Search

Objective:To develop a no-reference image quality assessment method using automated distortion recognition to boost MRI-guided radiotherapy precision.Methods:We analyzed 106,000 MR images from 10 patients with liver metastasis,captured with the Elekta Unity MR-LINAC.Our No-Reference Quality Assessment Model includes:1)image preprocessing to enhance visibility of key diagnostic features;2)feature extraction and directional analysis using MSCN coefficients across four directions to capture textural attributes and gradients,vital for identifying image features and potential distortions;3)integrative Quality Index(QI)calculation,which integrates features via AGGD parameter estimation and K-means clustering.The QI,based on a weighted MAD computation of directional scores,provides a comprehensive image quality measure,robust against outliers.LOO-CV assessed model generalizability and performance.Tumor tracking algorithm performance was compared with and without preprocessing to verify tracking accuracy enhancements.Results:Preprocessing significantly improved image quality,with the QI showing substantial positive changes and surpassing other metrics.After normalization,the QI's average value was 79.6 times higher than CNR,indicating improved image definition and contrast.It also showed higher sensitivity in detail recognition with average values 6.5 times and 1.7 times higher than Tenengrad gradient and entropy.The tumor tracking algorithm confirmed significant tracking accuracy improvements with preprocessed images,validating preprocessing effectiveness.Conclusions:This study introduces a novel no-reference image quality evaluation method based on automated distortion recognition,offering a new quality control tool for MRIgRT tumor tracking.It enhances clinical application accuracy and facilitates medical image quality assessment standardization, with significant clinical and research value.

Generating 3D meshes from a single image is an important but ill-posed task. Existing methods mainly adopt 2D multiview diffusion models to generate intermediate multiview images, and use the Large Reconstruction Model (LRM) to create the final meshes. However, the multiview images exhibit local inconsistencies, and the meshes often lack fidelity to the input image or look blurry. We propose Fancy123, featuring two enhancement modules and an unprojection operation to address the above three issues, respectively. The appearance enhancement module deforms the 2D multiview images to realign misaligned pixels for better multiview consistency. The fidelity enhancement module deforms the 3D mesh to match the input image. The unprojection of the input image and deformed multiview images onto LRM's generated mesh ensures high clarity, discarding LRM's predicted blurry-looking mesh colors. Extensive qualitative and quantitative experiments verify Fancy123's SoTA performance with significant improvement. Also, the two enhancement modules are plug-and-play and work at inference time, allowing seamless integration into various existing single-image-to-3D methods., Comment: Project page: https://github.com/YuQiao0303/Fancy123

Background and Purpose: Accurate motion tracking in MRI-guided Radiotherapy (MRIgRT) is essential for effective treatment delivery. This study aimed to enhance motion tracking precision in MRIgRT through an automatic real-time markerless tracking method using an enhanced Tracking-Learning-Detection (ETLD) framework with automatic segmentation. Materials and Methods: We developed a novel MRIgRT motion tracking and segmentation method by integrating the ETLD framework with an improved Chan-Vese model (ICV), named ETLD+ICV. The ETLD framework was upgraded for real-time cine MRI, including advanced image preprocessing, no-reference image quality assessment, an enhanced median-flow tracker, and a refined detector with dynamic search region adjustments. ICV was used for precise target volume coverage, refining the segmented region frame by frame using tracking results, with key parameters optimized. The method was tested on 3.5D MRI scans from 10 patients with liver metastases. Results: Evaluation of 106,000 frames across 77 treatment fractions showed sub-millimeter tracking errors of less than 0.8mm, with over 99% precision and 98% recall for all subjects in the Beam Eye View(BEV)/Beam Path View(BPV) orientation. The ETLD+ICV method achieved a dice global score of more than 82% for all subjects, demonstrating the method's extensibility and precise target volume coverage. Conclusion: This study successfully developed an automatic real-time markerless motion tracking method for MRIgRT that significantly outperforms current methods. The novel method not only delivers exceptional precision in tracking and segmentation but also shows enhanced adaptability to clinical demands, making it an indispensable asset in improving the efficacy of radiotherapy treatments.

