Token interaction operation is one of the core modules in MLP-based models to exchange and aggregate information between different spatial locations. However, the power of token interaction on the spatial dimension is highly dependent on the spatial resolution of the feature maps, which limits the model's expressive ability, especially in deep layers where the feature are down-sampled to a small spatial size. To address this issue, we present a novel method called Strip-MLP to enrich the token interaction power in three ways. Firstly, we introduce a new MLP paradigm called Strip MLP layer that allows the token to interact with other tokens in a cross-strip manner, enabling the tokens in a row (or column) contribute to the information aggregations in adjacent but different strips of rows (or columns). Secondly, a Cascade Group Strip Mixing Module (CGSMM) is proposed to overcome the performance degradation caused by small spatial feature size. The module allows tokens to interact more effectively in the manners of within-patch and cross-patch, which is independent to the feature spatial size. Finally, based on the Strip MLP layer, we propose a novel Local Strip Mixing Module (LSMM) to boost the token interaction power in the local region. Extensive experiments demonstrate that Strip-MLP significantly improves the performance of MLP-based models on small datasets and obtains comparable or even better results on ImageNet with great superiorities on the number of parameters and FLOPs. In particular, Strip-MLP models achieve higher average Top-1 accuracy than existing MLP-based models by +2.44% on Caltech-101 and +2.16% on CIFAR-100. The source codes will be available at https://github.com/Med-Process/Strip_MLP.
论文检索
输入标题、作者或关键词，从 2,156 篇学术成果中精准定位
In this paper, we consider a real-world scenario where a model that is trained on pre-defined classes continually encounters unlabeled data that contains both known and novel classes. The goal is to continually discover novel classes while maintaining the performance in known classes. We name the setting Continual Generalized Category Discovery (C-GCD). Existing methods for novel class discovery cannot directly handle the C-GCD setting due to some unrealistic assumptions, such as the unlabeled data only containing novel classes. Furthermore, they fail to discover novel classes in a continual fashion. In this work, we lift all these assumptions and propose an approach, called MetaGCD, to learn how to incrementally discover with less forgetting. Our proposed method uses a meta-learning framework and leverages the offline labeled data to simulate the testing incremental learning process. A meta-objective is defined to revolve around two conflicting learning objectives to achieve novel class discovery without forgetting. Furthermore, a soft neighborhood-based contrastive network is proposed to discriminate uncorrelated images while attracting correlated images. We build strong baselines and conduct extensive experiments on three widely used benchmarks to demonstrate the superiority of our method.
Self-supervised monocular depth estimation (SS-MDE) has the potential to scale to vast quantities of data. Unfortunately, existing approaches limit themselves to the automotive domain, resulting in models incapable of generalizing to complex environments such as natural or indoor settings. To address this, we propose a large-scale SlowTV dataset curated from YouTube, containing an order of magnitude more data than existing automotive datasets. SlowTV contains 1.7M images from a rich diversity of environments, such as worldwide seasonal hiking, scenic driving and scuba diving. Using this dataset, we train an SS-MDE model that provides zero-shot generalization to a large collection of indoor/outdoor datasets. The resulting model outperforms all existing SSL approaches and closes the gap on supervised SoTA, despite using a more efficient architecture. We additionally introduce a collection of best-practices to further maximize performance and zero-shot generalization. This includes 1) aspect ratio augmentation, 2) camera intrinsic estimation, 3) support frame randomization and 4) flexible motion estimation.
Although Deep Neural Networks (DNNs) have demonstrated excellent performance, they are vulnerable to adversarial patches that introduce perceptible and localized perturbations to the input. Generating adversarial patches on images has received much attention, while adversarial patches on videos have not been well investigated. Further, decision-based attacks, where attackers only access the predicted hard labels by querying threat models, have not been well explored on video models either, even if they are practical in real-world video recognition scenes. The absence of such studies leads to a huge gap in the robustness assessment for video models. To bridge this gap, this work first explores decision-based patch attacks on video models. We analyze that the huge parameter space brought by videos and the minimal information returned by decision-based models both greatly increase the attack difficulty and query burden. To achieve a query-efficient attack, we propose a spatial-temporal differential evolution (STDE) framework. First, STDE introduces target videos as patch textures and only adds patches on keyframes that are adaptively selected by temporal difference. Second, STDE takes minimizing the patch area as the optimization objective and adopts spatial-temporal mutation and crossover to search for the global optimum without falling into the local optimum. Experiments show STDE has demonstrated state-of-the-art performance in terms of threat, efficiency and imperceptibility. Hence, STDE has the potential to be a powerful tool for evaluating the robustness of video recognition models.
