Parameter-Efficient Fine-Tuning for Continual Learning: A Neural Tangent Kernel Perspective

Parameter-efficient fine-tuning for continual learning (PEFT-CL) has shown promise in adapting pre-trained models to sequential tasks while mitigating catastrophic forgetting problem. However, understanding the mechanisms that dictate continual performance in this paradigm remains elusive. To tackle this complexity, we undertake a rigorous analysis of PEFT-CL dynamics to derive relevant metrics for continual scenarios using Neural Tangent Kernel (NTK) theory. With the aid of NTK as a mathematical analysis tool, we recast the challenge of test-time forgetting into the quantifiable generalization gaps during training, identifying three key factors that influence these gaps and the performance of PEFT-CL: training sample size, task-level feature orthogonality, and regularization. To address these challenges, we introduce NTK-CL, a novel framework that eliminates task-specific parameter storage while adaptively generating task-relevant features. Aligning with theoretical guidance, NTK-CL triples the feature representation of each sample, theoretically and empirically reducing the magnitude of both task-interplay and task-specific generalization gaps. Grounded in NTK analysis, our approach imposes an adaptive exponential moving average mechanism and constraints on task-level feature orthogonality, maintaining intra-task NTK forms while attenuating inter-task NTK forms. Ultimately, by fine-tuning optimizable parameters with appropriate regularization, NTK-CL achieves state-of-the-art performance on established PEFT-CL benchmarks. This work provides a theoretical foundation for understanding and improving PEFT-CL models, offering insights into the interplay between feature representation, task orthogonality, and generalization, contributing to the development of more efficient continual learning systems.

PVUW 2024 Challenge on Complex Video Understanding: Methods and Results

Pixel-level Video Understanding in the Wild Challenge (PVUW) focus on complex video understanding. In this CVPR 2024 workshop, we add two new tracks, Complex Video Object Segmentation Track based on MOSE dataset and Motion Expression guided Video Segmentation track based on MeViS dataset. In the two new tracks, we provide additional videos and annotations that feature challenging elements, such as the disappearance and reappearance of objects, inconspicuous small objects, heavy occlusions, and crowded environments in MOSE. Moreover, we provide a new motion expression guided video segmentation dataset MeViS to study the natural language-guided video understanding in complex environments. These new videos, sentences, and annotations enable us to foster the development of a more comprehensive and robust pixel-level understanding of video scenes in complex environments and realistic scenarios. The MOSE challenge had 140 registered teams in total, 65 teams participated the validation phase and 12 teams made valid submissions in the final challenge phase. The MeViS challenge had 225 registered teams in total, 50 teams participated the validation phase and 5 teams made valid submissions in the final challenge phase.

1st Place Solution for MeViS Track in CVPR 2024 PVUW Workshop: Motion Expression guided Video Segmentation

Motion Expression guided Video Segmentation (MeViS), as an emerging task, poses many new challenges to the field of referring video object segmentation (RVOS). In this technical report, we investigated and validated the effectiveness of static-dominant data and frame sampling on this challenging setting. Our solution achieves a J&F score of 0.5447 in the competition phase and ranks 1st in the MeViS track of the PVUW Challenge. The code is available at: https://github.com/Tapall-AI/MeViS_Track_Solution_2024.

VIP: Versatile Image Outpainting Empowered by Multimodal Large Language Model

In this paper, we focus on resolving the problem of image outpainting, which aims to extrapolate the surrounding parts given the center contents of an image. Although recent works have achieved promising performance, the lack of versatility and customization hinders their practical applications in broader scenarios. Therefore, this work presents a novel image outpainting framework that is capable of customizing the results according to the requirement of users. First of all, we take advantage of a Multimodal Large Language Model (MLLM) that automatically extracts and organizes the corresponding textual descriptions of the masked and unmasked part of a given image. Accordingly, the obtained text prompts are introduced to endow our model with the capacity to customize the outpainting results. In addition, a special Cross-Attention module, namely Center-Total-Surrounding (CTS), is elaborately designed to enhance further the the interaction between specific space regions of the image and corresponding parts of the text prompts. Note that unlike most existing methods, our approach is very resource-efficient since it is just slightly fine-tuned on the off-the-shelf stable diffusion (SD) model rather than being trained from scratch. Finally, the experimental results on three commonly used datasets, i.e. Scenery, Building, and WikiArt, demonstrate our model significantly surpasses the SoTA methods. Moreover, versatile outpainting results are listed to show its customized ability.