Underwater images often suffer from quality degradation due to absorption and scattering effects. Most existing underwater image enhancement algorithms produce a single, fixed-color image, limiting user flexibility and application. To address this limitation, we propose a method called \textit{ColorCode}, which enhances underwater images while offering a range of controllable color outputs. Our approach involves recovering an underwater image to a reference enhanced image through supervised training and decomposing it into color and content codes via self-reconstruction and cross-reconstruction. The color code is explicitly constrained to follow a Gaussian distribution, allowing for efficient sampling and interpolation during inference. ColorCode offers three key features: 1) color enhancement, producing an enhanced image with a fixed color; 2) color adaptation, enabling controllable adjustments of long-wavelength color components using guidance images; and 3) color interpolation, allowing for the smooth generation of multiple colors through continuous sampling of the color code. Quantitative and visual evaluations on popular and challenging benchmark datasets demonstrate the superiority of ColorCode over existing methods in providing diverse, controllable, and color-realistic enhancement results. The source code is available at https://github.com/Xiaofeng-life/ColorCode.

Adversarial training has achieved remarkable advancements in defending against adversarial attacks. Among them, fast adversarial training (FAT) is gaining attention for its ability to achieve competitive robustness with fewer computing resources. Existing FAT methods typically employ a uniform strategy that optimizes all training data equally without considering the influence of different examples, which leads to an imbalanced optimization. However, this imbalance remains unexplored in the field of FAT. In this paper, we conduct a comprehensive study of the imbalance issue in FAT and observe an obvious class disparity regarding their performances. This disparity could be embodied from a perspective of alignment between clean and robust accuracy. Based on the analysis, we mainly attribute the observed misalignment and disparity to the imbalanced optimization in FAT, which motivates us to optimize different training data adaptively to enhance robustness. Specifically, we take disparity and misalignment into consideration. First, we introduce self-knowledge guided regularization, which assigns differentiated regularization weights to each class based on its training state, alleviating class disparity. Additionally, we propose self-knowledge guided label relaxation, which adjusts label relaxation according to the training accuracy, alleviating the misalignment and improving robustness. By combining these methods, we formulate the Self-Knowledge Guided FAT (SKG-FAT), leveraging naturally generated knowledge during training to enhance the adversarial robustness without compromising training efficiency. Extensive experiments on four standard datasets demonstrate that the SKG-FAT improves the robustness and preserves competitive clean accuracy, outperforming the state-of-the-art methods., Comment: 13 pages

As a point defect with unique spin and optical properties, nitrogen-vacancy (NV) center in diamond has attracted much attention in the fields of quantum sensing, quantum simulation, and quantum networks. The optical properties of an NV center are crucial for all these quantum applications. However, NV centers fabricated by destructive methods such as electron irradiation or ion implantation usually exhibit poor optical coherence. In this work, we demonstrate a non-destructive method to fabricate optically coherent NV centers. High-purity single crystal diamonds are annealed under high pressure and high temperature (1700 $^{\circ}$C, 5.5 GPa), and individually resolvable NV centers with narrow PLE linewidth (<100 MHz) are produced. The high-pressure condition prevents the conversion of diamond to graphite during high-temperature annealing, significantly expanding the parameter space for creating high-performance artificial defects for quantum information science. These findings deepen our understanding of NV center formation in diamond and have implications for the optimization of color centers in solids, including silicon carbide and hexagonal boron nitride., Comment: 11 pages,4 figures

Finger vein recognition technology has become one of the primary solutions for high-security identification systems. However, it still has information leakage problems, which seriously jeopardizes users privacy and anonymity and cause great security risks. In addition, there is no work to consider a fully integrated secure finger vein recognition system. So, different from the previous systems, we integrate preprocessing and template protection into an integrated deep learning model. We propose an end-to-end cancelable finger vein network (CFVNet), which can be used to design an secure finger vein recognition system.It includes a plug-and-play BWR-ROIAlign unit, which consists of three sub-modules: Localization, Compression and Transformation. The localization module achieves automated localization of stable and unique finger vein ROI. The compression module losslessly removes spatial and channel redundancies. The transformation module uses the proposed BWR method to introduce unlinkability, irreversibility and revocability to the system. BWR-ROIAlign can directly plug into the model to introduce the above features for DCNN-based finger vein recognition systems. We perform extensive experiments on four public datasets to study the performance and cancelable biometric attributes of the CFVNet-based recognition system. The average accuracy, EERs and Dsys on the four datasets are 99.82%, 0.01% and 0.025, respectively, and achieves competitive performance compared with the state-of-the-arts.

