded with DETR-QFormer to capture fine-grained details from short-term observations. This design reduces token count by 22x over existing methods in representing one hour of long-term observations while effectively encoding fine-granularity of the present. By interleaving these tokens in our multimodal cache, ProVideLLM ensures sub-linear scaling of memory and compute with video length, enabling per-frame streaming inference at 10 FPS and streaming dialogue at 25 FPS, with a minimal 2GB GPU memory footprint. ProVideLLM also sets new state-of-the-art results on six procedural tasks across four datasets.

Zhanzhong Pang, Fadime Sener, Shrinivas Ramasubramanian et al.

Temporal action segmentation in untrimmed procedural videos aims to densely label frames into action classes. These videos inherently exhibit long-tailed distributions, where actions vary widely in frequency and duration. In temporal action segmentation approaches, we identified a bi-level learning bias. This bias encompasses (1) a class-level bias, stemming from class imbalance favoring head classes, and (2) a transition-level bias arising from variations in transitions, prioritizing commonly observed transitions. As a remedy, we introduce a constrained optimization problem to alleviate both biases. We define learning states for action classes and their associated transitions and integrate them into the optimization process. We propose a novel cost-sensitive loss function formulated as a weighted cross-entropy loss, with weights adaptively adjusted based on the learning state of actions and their transitions. Experiments on three challenging temporal segmentation benchmarks and various frameworks demonstrate the effectiveness of our approach, resulting in significant improvements in both per-class frame-wise and segment-wise performance.

Md Salman Shamil, Dibyadip Chatterjee, Fadime Sener et al.

3D hand pose is an underexplored modality for action recognition. Poses are compact yet informative and can greatly benefit applications with limited compute budgets. However, poses alone offer an incomplete understanding of actions, as they cannot fully capture objects and environments with which humans interact. We propose HandFormer, a novel multimodal transformer, to efficiently model hand-object interactions. HandFormer combines 3D hand poses at a high temporal resolution for fine-grained motion modeling with sparsely sampled RGB frames for encoding scene semantics. Observing the unique characteristics of hand poses, we temporally factorize hand modeling and represent each joint by its short-term trajectories. This factorized pose representation combined with sparse RGB samples is remarkably efficient and highly accurate. Unimodal HandFormer with only hand poses outperforms existing skeleton-based methods at 5x fewer FLOPs. With RGB, we achieve new state-of-the-art performance on Assembly101 and H2O with significant improvements in egocentric action recognition.

Abhinav Rai, Fadime Sener, Angela Yao

Modeling the visual changes that an action brings to a scene is critical for video understanding. Currently, CNNs process one local neighbourhood at a time, thus contextual relationships over longer ranges, while still learnable, are indirect. We present TROI, a plug-and-play module for CNNs to reason between mid-level feature representations that are otherwise separated in space and time. The module relates localized visual entities such as hands and interacting objects and transforms their corresponding regions of interest directly in the feature maps of convolutional layers. With TROI, we achieve state-of-the-art action recognition results on the large-scale datasets Something-Something-V2 and EPIC-Kitchens-100.

Fadime Sener, Rishabh Saraf, Angela Yao

Can we teach a robot to recognize and make predictions for activities that it has never seen before? We tackle this problem by learning models for video from text. This paper presents a hierarchical model that generalizes instructional knowledge from large-scale text corpora and transfers the knowledge to video. Given a portion of an instructional video, our model recognizes and predicts coherent and plausible actions multiple steps into the future, all in rich natural language. To demonstrate the capabilities of our model, we introduce the \emph{Tasty Videos Dataset V2}, a collection of 4022 recipes for zero-shot learning, recognition and anticipation. Extensive experiments with various evaluation metrics demonstrate the potential of our method for generalization, given limited video data for training models.

Fadime Sener, Dibyadip Chatterjee, Angela Yao

This technical report extends our work presented in [9] with more experiments. In [9], we tackle long-term video understanding, which requires reasoning from current and past or future observations and raises several fundamental questions. How should temporal or sequential relationships be modelled? What temporal extent of information and context needs to be processed? At what temporal scale should they be derived? [9] addresses these questions with a flexible multi-granular temporal aggregation framework. In this report, we conduct further experiments with this framework on different tasks and a new dataset, EPIC-KITCHENS-100.