YOLOv10: Real-Time End-to-End Object Detection

Over the past years, YOLOs have emerged as the predominant paradigm in the field of real-time object detection owing to their effective balance between computational cost and detection performance. Researchers have explored the architectural designs, optimization objectives, data augmentation strategies, and others for YOLOs, achieving notable progress. However, the reliance on the non-maximum suppression (NMS) for post-processing hampers the end-to-end deployment of YOLOs and adversely impacts the inference latency. Besides, the design of various components in YOLOs lacks the comprehensive and thorough inspection, resulting in noticeable computational redundancy and limiting the model's capability. It renders the suboptimal efficiency, along with considerable potential for performance improvements. In this work, we aim to further advance the performance-efficiency boundary of YOLOs from both the post-processing and model architecture. To this end, we first present the consistent dual assignments for NMS-free training of YOLOs, which brings competitive performance and low inference latency simultaneously. Moreover, we introduce the holistic efficiency-accuracy driven model design strategy for YOLOs. We comprehensively optimize various components of YOLOs from both efficiency and accuracy perspectives, which greatly reduces the computational overhead and enhances the capability. The outcome of our effort is a new generation of YOLO series for real-time end-to-end object detection, dubbed YOLOv10. Extensive experiments show that YOLOv10 achieves state-of-the-art performance and efficiency across various model scales. For example, our YOLOv10-S is 1.8$\times$ faster than RT-DETR-R18 under the similar AP on COCO, meanwhile enjoying 2.8$\times$ smaller number of parameters and FLOPs. Compared with YOLOv9-C, YOLOv10-B has 46\% less latency and 25\% fewer parameters for the same performance.

Modality Prompts for Arbitrary Modality Salient Object Detection

This paper delves into the task of arbitrary modality salient object detection (AM SOD), aiming to detect salient objects from arbitrary modalities, eg RGB images, RGB-D images, and RGB-D-T images. A novel modality-adaptive Transformer (MAT) will be proposed to investigate two fundamental challenges of AM SOD, ie more diverse modality discrepancies caused by varying modality types that need to be processed, and dynamic fusion design caused by an uncertain number of modalities present in the inputs of multimodal fusion strategy. Specifically, inspired by prompt learning's ability of aligning the distributions of pre-trained models to the characteristic of downstream tasks by learning some prompts, MAT will first present a modality-adaptive feature extractor (MAFE) to tackle the diverse modality discrepancies by introducing a modality prompt for each modality. In the training stage, a new modality translation contractive (MTC) loss will be further designed to assist MAFE in learning those modality-distinguishable modality prompts. Accordingly, in the testing stage, MAFE can employ those learned modality prompts to adaptively adjust its feature space according to the characteristics of the input modalities, thus being able to extract discriminative unimodal features. Then, MAFE will present a channel-wise and spatial-wise fusion hybrid (CSFH) strategy to meet the demand for dynamic fusion. For that, CSFH dedicates a channel-wise dynamic fusion module (CDFM) and a novel spatial-wise dynamic fusion module (SDFM) to fuse the unimodal features from varying numbers of modalities and meanwhile effectively capture cross-modal complementary semantic and detail information, respectively. Moreover, CSFH will carefully align CDFM and SDFM to different levels of unimodal features based on their characteristics for more effective complementary information exploitation.

Salient Object Detection From Arbitrary Modalities

Toward desirable saliency prediction, the types and numbers of inputs for a salient object detection (SOD) algorithm may dynamically change in many real-life applications. However, existing SOD algorithms are mainly designed or trained for one particular type of inputs, failing to be generalized to other types of inputs. Consequentially, more types of SOD algorithms need to be prepared in advance for handling different types of inputs, raising huge hardware and research costs. Differently, in this paper, we propose a new type of SOD task, termed Arbitrary Modality SOD (AM SOD). The most prominent characteristics of AM SOD are that the modality types and modality numbers will be arbitrary or dynamically changed. The former means that the inputs to the AM SOD algorithm may be arbitrary modalities such as RGB, depths, or even any combination of them. While, the latter indicates that the inputs may have arbitrary modality numbers as the input type is changed, e.g. single-modality RGB image, dual-modality RGB-Depth (RGB-D) images or triple-modality RGB-Depth-Thermal (RGB-D-T) images. Accordingly, a preliminary solution to the above challenges, \i.e. a modality switch network (MSN), is proposed in this paper. In particular, a modality switch feature extractor (MSFE) is first designed to extract discriminative features from each modality effectively by introducing some modality indicators, which will generate some weights for modality switching. Subsequently, a dynamic fusion module (DFM) is proposed to adaptively fuse features from a variable number of modalities based on a novel Transformer structure. Finally, a new dataset, named AM-XD, is constructed to facilitate research on AM SOD. Extensive experiments demonstrate that our AM SOD method can effectively cope with changes in the type and number of input modalities for robust salient object detection.