Learning-based methods for underwater image enhancement (UWIE) have undergone extensive exploration. However, learning-based models are usually vulnerable to adversarial examples so as the UWIE models. To the best of our knowledge, there is no comprehensive study on the adversarial robustness of UWIE models, which indicates that UWIE models are potentially under the threat of adversarial attacks. In this paper, we propose a general adversarial attack protocol. We make a first attempt to conduct adversarial attacks on five well-designed UWIE models on three common underwater image benchmark datasets. Considering the scattering and absorption of light in the underwater environment, there exists a strong correlation between color correction and underwater image enhancement. On the basis of that, we also design two effective UWIE-oriented adversarial attack methods Pixel Attack and Color Shift Attack targeting different color spaces. The results show that five models exhibit varying degrees of vulnerability to adversarial attacks and well-designed small perturbations on degraded images are capable of preventing UWIE models from generating enhanced results. Further, we conduct adversarial training on these models and successfully mitigated the effectiveness of adversarial attacks. In summary, we reveal the adversarial vulnerability of UWIE models and propose a new evaluation dimension of UWIE models.

Enabling Large Language Models (LLMs) to comprehend the 3D physical world remains a significant challenge. Due to the lack of large-scale 3D-text pair datasets, the success of LLMs has yet to be replicated in 3D understanding. In this paper, we rethink this issue and propose a new task: 3D Data-Efficient Point-Language Understanding. The goal is to enable LLMs to achieve robust 3D object understanding with minimal 3D point cloud and text data pairs. To address this task, we introduce GreenPLM, which leverages more text data to compensate for the lack of 3D data. First, inspired by using CLIP to align images and text, we utilize a pre-trained point cloud-text encoder to map the 3D point cloud space to the text space. This mapping leaves us to seamlessly connect the text space with LLMs. Once the point-text-LLM connection is established, we further enhance text-LLM alignment by expanding the intermediate text space, thereby reducing the reliance on 3D point cloud data. Specifically, we generate 6M free-text descriptions of 3D objects, and design a three-stage training strategy to help LLMs better explore the intrinsic connections between different modalities. To achieve efficient modality alignment, we design a zero-parameter cross-attention module for token pooling. Extensive experimental results show that GreenPLM requires only 12% of the 3D training data used by existing state-of-the-art models to achieve superior 3D understanding. Remarkably, GreenPLM also achieves competitive performance using text-only data. The code and weights are available at: https://github.com/TangYuan96/GreenPLM.

While adversarial training is an effective defense method against adversarial attacks, it notably increases the training cost. To this end, fast adversarial training (FAT) is presented for efficient training and has become a hot research topic. However, FAT suffers from catastrophic overfitting, which leads to a performance drop compared with multi-step adversarial training. However, the cause of catastrophic overfitting remains unclear and lacks exploration. In this paper, we present an example taxonomy in FAT, which identifies that catastrophic overfitting is caused by the imbalance between the inner and outer optimization in FAT. Furthermore, we investigated the impact of varying degrees of training loss, revealing a correlation between training loss and catastrophic overfitting. Based on these observations, we redesign the loss function in FAT with the proposed dynamic label relaxation to concentrate the loss range and reduce the impact of misclassified examples. Meanwhile, we introduce batch momentum initialization to enhance the diversity to prevent catastrophic overfitting in an efficient manner. Furthermore, we also propose Catastrophic Overfitting aware Loss Adaptation (COLA), which employs a separate training strategy for examples based on their loss degree. Our proposed method, named example taxonomy aware FAT (ETA), establishes an improved paradigm for FAT. Experiment results demonstrate our ETA achieves state-of-the-art performance. Comprehensive experiments on four standard datasets demonstrate the competitiveness of our proposed method., Comment: 15 pages

Reconstructing textured meshes from colored point clouds is an important but challenging task. Most existing methods yield blurry-looking textures or rely on 3D training data that are hard to acquire. Regarding this, we propose PointDreamer, a novel framework for textured mesh reconstruction from colored point cloud via diffusion-based 2D inpainting. Specifically, we first reconstruct an untextured mesh. Next, we project the input point cloud into 2D space to generate sparse multi-view images, and then inpaint empty pixels utilizing a pre-trained 2D diffusion model. After that, we unproject the colors of the inpainted dense images onto the untextured mesh, thus obtaining the final textured mesh. This project-inpaint-unproject pipeline bridges the gap between 3D point clouds and 2D diffusion models for the first time. Thanks to the powerful 2D diffusion model pre-trained on extensive 2D data, PointDreamer reconstructs clear, high-quality textures with high robustness to sparse or noisy input. Also, it's zero-shot requiring no extra training. In addition, we design Non-Border-First unprojection strategy to address the border-area inconsistency issue, which is less explored but commonly-occurred in methods that generate 3D textures from multiview images. Extensive qualitative and quantitative experiments on various synthetic and real-scanned datasets show the SoTA performance of PointDreamer, by significantly outperforming baseline methods with 30% improvement in LPIPS score (from 0.118 to 0.068). Code at: https://github.com/YuQiao0303/PointDreamer.

