Few-Shot Action Recognition (FSAR) aims to train a model with only a few labeled video instances. A key challenge in FSAR is handling divergent narrative trajectories for precise video matching. While the frame- and tuple-level alignment approaches have been promising, their methods heavily rely on pre-defined and length-dependent alignment units (e.g., frames or tuples), which limits flexibility for actions of varying lengths and speeds. In this work, we introduce a novel TEmporal Alignment-free Matching (TEAM) approach, which eliminates the need for temporal units in action representation and brute-force alignment during matching. Specifically, TEAM represents each video with a fixed set of pattern tokens that capture globally discriminative clues within the video instance regardless of action length or speed, ensuring its flexibility. Furthermore, TEAM is inherently efficient, using token-wise comparisons to measure similarity between videos, unlike existing methods that rely on pairwise comparisons for temporal alignment. Additionally, we propose an adaptation process that identifies and removes common information across classes, establishing clear boundaries even between novel categories. Extensive experiments demonstrate the effectiveness of TEAM. Codes are available at github.com/leesb7426/TEAM.
Large-scale pre-trained models have achieved remarkable success in language and image tasks, leading an increasing number of studies to explore the application of pre-trained image models, such as CLIP, in the domain of few-shot action recognition (FSAR). However, current methods generally suffer from several problems: 1) Direct fine-tuning often undermines the generalization capability of the pre-trained model; 2) The exploration of task-specific information is insufficient in the visual tasks; 3) The semantic order information is typically overlooked during text modeling; 4) Existing cross-modal alignment techniques ignore the temporal coupling of multimodal information. To address these, we propose Task-Adapter++, a parameter-efficient dual adaptation method for both image and text encoders. Specifically, to make full use of the variations across different few-shot learning tasks, we design a task-specific adaptation for the image encoder so that the most discriminative information can be well noticed during feature extraction. Furthermore, we leverage large language models (LLMs) to generate detailed sequential sub-action descriptions for each action class, and introduce semantic order adapters into the text encoder to effectively model the sequential relationships between these sub-actions. Finally, we develop an innovative fine-grained cross-modal alignment strategy that actively maps visual features to reside in the same temporal stage as semantic descriptions. Extensive experiments fully demonstrate the effectiveness and superiority of the proposed method, which achieves state-of-the-art performance on 5 benchmarks consistently. The code is open-sourced at https://github.com/Jaulin-Bage/Task-Adapter-pp.
Going beyond few-shot action recognition (FSAR), cross-domain FSAR (CDFSAR) has attracted recent research interests by solving the domain gap lying in source-to-target transfer learning. Existing CDFSAR methods mainly focus on joint training of source and target data to mitigate the side effect of domain gap. However, such kind of methods suffer from two limitations: First, pair-wise joint training requires retraining deep models in case of one source data and multiple target ones, which incurs heavy computation cost, especially for large source and small target data. Second, pre-trained models after joint training are adopted to target domain in a straightforward manner, hardly taking full potential of pre-trained models and then limiting recognition performance. To overcome above limitations, this paper proposes a simple yet effective baseline, namely Temporal-Aware Model Tuning (TAMT) for CDFSAR. Specifically, our TAMT involves a decoupled paradigm by performing pre-training on source data and fine-tuning target data, which avoids retraining for multiple target data with single source. To effectively and efficiently explore the potential of pre-trained models in transferring to target domain, our TAMT proposes a Hierarchical Temporal Tuning Network (HTTN), whose core involves local temporal-aware adapters (TAA) and a global temporal-aware moment tuning (GTMT). Particularly, TAA learns few parameters to recalibrate the intermediate features of frozen pre-trained models, enabling efficient adaptation to target domains. Furthermore, GTMT helps to generate powerful video representations, improving match performance on the target domain. Experiments on several widely used video benchmarks show our TAMT outperforms the recently proposed counterparts by 13%$\sim$31%, achieving new state-of-the-art CDFSAR results.