Fadime Sener, Dipika Singhania, Angela Yao

Future prediction, especially in long-range videos, requires reasoning from current and past observations. In this work, we address questions of temporal extent, scaling, and level of semantic abstraction with a flexible multi-granular temporal aggregation framework. We show that it is possible to achieve state of the art in both next action and dense anticipation with simple techniques such as max-pooling and attention. To demonstrate the anticipation capabilities of our model, we conduct experiments on Breakfast, 50Salads, and EPIC-Kitchens datasets, where we achieve state-of-the-art results. With minimal modifications, our model can also be extended for video segmentation and action recognition.

Anna Kukleva, Hilde Kuehne, Fadime Sener et al.

The task of temporally detecting and segmenting actions in untrimmed videos has seen an increased attention recently. One problem in this context arises from the need to define and label action boundaries to create annotations for training which is very time and cost intensive. To address this issue, we propose an unsupervised approach for learning action classes from untrimmed video sequences. To this end, we use a continuous temporal embedding of framewise features to benefit from the sequential nature of activities. Based on the latent space created by the embedding, we identify clusters of temporal segments across all videos that correspond to semantic meaningful action classes. The approach is evaluated on three challenging datasets, namely the Breakfast dataset, YouTube Instructions, and the 50Salads dataset. While previous works assumed that the videos contain the same high level activity, we furthermore show that the proposed approach can also be applied to a more general setting where the content of the videos is unknown.

Divyansh Aggarwal, Elchin Valiyev, Fadime Sener et al.

When judging style, a key question that often arises is whether or not a pair of objects are compatible with each other. In this paper we investigate how Siamese networks can be used efficiently for assessing the style compatibility between images of furniture items. We show that the middle layers of pretrained CNNs can capture essential information about furniture style, which allows for efficient applications of such networks for this task. We also use a joint image-text embedding method that allows for the querying of stylistically compatible furniture items, along with additional attribute constraints based on text. To evaluate our methods, we collect and present a large scale dataset of images of furniture of different style categories accompanied by text attributes.

Fadime Sener, Angela Yao

How can we teach a robot to predict what will happen next for an activity it has never seen before? We address this problem of zero-shot anticipation by presenting a hierarchical model that generalizes instructional knowledge from large-scale text-corpora and transfers the knowledge to the visual domain. Given a portion of an instructional video, our model predicts coherent and plausible actions multiple steps into the future, all in rich natural language. To demonstrate the anticipation capabilities of our model, we introduce the Tasty Videos dataset, a collection of 2511 recipes for zero-shot learning, recognition and anticipation.

Fadime Sener, Angela Yao

This paper presents a new method for unsupervised segmentation of complex activities from video into multiple steps, or sub-activities, without any textual input. We propose an iterative discriminative-generative approach which alternates between discriminatively learning the appearance of sub-activities from the videos' visual features to sub-activity labels and generatively modelling the temporal structure of sub-activities using a Generalized Mallows Model. In addition, we introduce a model for background to account for frames unrelated to the actual activities. Our approach is validated on the challenging Breakfast Actions and Inria Instructional Videos datasets and outperforms both unsupervised and weakly-supervised state of the art.

Samet Hicsonmez, Nermin Samet, Fadime Sener et al.

This paper is motivated from a young boy's capability to recognize an illustrator's style in a totally different context. In the book "We are All Born Free" [1], composed of selected rights from the Universal Declaration of Human Rights interpreted by different illustrators, the boy was surprised to see a picture similar to the ones in the "Winnie the Witch" series drawn by Korky Paul (Figure 1). The style was noticeable in other characters of the same illustrator in different books as well. The capability of a child to easily spot the style was shown to be valid for other illustrators such as Axel Scheffler and Debi Gliori. The boy's enthusiasm let us to start the journey to explore the capabilities of machines to recognize the style of illustrators. We collected pages from children's books to construct a new illustrations dataset consisting of about 6500 pages from 24 artists. We exploited deep networks for categorizing illustrators and with around 94% classification performance our method over-performed the traditional methods by more than 10%. Going beyond categorization we explored transferring style. The classification performance on the transferred images has shown the ability of our system to capture the style. Furthermore, we discovered representative illustrations and discriminative stylistic elements.