Large 2D vision-language models (2D-LLMs) have gained significant attention by bridging Large Language Models (LLMs) with images using a simple projector. Inspired by their success, large 3D point cloud-language models (3D-LLMs) also integrate point clouds into LLMs. However, directly aligning point clouds with LLM requires expensive training costs, typically in hundreds of GPU-hours on A100, which hinders the development of 3D-LLMs. In this paper, we introduce MiniGPT-3D, an efficient and powerful 3D-LLM that achieves multiple SOTA results while training for only 27 hours on one RTX 3090. Specifically, we propose to align 3D point clouds with LLMs using 2D priors from 2D-LLMs, which can leverage the similarity between 2D and 3D visual information. We introduce a novel four-stage training strategy for modality alignment in a cascaded way, and a mixture of query experts module to adaptively aggregate features with high efficiency. Moreover, we utilize parameter-efficient fine-tuning methods LoRA and Norm fine-tuning, resulting in only 47.8M learnable parameters, which is up to 260x fewer than existing methods. Extensive experiments show that MiniGPT-3D achieves SOTA on 3D object classification and captioning tasks, with significantly cheaper training costs. Notably, MiniGPT-3D gains an 8.12 increase on GPT-4 evaluation score for the challenging object captioning task compared to ShapeLLM-13B, while the latter costs 160 total GPU-hours on 8 A800. We are the first to explore the efficient 3D-LLM, offering new insights to the community. Code and weights are available at https://github.com/TangYuan96/MiniGPT-3D., Comment: 17 pages, 9 figures

Existing Transformer-based models for point cloud analysis suffer from quadratic complexity, leading to compromised point cloud resolution and information loss. In contrast, the newly proposed Mamba model, based on state space models (SSM), outperforms Transformer in multiple areas with only linear complexity. However, the straightforward adoption of Mamba does not achieve satisfactory performance on point cloud tasks. In this work, we present Mamba3D, a state space model tailored for point cloud learning to enhance local feature extraction, achieving superior performance, high efficiency, and scalability potential. Specifically, we propose a simple yet effective Local Norm Pooling (LNP) block to extract local geometric features. Additionally, to obtain better global features, we introduce a bidirectional SSM (bi-SSM) with both a token forward SSM and a novel backward SSM that operates on the feature channel. Extensive experimental results show that Mamba3D surpasses Transformer-based counterparts and concurrent works in multiple tasks, with or without pre-training. Notably, Mamba3D achieves multiple SoTA, including an overall accuracy of 92.6% (train from scratch) on the ScanObjectNN and 95.1% (with single-modal pre-training) on the ModelNet40 classification task, with only linear complexity. Our code and weights are available at https://github.com/xhanxu/Mamba3D., Comment: ACM MM 2024. Code and weights are available at https://github.com/xhanxu/Mamba3D

This paper presents a reexamination of the research paper titled "Communication-Avoiding Parallel Algorithms for \proc{TRSM}" by Wicky et al. We focus on the communication bandwidth cost analysis presented in the original work and identify potential issues that require clarification or revision. The problem at hand is the need to address inconsistencies and miscalculations found in the analysis, particularly in the categorization of costs into three scenarios based on the relationship between matrix dimensions and processor count. Our findings contribute to the ongoing discourse in the field and pave the way for further improvements in this area of research., Comment: 2 pages, comment on arXiv:1612.01855

This article conducts a reexamination of the research conducted by Kwasniewski et al., focusing on their adaptation of the 2.5D LU factorization algorithm with tournament pivoting, known as \func{COnfLUX}. Our reexamination reveals potential concerns regarding the upper bound, empirical investigation methods, and lower bound, despite the original study providing a theoretical foundation and an instantiation of the proposed algorithm. This paper offers a reexamination of these matters, highlighting probable shortcomings in the original investigation. Our observations are intended to enhance the development and comprehension of parallel matrix factorization algorithms., Comment: 4 pages, comment paper