Face Anti-Spoofing (FAS) is recently studied under the continual learning setting, where the FAS models are expected to evolve after encountering data from new domains. However, existing methods need extra replay buffers to store previous data for rehearsal, which becomes infeasible when previous data is unavailable because of privacy issues. In this paper, we propose the first rehearsal-free method for Domain Continual Learning (DCL) of FAS, which deals with catastrophic forgetting and unseen domain generalization problems simultaneously. For better generalization to unseen domains, we design the Dynamic Central Difference Convolutional Adapter (DCDCA) to adapt Vision Transformer (ViT) models during the continual learning sessions. To alleviate the forgetting of previous domains without using previous data, we propose the Proxy Prototype Contrastive Regularization (PPCR) to constrain the continual learning with previous domain knowledge from the proxy prototypes. Simulating practical DCL scenarios, we devise two new protocols which evaluate both generalization and anti-forgetting performance. Extensive experimental results show that our proposed method can improve the generalization performance in unseen domains and alleviate the catastrophic forgetting of previous knowledge. The code and protocol files are released on https://github.com/RizhaoCai/DCL-FAS-ICCV2023.
The current popular methods for video object segmentation (VOS) implement feature matching through several hand-crafted modules that separately perform feature extraction and matching. However, the above hand-crafted designs empirically cause insufficient target interaction, thus limiting the dynamic target-aware feature learning in VOS. To tackle these limitations, this paper presents a scalable Simplified VOS (SimVOS) framework to perform joint feature extraction and matching by leveraging a single transformer backbone. Specifically, SimVOS employs a scalable ViT backbone for simultaneous feature extraction and matching between query and reference features. This design enables SimVOS to learn better target-ware features for accurate mask prediction. More importantly, SimVOS could directly apply well-pretrained ViT backbones (e.g., MAE) for VOS, which bridges the gap between VOS and large-scale self-supervised pre-training. To achieve a better performance-speed trade-off, we further explore within-frame attention and propose a new token refinement module to improve the running speed and save computational cost. Experimentally, our SimVOS achieves state-of-the-art results on popular video object segmentation benchmarks, i.e., DAVIS-2017 (88.0% J&F), DAVIS-2016 (92.9% J&F) and YouTube-VOS 2019 (84.2% J&F), without applying any synthetic video or BL30K pre-training used in previous VOS approaches. Our code and models are available at https://github.com/jimmy-dq/SimVOS.git.
Continual Learning (CL) involves training a machine learning model in a sequential manner to learn new information while retaining previously learned tasks without the presence of previous training data. Although there has been significant interest in CL, most recent CL approaches in computer vision have focused on convolutional architectures only. However, with the recent success of vision transformers, there is a need to explore their potential for CL. Although there have been some recent CL approaches for vision transformers, they either store training instances of previous tasks or require a task identifier during test time, which can be limiting. This paper proposes a new exemplar-free approach for class/task incremental learning called ConTraCon, which does not require task-id to be explicitly present during inference and avoids the need for storing previous training instances. The proposed approach leverages the transformer architecture and involves re-weighting the key, query, and value weights of the multi-head self-attention layers of a transformer trained on a similar task. The re-weighting is done using convolution, which enables the approach to maintain low parameter requirements per task. Additionally, an image augmentation-based entropic task identification approach is used to predict tasks without requiring task-ids during inference. Experiments on four benchmark datasets demonstrate that the proposed approach outperforms several competitive approaches while requiring fewer parameters.