Raformer: Redundancy-Aware Transformer for Video Wire Inpainting

Video Wire Inpainting (VWI) is a prominent application in video inpainting, aimed at flawlessly removing wires in films or TV series, offering significant time and labor savings compared to manual frame-by-frame removal. However, wire removal poses greater challenges due to the wires being longer and slimmer than objects typically targeted in general video inpainting tasks, and often intersecting with people and background objects irregularly, which adds complexity to the inpainting process. Recognizing the limitations posed by existing video wire datasets, which are characterized by their small size, poor quality, and limited variety of scenes, we introduce a new VWI dataset with a novel mask generation strategy, namely Wire Removal Video Dataset 2 (WRV2) and Pseudo Wire-Shaped (PWS) Masks. WRV2 dataset comprises over 4,000 videos with an average length of 80 frames, designed to facilitate the development and efficacy of inpainting models. Building upon this, our research proposes the Redundancy-Aware Transformer (Raformer) method that addresses the unique challenges of wire removal in video inpainting. Unlike conventional approaches that indiscriminately process all frame patches, Raformer employs a novel strategy to selectively bypass redundant parts, such as static background segments devoid of valuable information for inpainting. At the core of Raformer is the Redundancy-Aware Attention (RAA) module, which isolates and accentuates essential content through a coarse-grained, window-based attention mechanism. This is complemented by a Soft Feature Alignment (SFA) module, which refines these features and achieves end-to-end feature alignment. Extensive experiments on both the traditional video inpainting datasets and our proposed WRV2 dataset demonstrate that Raformer outperforms other state-of-the-art methods.

On Exploring PDE Modeling for Point Cloud Video Representation Learning

Point cloud video representation learning is challenging due to complex structures and unordered spatial arrangement. Traditional methods struggle with frame-to-frame correlations and point-wise correspondence tracking. Recently, partial differential equations (PDE) have provided a new perspective in uniformly solving spatial-temporal data information within certain constraints. While tracking tangible point correspondence remains challenging, we propose to formalize point cloud video representation learning as a PDE-solving problem. Inspired by fluid analysis, where PDEs are used to solve the deformation of spatial shape over time, we employ PDE to solve the variations of spatial points affected by temporal information. By modeling spatial-temporal correlations, we aim to regularize spatial variations with temporal features, thereby enhancing representation learning in point cloud videos. We introduce Motion PointNet composed of a PointNet-like encoder and a PDE-solving module. Initially, we construct a lightweight yet effective encoder to model an initial state of the spatial variations. Subsequently, we develop our PDE-solving module in a parameterized latent space, tailored to address the spatio-temporal correlations inherent in point cloud video. The process of solving PDE is guided and refined by a contrastive learning structure, which is pivotal in reshaping the feature distribution, thereby optimizing the feature representation within point cloud video data. Remarkably, our Motion PointNet achieves an impressive accuracy of 97.52% on the MSRAction-3D dataset, surpassing the current state-of-the-art in all aspects while consuming minimal resources (only 0.72M parameters and 0.82G FLOPs).

WaveFace: Authentic Face Restoration with Efficient Frequency Recovery

Although diffusion models are rising as a powerful solution for blind face restoration, they are criticized for two problems: 1) slow training and inference speed, and 2) failure in preserving identity and recovering fine-grained facial details. In this work, we propose WaveFace to solve the problems in the frequency domain, where low- and high-frequency components decomposed by wavelet transformation are considered individually to maximize authenticity as well as efficiency. The diffusion model is applied to recover the low-frequency component only, which presents general information of the original image but 1/16 in size. To preserve the original identity, the generation is conditioned on the low-frequency component of low-quality images at each denoising step. Meanwhile, high-frequency components at multiple decomposition levels are handled by a unified network, which recovers complex facial details in a single step. Evaluations on four benchmark datasets show that: 1) WaveFace outperforms state-of-the-art methods in authenticity, especially in terms of identity preservation, and 2) authentic images are restored with the efficiency 10x faster than existing diffusion model-based BFR methods.