Existing point cloud semantic segmentation networks cannot identify unknown classes and update their knowledge, due to a closed-set and static perspective of the real world, which would induce the intelligent agent to make bad decisions. To address this problem, we propose a Probability-Driven Framework (PDF) for open world semantic segmentation that includes (i) a lightweight U-decoder branch to identify unknown classes by estimating the uncertainties, (ii) a flexible pseudo-labeling scheme to supply geometry features along with probability distribution features of unknown classes by generating pseudo labels, and (iii) an incremental knowledge distillation strategy to incorporate novel classes into the existing knowledge base gradually. Our framework enables the model to behave like human beings, which could recognize unknown objects and incrementally learn them with the corresponding knowledge. Experimental results on the S3DIS and ScanNetv2 datasets demonstrate that the proposed PDF outperforms other methods by a large margin in both important tasks of open world semantic segmentation.

Machine Learning (ML) has become ubiquitous, fueling data-driven applications across various organizations. Contrary to the traditional perception of ML in research, ML workflows can be complex, resource-intensive, and time-consuming. Expanding an ML workflow to encompass a wider range of data infrastructure and data types may lead to larger workloads and increased deployment costs. Currently, numerous workflow engines are available (with over ten being widely recognized). This variety poses a challenge for end-users in terms of mastering different engine APIs. While efforts have primarily focused on optimizing ML Operations (MLOps) for a specific workflow engine, current methods largely overlook workflow optimization across different engines. In this work, we design and implement Couler, a system designed for unified ML workflow optimization in the cloud. Our main insight lies in the ability to generate an ML workflow using natural language (NL) descriptions. We integrate Large Language Models (LLMs) into workflow generation, and provide a unified programming interface for various workflow engines. This approach alleviates the need to understand various workflow engines' APIs. Moreover, Couler enhances workflow computation efficiency by introducing automated caching at multiple stages, enabling large workflow auto-parallelization and automatic hyperparameters tuning. These enhancements minimize redundant computational costs and improve fault tolerance during deep learning workflow training. Couler is extensively deployed in real-world production scenarios at Ant Group, handling approximately 22k workflows daily, and has successfully improved the CPU/Memory utilization by more than 15% and the workflow completion rate by around 17%.

Humans solving algorithmic (or) reasoning problems typically exhibit solution times that grow as a function of problem difficulty. Adaptive recurrent neural networks have been shown to exhibit this property for various language-processing tasks. However, little work has been performed to assess whether such adaptive computation can also enable vision models to extrapolate solutions beyond their training distribution's difficulty level, with prior work focusing on very simple tasks. In this study, we investigate a critical functional role of such adaptive processing using recurrent neural networks: to dynamically scale computational resources conditional on input requirements that allow for zero-shot generalization to novel difficulty levels not seen during training using two challenging visual reasoning tasks: PathFinder and Mazes. We combine convolutional recurrent neural networks (ConvRNNs) with a learnable halting mechanism based on Graves (2016). We explore various implementations of such adaptive ConvRNNs (AdRNNs) ranging from tying weights across layers to more sophisticated biologically inspired recurrent networks that possess lateral connections and gating. We show that 1) AdRNNs learn to dynamically halt processing early (or late) to solve easier (or harder) problems, 2) these RNNs zero-shot generalize to more difficult problem settings not shown during training by dynamically increasing the number of recurrent iterations at test time. Our study provides modeling evidence supporting the hypothesis that recurrent processing enables the functional advantage of adaptively allocating compute resources conditional on input requirements and hence allowing generalization to harder difficulty levels of a visual reasoning problem without training., Comment: 37th Conference on Neural Information Processing Systems (NeurIPS 2023)

In a series of papers (i.e., Part I--III), we presented our systematic study on nitrogen gas propagation in an evacuated copper tube cooled by liquid helium. This research aims to gain insights into the coupled heat and mass transfer processes involved in a catastrophic vacuum loss in superconducting particle accelerator beamline tubes. Our experimental measurements on nitrogen gas propagation in both normal liquid helium (He I) and superfluid helium (He II) cooled tube revealed a nearly exponential deceleration of the gas propagation. A theoretical model that accounts for the gas dynamics, heat transfer, and condensation was also developed, and various key experimental observations were nicely reproduced in our model simulations. A particularly interesting phenomenon uncovered in our study is that the gas propagation appears to nearly stop beyond a certain distance from the location where condensation starts. We refer to this distance as the freeze range. In this paper, we present our detailed analysis of the freeze range at various inlet mass fluxes and tube diameters. We show that the results can be well described by a simple correlation. The physical mechanism underlying this useful correlation is explained. Knowing the freeze range may allow accelerator engineers to develop protocols for controlling frost-layer contamination in the beamline tubes, which should be of practical importance., Comment: 8 pages, 10 figures