High frame-rate (HFR) videos of action recognition improve fine-grained expression while reducing the spatio-temporal relation and motion information density. Thus, large amounts of video samples are continuously required for traditional data-driven training. However, samples are not always sufficient in real-world scenarios, promoting few-shot action recognition (FSAR) research. We observe that most recent FSAR works build spatio-temporal relation of video samples via temporal alignment after spatial feature extraction, cutting apart spatial and temporal features within samples. They also capture motion information via narrow perspectives between adjacent frames without considering density, leading to insufficient motion information capturing. Therefore, we propose a novel plug-and-play architecture for FSAR called Spatio-tempOral frAme tuPle enhancer (SOAP) in this paper. The model we designed with such architecture refers to SOAP-Net. Temporal connections between different feature channels and spatio-temporal relation of features are considered instead of simple feature extraction. Comprehensive motion information is also captured, using frame tuples with multiple frames containing more motion information than adjacent frames. Combining frame tuples of diverse frame counts further provides a broader perspective. SOAP-Net achieves new state-of-the-art performance across well-known benchmarks such as SthSthV2, Kinetics, UCF101, and HMDB51. Extensive empirical evaluations underscore the competitiveness, pluggability, generalization, and robustness of SOAP. The code is released at https://github.com/wenbohuang1002/SOAP.
Adapting large pre-trained image models to few-shot action recognition has proven to be an effective and efficient strategy for learning robust feature extractors, which is essential for few-shot learning. Typical fine-tuning based adaptation paradigm is prone to overfitting in the few-shot learning scenarios and offers little modeling flexibility for learning temporal features in video data. In this work we present the Disentangled-and-Deformable Spatio-Temporal Adapter (D$^2$ST-Adapter), which is a novel adapter tuning framework well-suited for few-shot action recognition due to lightweight design and low parameter-learning overhead. It is designed in a dual-pathway architecture to encode spatial and temporal features in a disentangled manner. In particular, we devise the anisotropic Deformable Spatio-Temporal Attention module as the core component of D$^2$ST-Adapter, which can be tailored with anisotropic sampling densities along spatial and temporal domains to learn spatial and temporal features specifically in corresponding pathways, allowing our D$^2$ST-Adapter to encode features in a global view in 3D spatio-temporal space while maintaining a lightweight design. Extensive experiments with instantiations of our method on both pre-trained ResNet and ViT demonstrate the superiority of our method over state-of-the-art methods for few-shot action recognition. Our method is particularly well-suited to challenging scenarios where temporal dynamics are critical for action recognition.
Few-shot video action recognition is an effective approach to recognizing new categories with only a few labeled examples, thereby reducing the challenges associated with collecting and annotating large-scale video datasets. Existing methods in video action recognition rely on large labeled datasets from the same domain. However, this setup is not realistic as novel categories may come from different data domains that may have different spatial and temporal characteristics. This dissimilarity between the source and target domains can pose a significant challenge, rendering traditional few-shot action recognition techniques ineffective. To address this issue, in this work, we propose a novel cross-domain few-shot video action recognition method that leverages self-supervised learning and curriculum learning to balance the information from the source and target domains. To be particular, our method employs a masked autoencoder-based self-supervised training objective to learn from both source and target data in a self-supervised manner. Then a progressive curriculum balances learning the discriminative information from the source dataset with the generic information learned from the target domain. Initially, our curriculum utilizes supervised learning to learn class discriminative features from the source data. As the training progresses, we transition to learning target-domain-specific features. We propose a progressive curriculum to encourage the emergence of rich features in the target domain based on class discriminative supervised features in the source domain. We evaluate our method on several challenging benchmark datasets and demonstrate that our approach outperforms existing cross-domain few-shot learning techniques. Our code is available at https://github.com/Sarinda251/CDFSL-V
Class prototype construction and matching are core aspects of few-shot action recognition. Previous methods mainly focus on designing spatiotemporal relation modeling modules or complex temporal alignment algorithms. Despite the promising results, they ignored the value of class prototype construction and matching, leading to unsatisfactory performance in recognizing similar categories in every task. In this paper, we propose GgHM, a new framework with Graph-guided Hybrid Matching. Concretely, we learn task-oriented features by the guidance of a graph neural network during class prototype construction, optimizing the intra- and inter-class feature correlation explicitly. Next, we design a hybrid matching strategy, combining frame-level and tuple-level matching to classify videos with multivariate styles. We additionally propose a learnable dense temporal modeling module to enhance the video feature temporal representation to build a more solid foundation for the matching process. GgHM shows consistent improvements over other challenging baselines on several few-shot datasets, demonstrating the effectiveness of our method. The code will be publicly available at https://github.com/jiazheng-xing/GgHM.