Prototype Reminiscence and Augmented Asymmetric Knowledge Aggregation for Non-Exemplar Class-Incremental Learning
PDF ↗Non-exemplar class-incremental learning (NECIL) requires deep models to maintain existing knowledge while continuously learning new classes without saving old class samples. In NECIL methods, prototypical representations are usually stored, which inject information from former classes to resist catastrophic forgetting in subsequent incremental learning. However, since the model continuously learns new knowledge, the stored prototypical representations cannot correctly model the properties of old classes in the existence of knowledge updates. To address this problem, we propose a novel prototype reminiscence mechanism that incorporates the previous class prototypes with arriving new class features to dynamically reshape old class feature distributions thus preserving the decision boundaries of previous tasks. In addition, to improve the model generalization on both newly arriving classes and old classes, we contribute an augmented asymmetric knowledge aggregation approach, which aggregates the overall knowledge of the current task and extracts the valuable knowledge of the past tasks, on top of self-supervised label augmentation. Experimental results on three benchmarks suggest the superior performance of our approach over the SOTA methods.
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.
We demonstrate a compact, cost-effective snapshot spectral imaging system named Aperture Diffraction Imaging Spectrometer (ADIS), which consists only of an imaging lens with an ultra-thin orthogonal aperture mask and a mosaic filter sensor, requiring no additional physical footprint compared to common RGB cameras. Then we introduce a new optical design that each point in the object space is multiplexed to discrete encoding locations on the mosaic filter sensor by diffraction-based spatial-spectral projection engineering generated from the orthogonal mask. The orthogonal projection is uniformly accepted to obtain a weakly calibration-dependent data form to enhance modulation robustness. Meanwhile, the Cascade Shift-Shuffle Spectral Transformer (CSST) with strong perception of the diffraction degeneration is designed to solve a sparsity-constrained inverse problem, realizing the volume reconstruction from 2D measurements with Large amount of aliasing. Our system is evaluated by elaborating the imaging optical theory and reconstruction algorithm with demonstrating the experimental imaging under a single exposure. Ultimately, we achieve the sub-super-pixel spatial resolution and high spectral resolution imaging. The code will be available at: https://github.com/Krito-ex/CSST.
Thanks to the large pre-trained vision-language models (VLMs) like CLIP, we can craft a zero-shot classifier by discrete prompt design, e.g., the confidence score of an image being "[CLASS]" can be obtained by using the VLM provided similarity between the image and the prompt sentence "a photo of a [CLASS]". Furthermore, prompting shows great potential for fast adaptation of VLMs to downstream tasks if we fine-tune the soft prompts with few samples. However, we find a common failure that improper fine-tuning or learning with extremely few-shot samples may even under-perform the zero-shot prediction. Existing methods still address this problem by using traditional anti-overfitting techniques such as early stopping and data augmentation, which lack a principled solution specific to prompting. In this paper, we present Prompt-aligned Gradient, dubbed ProGrad to prevent prompt tuning from forgetting the general knowledge learned from VLMs. In particular, ProGrad only updates the prompt whose gradient is aligned (or non-conflicting) to the general knowledge, which is represented as the optimization direction offered by the predefined prompt predictions. Extensive experiments under the few-shot learning, domain generalization, base-to-new generalization and cross-dataset transfer settings demonstrate the stronger few-shot generalization ability of ProGrad over state-of-the-art prompt tuning methods. Codes and theoretical proof are in Appendix.
Interactions between humans are diverse and context-dependent, but previous works have treated them as categorical, disregarding the heavy tail of possible interactions. We propose a new paradigm of learning human-human interactions as free text from a single still image, allowing for flexibility in modeling the unlimited space of situations and relationships between people. To overcome the absence of data labelled specifically for this task, we use knowledge distillation applied to synthetic caption data produced by a large language model without explicit supervision. We show that the pseudo-labels produced by this procedure can be used to train a captioning model to effectively understand human-human interactions in images, as measured by a variety of metrics that measure textual and semantic faithfulness and factual groundedness of our predictions. We further show that our approach outperforms SOTA image captioning and situation recognition models on this task. We will release our code and pseudo-labels along with Waldo and Wenda, a manually-curated test set for still image human-human interaction understanding.