In the research field of few-shot learning, the main difference between image-based and video-based is the additional temporal dimension. In recent years, some works have used the Transformer to deal with frames, then get the attention feature and the enhanced prototype, and the results are competitive. However, some video frames may relate little to the action, and only using single frame-level or segment-level features may not mine enough information. We address these problems sequentially through an end-to-end method named "Task-Specific Alignment and Multiple-level Transformer Network (TSA-MLT)". The first module (TSA) aims at filtering the action-irrelevant frames for action duration alignment. Affine Transformation for frame sequence in the time dimension is used for linear sampling. The second module (MLT) focuses on the Multiple-level feature of the support prototype and query sample to mine more information for the alignment, which operates on different level features. We adopt a fusion loss according to a fusion distance that fuses the L2 sequence distance, which focuses on temporal order alignment, and the Optimal Transport distance, which focuses on measuring the gap between the appearance and semantics of the videos. Extensive experiments show our method achieves state-of-the-art results on the HMDB51 and UCF101 datasets and a competitive result on the benchmark of Kinetics and something 2-something V2 datasets. Our code is available at the URL: https://github.com/cofly2014/tsa-mlt.git
Current state-of-the-art approaches for few-shot action recognition achieve promising performance by conducting frame-level matching on learned visual features. However, they generally suffer from two limitations: i) the matching procedure between local frames tends to be inaccurate due to the lack of guidance to force long-range temporal perception; ii) explicit motion learning is usually ignored, leading to partial information loss. To address these issues, we develop a Motion-augmented Long-short Contrastive Learning (MoLo) method that contains two crucial components, including a long-short contrastive objective and a motion autodecoder. Specifically, the long-short contrastive objective is to endow local frame features with long-form temporal awareness by maximizing their agreement with the global token of videos belonging to the same class. The motion autodecoder is a lightweight architecture to reconstruct pixel motions from the differential features, which explicitly embeds the network with motion dynamics. By this means, MoLo can simultaneously learn long-range temporal context and motion cues for comprehensive few-shot matching. To demonstrate the effectiveness, we evaluate MoLo on five standard benchmarks, and the results show that MoLo favorably outperforms recent advanced methods. The source code is available at https://github.com/alibaba-mmai-research/MoLo.
Large scale Vision-Language (VL) models have shown tremendous success in aligning representations between visual and text modalities. This enables remarkable progress in zero-shot recognition, image generation & editing, and many other exciting tasks. However, VL models tend to over-represent objects while paying much less attention to verbs, and require additional tuning on video data for best zero-shot action recognition performance. While previous work relied on large-scale, fully-annotated data, in this work we propose an unsupervised approach. We adapt a VL model for zero-shot and few-shot action recognition using a collection of unlabeled videos and an unpaired action dictionary. Based on that, we leverage Large Language Models and VL models to build a text bag for each unlabeled video via matching, text expansion and captioning. We use those bags in a Multiple Instance Learning setup to adapt an image-text backbone to video data. Although finetuned on unlabeled video data, our resulting models demonstrate high transferability to numerous unseen zero-shot downstream tasks, improving the base VL model performance by up to 14\%, and even comparing favorably to fully-supervised baselines in both zero-shot and few-shot video recognition transfer. The code will be released later at \url{https://github.com/wlin-at/MAXI}.
Learning from large-scale contrastive language-image pre-training like CLIP has shown remarkable success in a wide range of downstream tasks recently, but it is still under-explored on the challenging few-shot action recognition (FSAR) task. In this work, we aim to transfer the powerful multimodal knowledge of CLIP to alleviate the inaccurate prototype estimation issue due to data scarcity, which is a critical problem in low-shot regimes. To this end, we present a CLIP-guided prototype modulating framework called CLIP-FSAR, which consists of two key components: a video-text contrastive objective and a prototype modulation. Specifically, the former bridges the task discrepancy between CLIP and the few-shot video task by contrasting videos and corresponding class text descriptions. The latter leverages the transferable textual concepts from CLIP to adaptively refine visual prototypes with a temporal Transformer. By this means, CLIP-FSAR can take full advantage of the rich semantic priors in CLIP to obtain reliable prototypes and achieve accurate few-shot classification. Extensive experiments on five commonly used benchmarks demonstrate the effectiveness of our proposed method, and CLIP-FSAR significantly outperforms existing state-of-the-art methods under various settings. The source code and models will be publicly available at https://github.com/alibaba-mmai-research/CLIP-FSAR.
Recent attempts mainly focus on learning deep representations for each video individually under the episodic meta-learning regime and then performing temporal alignment to match query and support videos. However, they still suffer from two drawbacks: (i) learning individual features without considering the entire task may result in limited representation capability, and (ii) existing alignment strategies are sensitive to noises and misaligned instances. To handle the two limitations, we propose a novel Hybrid Relation guided temporal Set Matching (HyRSM++) approach for few-shot action recognition. The core idea of HyRSM++ is to integrate all videos within the task to learn discriminative representations and involve a robust matching technique. To be specific, HyRSM++ consists of two key components, a hybrid relation module and a temporal set matching metric. Given the basic representations from the feature extractor, the hybrid relation module is introduced to fully exploit associated relations within and cross videos in an episodic task and thus can learn task-specific embeddings. Subsequently, in the temporal set matching metric, we carry out the distance measure between query and support videos from a set matching perspective and design a Bi-MHM to improve the resilience to misaligned instances. In addition, we explicitly exploit the temporal coherence in videos to regularize the matching process. Furthermore, we extend the proposed HyRSM++ to deal with the more challenging semi-supervised few-shot action recognition and unsupervised few-shot action recognition tasks. Experimental results on multiple benchmarks demonstrate that our method achieves state-of-the-art performance under various few-shot settings. The source code is available at https://github.com/alibaba-mmai-research/HyRSMPlusPlus.
Video representation learning has been successful in video-text pre-training for zero-shot transfer, where each sentence is trained to be close to the paired video clips in a common feature space. For long videos, given a paragraph of description where the sentences describe different segments of the video, by matching all sentence-clip pairs, the paragraph and the full video are aligned implicitly. However, such unit-level comparison may ignore global temporal context, which inevitably limits the generalization ability. In this paper, we propose a contrastive learning framework TempCLR to compare the full video and the paragraph explicitly. As the video/paragraph is formulated as a sequence of clips/sentences, under the constraint of their temporal order, we use dynamic time warping to compute the minimum cumulative cost over sentence-clip pairs as the sequence-level distance. To explore the temporal dynamics, we break the consistency of temporal succession by shuffling video clips w.r.t. temporal granularity. Then, we obtain the representations for clips/sentences, which perceive the temporal information and thus facilitate the sequence alignment. In addition to pre-training on the video and paragraph, our approach can also generalize on the matching between video instances. We evaluate our approach on video retrieval, action step localization, and few-shot action recognition, and achieve consistent performance gain over all three tasks. Detailed ablation studies are provided to justify the approach design.