The field of visual representation learning has seen explosive growth in the past years, but its benefits in robotics have been surprisingly limited so far. Prior work uses generic visual representations as a basis to learn (task-specific) robot action policies (e.g., via behavior cloning). While the visual representations do accelerate learning, they are primarily used to encode visual observations. Thus, action information has to be derived purely from robot data, which is expensive to collect! In this work, we present a scalable alternative where the visual representations can help directly infer robot actions. We observe that vision encoders express relationships between image observations as distances (e.g., via embedding dot product) that could be used to efficiently plan robot behavior. We operationalize this insight and develop a simple algorithm for acquiring a distance function and dynamics predictor, by fine-tuning a pre-trained representation on human collected video sequences. The final method is able to substantially outperform traditional robot learning baselines (e.g., 70% success v.s. 50% for behavior cloning on pick-place) on a suite of diverse real-world manipulation tasks. It can also generalize to novel objects, without using any robot demonstrations during train time. For visualizations of the learned policies please check: https://agi-labs.github.io/manipulate-by-seeing/
Surrogate gradient (SG) is one of the most effective approaches for training spiking neural networks (SNNs). While assisting SNNs to achieve classification performance comparable to artificial neural networks, SG suffers from the problem of time-consuming training, preventing it from efficient learning. In this paper, we formally analyze the backward process of classic SG and find that the membrane accumulation through time leads to exponential growth of training time. With this discovery, we propose Stabilized Spiking Flow (SSF), a simple yet effective approach to accelerate training of SG-based SNNs. For each spiking neuron, SSF averages its input and output activations over time to yield stabilized input and output, respectively. Then, instead of back propagating all errors that are related to current neuron and inherently entangled in time domain, the auxiliary gradient is directly propagated from the stabilized output to input through a devised relationship mapping. Additionally, SSF method is suitable to different neuron models. Extensive experiments on both static and neuromorphic datasets demonstrate that SNNs trained with SSF approach can achieve performance comparable to the original counterparts, while reducing the training time significantly. In particular, SSF speeds up the training process of state-of-the-art SNN models up to 10x when time steps equal to 80.
Knowledge-Spreader: Learning Semi-Supervised Facial Action Dynamics by Consistifying Knowledge Granularity
PDF ↗Recent studies on dynamic facial action unit (AU) detection have extensively relied on dense annotations. However, manual annotations are difficult, time-consuming, and costly. The canonical semi-supervised learning (SSL) methods ignore the consistency, extensibility, and adaptability of structural knowledge across spatial-temporal domains. Furthermore, the reliance on offline design and excessive parameters hinder the efficiency of the learning process. To remedy these issues, we propose a lightweight and on-line semi-supervised framework, a so-called Knowledge-Spreader (KS), to learn AU dynamics with sparse annotations. By formulating SSL as a Progressive Knowledge Distillation (PKD) problem, we aim to infer cross-domain information, specifically from spatial to temporal domains, by consistifying knowledge granularity within Teacher-Students Network. Specifically, KS employs sparsely annotated key-frames to learn AU dependencies as the privileged knowledge. Then, the model spreads the learned knowledge to their unlabeled neighbours by jointly applying knowledge distillation and pseudo-labeling, and completes the temporal information as the expanded knowledge. We term the progressive knowledge distillation as "Knowledge Spreading", which allows our model to learn spatial-temporal knowledge from video clips with only one label allocated. Extensive experiments demonstrate that KS achieves competitive performance as compared to the state of the arts under the circumstances of using only 2% labels on BP4D and 5% labels on DISFA. In addition, we have tested it on our newly developed large-scale comprehensive emotion database BP4D++, which contains considerable samples across well-synchronized and aligned sensor modalities for alleviating the scarcity issue of annotations and identities.
Tangent Model Composition (TMC) is a method to combine component models independently fine-tuned around a pre-trained point. Component models are tangent vectors to the pre-trained model that can be added, scaled, or subtracted to support incremental learning, ensembling, or unlearning. Component models are composed at inference time via scalar combination, reducing the cost of ensembling to that of a single model. TMC improves accuracy by 4.2% compared to ensembling non-linearly fine-tuned models at a 2.5x to 10x reduction of inference cost, growing linearly with the number of component models. Each component model can be forgotten at zero cost, with no residual effect on the resulting inference. When used for continual fine-tuning, TMC is not constrained by sequential bias and can be executed in parallel on federated data. TMC outperforms recently published continual fine-tuning methods almost uniformly on each setting -- task-incremental, class-incremental, and data-incremental -- on a total of 13 experiments across 3 benchmark datasets, despite not using any replay buffer. TMC is designed for composing models that are local to a pre-trained embedding, but could be extended to more general settings.
Multi-object tracking (MOT) at low frame rates can reduce computational, storage and power overhead to better meet the constraints of edge devices. Many existing MOT methods suffer from significant performance degradation in low-frame-rate videos due to significant location and appearance changes between adjacent frames. To this end, we propose to explore collaborative tracking learning (CoTracker) for frame-rate-insensitive MOT in a query-based end-to-end manner. Multiple historical queries of the same target jointly track it with richer temporal descriptions. Meanwhile, we insert an information refinement module between every two temporal blocking decoders to better fuse temporal clues and refine features. Moreover, a tracking object consistency loss is proposed to guide the interaction between historical queries. Extensive experimental results demonstrate that in high-frame-rate videos, CoTracker obtains higher performance than state-of-the-art methods on large-scale datasets Dancetrack and BDD100K, and outperforms the existing end-to-end methods on MOT17. More importantly, CoTracker has a significant advantage over state-of-the-art methods in low-frame-rate videos, which allows it to obtain faster processing speeds by reducing frame-rate requirements while maintaining higher performance. Code will be released at https://github.com/yolomax/ColTrack
Mitochondria, as tiny structures within the cell, are of significant importance to study cell functions for biological and clinical analysis. And exploring how to automatically segment mitochondria in electron microscopy (EM) images has attracted increasing attention. However, most of existing methods struggle to adapt to different scales and appearances of the input due to the inherent limitations of the traditional CNN architecture. To mitigate these limitations, we propose a novel adaptive template transformer (ATFormer) for mitochondria segmentation. The proposed ATFormer model enjoys several merits. First, the designed structural template learning module can acquire appearance-adaptive templates of background, foreground and contour to sense the characteristics of different shapes of mitochondria. And we further adopt an optimal transport algorithm to enlarge the discrepancy among diverse templates to fully activate corresponding regions. Second, we introduce a hierarchical attention learning mechanism to absorb multi-level information for templates to be adaptive scale-aware classifiers for dense prediction. Extensive experimental results on three challenging benchmarks including MitoEM, Lucchi and NucMM-Z datasets demonstrate that our ATFormer performs favorably against state-of-the-art mitochondria segmentation methods.
Task-Free Continual Learning (TFCL) aims to learn new concepts from a stream of data without any task information. The Dynamic Expansion Model (DEM) has shown promising results in TFCL by dynamically expanding the model's capacity to deal with shifts in the data distribution. However, existing approaches only consider the recognition of the input shift as the expansion signal and ignore the correlation between the newly incoming data and previously learned knowledge, resulting in adding and training unnecessary parameters. In this paper, we propose a novel and effective framework for TFCL, which dynamically expands the architecture of a DEM model through a self-assessment mechanism evaluating the diversity of knowledge among existing experts as expansion signals. This mechanism ensures learning additional underlying data distributions with a compact model structure. A novelty-aware sample selection approach is proposed to manage the memory buffer that forces the newly added expert to learn novel information from a data stream, which further promotes the diversity among experts. Moreover, we also propose to reuse previously learned representation information for learning new incoming data by using knowledge transfer in TFCL, which has not been explored before. The DEM expansion and training are regularized through a gradient updating mechanism to gradually explore the positive forward transfer, further improving the performance. Empirical results on TFCL benchmarks show that the proposed framework outperforms the state-of-the-art while using a reasonable number of parameters. The code is available at https://github.com/dtuzi123/SEDEM/.
We propose an Unsupervised Domain Adaptation (UDA) method by making use of Energy-Based Learning (EBL) and demonstrate 1. EBL can be used to improve the instance selection for a self-training task on the unlabelled target domain, and 2. alignment and normalizing energy scores can learn domain-invariant representations. For the former, we show that an energy-based selection criterion can be used to model instance selections by mimicking the joint distribution between data and predictions in the target domain. As per learning domain invariant representations, we show that stable domain alignment can be achieved by a combined energy alignment and an energy normalization process. We implement our method in consistent with the vision-transformer (ViT) backbone and empirically show that our proposed method can outperform state-of-the-art ViT based UDA methods on diverse benchmarks (DomainNet, OfficeHome, and VISDA2017).