Dynamic Time Warping (DTW) is used for matching pairs of sequences and celebrated in applications such as forecasting the evolution of time series, clustering time series or even matching sequence pairs in few-shot action recognition. The transportation plan of DTW contains a set of paths; each path matches frames between two sequences under a varying degree of time warping, to account for varying temporal intra-class dynamics of actions. However, as DTW is the smallest distance among all paths, it may be affected by the feature uncertainty which varies across time steps/frames. Thus, in this paper, we propose to model the so-called aleatoric uncertainty of a differentiable (soft) version of DTW. To this end, we model the heteroscedastic aleatoric uncertainty of each path by the product of likelihoods from Normal distributions, each capturing variance of pair of frames. (The path distance is the sum of base distances between features of pairs of frames of the path.) The Maximum Likelihood Estimation (MLE) applied to a path yields two terms: (i) a sum of Euclidean distances weighted by the variance inverse, and (ii) a sum of log-variance regularization terms. Thus, our uncertainty-DTW is the smallest weighted path distance among all paths, and the regularization term (penalty for the high uncertainty) is the aggregate of log-variances along the path. The distance and the regularization term can be used in various objectives. We showcase forecasting the evolution of time series, estimating the Fr\'echet mean of time series, and supervised/unsupervised few-shot action recognition of the articulated human 3D body joints.
A primary challenge faced in few-shot action recognition is inadequate video data for training. To address this issue, current methods in this field mainly focus on devising algorithms at the feature level while little attention is paid to processing input video data. Moreover, existing frame sampling strategies may omit critical action information in temporal and spatial dimensions, which further impacts video utilization efficiency. In this paper, we propose a novel video frame sampler for few-shot action recognition to address this issue, where task-specific spatial-temporal frame sampling is achieved via a temporal selector (TS) and a spatial amplifier (SA). Specifically, our sampler first scans the whole video at a small computational cost to obtain a global perception of video frames. The TS plays its role in selecting top-T frames that contribute most significantly and subsequently. The SA emphasizes the discriminative information of each frame by amplifying critical regions with the guidance of saliency maps. We further adopt task-adaptive learning to dynamically adjust the sampling strategy according to the episode task at hand. Both the implementations of TS and SA are differentiable for end-to-end optimization, facilitating seamless integration of our proposed sampler with most few-shot action recognition methods. Extensive experiments show a significant boost in the performances on various benchmarks including long-term videos.The code is available at https://github.com/R00Kie-Liu/Sampler
We propose a Multi-level Second-order (MlSo) few-shot learning network for supervised or unsupervised few-shot image classification and few-shot action recognition. We leverage so-called power-normalized second-order base learner streams combined with features that express multiple levels of visual abstraction, and we use self-supervised discriminating mechanisms. As Second-order Pooling (SoP) is popular in image recognition, we employ its basic element-wise variant in our pipeline. The goal of multi-level feature design is to extract feature representations at different layer-wise levels of CNN, realizing several levels of visual abstraction to achieve robust few-shot learning. As SoP can handle convolutional feature maps of varying spatial sizes, we also introduce image inputs at multiple spatial scales into MlSo. To exploit the discriminative information from multi-level and multi-scale features, we develop a Feature Matching (FM) module that reweights their respective branches. We also introduce a self-supervised step, which is a discriminator of the spatial level and the scale of abstraction. Our pipeline is trained in an end-to-end manner. With a simple architecture, we demonstrate respectable results on standard datasets such as Omniglot, mini-ImageNet, tiered-ImageNet, Open MIC, fine-grained datasets such as CUB Birds, Stanford Dogs and Cars, and action recognition datasets such as HMDB51, UCF101, and mini-MIT.
We propose a novel few-shot action recognition framework, STRM, which enhances class-specific feature discriminability while simultaneously learning higher-order temporal representations. The focus of our approach is a novel spatio-temporal enrichment module that aggregates spatial and temporal contexts with dedicated local patch-level and global frame-level feature enrichment sub-modules. Local patch-level enrichment captures the appearance-based characteristics of actions. On the other hand, global frame-level enrichment explicitly encodes the broad temporal context, thereby capturing the relevant object features over time. The resulting spatio-temporally enriched representations are then utilized to learn the relational matching between query and support action sub-sequences. We further introduce a query-class similarity classifier on the patch-level enriched features to enhance class-specific feature discriminability by reinforcing the feature learning at different stages in the proposed framework. Experiments are performed on four few-shot action recognition benchmarks: Kinetics, SSv2, HMDB51 and UCF101. Our extensive ablation study reveals the benefits of the proposed contributions. Furthermore, our approach sets a new state-of-the-art on all four benchmarks. On the challenging SSv2 benchmark, our approach achieves an absolute gain of $3.5\%$ in classification accuracy, as compared to the best existing method in the literature. Our code and models are available at https://github.com/Anirudh257/strm.
Recognizing human actions is fundamentally a spatio-temporal reasoning problem, and should be, at least to some extent, invariant to the appearance of the human and the objects involved. Motivated by this hypothesis, in this work, we take an object-centric approach to action recognition. Multiple works have studied this setting before, yet it remains unclear (i) how well a carefully crafted, spatio-temporal layout-based method can recognize human actions, and (ii) how, and when, to fuse the information from layout and appearance-based models. The main focus of this paper is compositional/few-shot action recognition, where we advocate the usage of multi-head attention (proven to be effective for spatial reasoning) over spatio-temporal layouts, i.e., configurations of object bounding boxes. We evaluate different schemes to inject video appearance information to the system, and benchmark our approach on background cluttered action recognition. On the Something-Else and Action Genome datasets, we demonstrate (i) how to extend multi-head attention for spatio-temporal layout-based action recognition, (ii) how to improve the performance of appearance-based models by fusion with layout-based models, (iii) that even on non-compositional background-cluttered video datasets, a fusion between layout- and appearance-based models improves the performance.
Zero-shot action recognition is the task of classifying action categories that are not available in the training set. In this setting, the standard evaluation protocol is to use existing action recognition datasets(e.g. UCF101) and randomly split the classes into seen and unseen. However, most recent work builds on representations pre-trained on the Kinetics dataset, where classes largely overlap with classes in the zero-shot evaluation datasets. As a result, classes which are supposed to be unseen, are present during supervised pre-training, invalidating the condition of the zero-shot setting. A similar concern was previously noted several years ago for image based zero-shot recognition but has not been considered by the zero-shot action recognition community. In this paper, we propose a new split for true zero-shot action recognition with no overlap between unseen test classes and training or pre-training classes. We benchmark several recent approaches on the proposed True Zero-Shot(TruZe) Split for UCF101 and HMDB51, with zero-shot and generalized zero-shot evaluation. In our extensive analysis, we find that our TruZesplits are significantly harder than comparable random splits as nothing is leaking from pre-training, i.e. unseen performance is consistently lower,up to 8.9% for zero-shot action recognition. In an additional evaluation we also find that similar issues exist in the splits used in few-shot action recognition, here we see differences of up to 17.1%. We publish oursplits1and hope that our benchmark analysis will change how the field is evaluating zero- and few-shot action recognition moving forward.
Few-shot action recognition aims to recognize novel action classes (query) using just a few samples (support). The majority of current approaches follow the metric learning paradigm, which learns to compare the similarity between videos. Recently, it has been observed that directly measuring this similarity is not ideal since different action instances may show distinctive temporal distribution, resulting in severe misalignment issues across query and support videos. In this paper, we arrest this problem from two distinct aspects -- action duration misalignment and action evolution misalignment. We address them sequentially through a Two-stage Action Alignment Network (TA2N). The first stage locates the action by learning a temporal affine transform, which warps each video feature to its action duration while dismissing the action-irrelevant feature (e.g. background). Next, the second stage coordinates query feature to match the spatial-temporal action evolution of support by performing temporally rearrange and spatially offset prediction. Extensive experiments on benchmark datasets show the potential of the proposed method in achieving state-of-the-art performance for few-shot action recognition.The code of this project can be found at https://github.com/R00Kie-Liu/TA2N