On the one hand, the dehazing task is an illposedness problem, which means that no unique solution exists. On the other hand, the dehazing task should take into account the subjective factor, which is to give the user selectable dehazed images rather than a single result. Therefore, this paper proposes a multi-output dehazing network by introducing illumination controllable ability, called IC-Dehazing. The proposed IC-Dehazing can change the illumination intensity by adjusting the factor of the illumination controllable module, which is realized based on the interpretable Retinex theory. Moreover, the backbone dehazing network of IC-Dehazing consists of a Transformer with double decoders for high-quality image restoration. Further, the prior-based loss function and unsupervised training strategy enable IC-Dehazing to complete the parameter learning process without the need for paired data. To demonstrate the effectiveness of the proposed IC-Dehazing, quantitative and qualitative experiments are conducted on image dehazing, semantic segmentation, and object detection tasks. Code is available at https://github.com/Xiaofeng-life/ICDehazing.

Weather recognition is an essential support for many practical life applications, including traffic safety, environment, and meteorology. However, many existing related works cannot comprehensively describe weather conditions due to their complex co-occurrence dependencies. This paper proposes a novel multi-label weather recognition model considering these dependencies. The proposed model called MASK-Convolutional Neural Network-Transformer (MASK-CT) is based on the Transformer, the convolutional process, and the MASK mechanism. The model employs multiple convolutional layers to extract features from weather images and a Transformer encoder to calculate the probability of each weather condition based on the extracted features. To improve the generalization ability of MASK-CT, a MASK mechanism is used during the training phase. The effect of the MASK mechanism is explored and discussed. The Mask mechanism randomly withholds some information from one-pair training instances (one image and its corresponding label). There are two types of MASK methods. Specifically, MASK-I is designed and deployed on the image before feeding it into the weather feature extractor and MASK-II is applied to the image label. The Transformer encoder is then utilized on the randomly masked image features and labels. The experimental results from various real-world weather recognition datasets demonstrate that the proposed MASK-CT model outperforms state-of-the-art methods. Furthermore, the high-speed dynamic real-time weather recognition capability of the MASK-CT is evaluated., Comment: Have been accepted. Appears in Knowledge-Based Systems, see https://www.sciencedirect.com/science/article/pii/S0950705123006317

The research on the single image dehazing task has been widely explored. However, as far as we know, no comprehensive study has been conducted on the robustness of the well-trained dehazing models. Therefore, there is no evidence that the dehazing networks can resist malicious attacks. In this paper, we focus on designing a group of attack methods based on first order gradient to verify the robustness of the existing dehazing algorithms. By analyzing the general purpose of image dehazing task, four attack methods are proposed, which are predicted dehazed image attack, hazy layer mask attack, haze-free image attack and haze-preserved attack. The corresponding experiments are conducted on six datasets with different scales. Further, the defense strategy based on adversarial training is adopted for reducing the negative effects caused by malicious attacks. In summary, this paper defines a new challenging problem for the image dehazing area, which can be called as adversarial attack on dehazing networks (AADN). Code and Supplementary Material are available at https://github.com/Xiaofeng-life/AADN Dehazing., Comment: This paper is accepted by IEEE Transactions on Circuits and Systems for Video Technology (TCSVT)

Recently, masked image modeling (MIM), which learns visual representations by reconstructing the masked patches of an image, has dominated self-supervised learning in computer vision. However, the pre-training of MIM always takes massive time due to the large-scale data and large-size backbones. We mainly attribute it to the random patch masking in previous MIM works, which fails to leverage the crucial semantic information for effective visual representation learning. To tackle this issue, we propose the Frequency \& Attention-driven Masking and Throwing Strategy (FAMT), which can extract semantic patches and reduce the number of training patches to boost model performance and training efficiency simultaneously. Specifically, FAMT utilizes the self-attention mechanism to extract semantic information from the image for masking during training in an unsupervised manner. However, attention alone could sometimes focus on inappropriate areas regarding the semantic information. Thus, we are motivated to incorporate the information from the frequency domain into the self-attention mechanism to derive the sampling weights for masking, which captures semantic patches for visual representation learning. Furthermore, we introduce a patch throwing strategy based on the derived sampling weights to reduce the training cost. FAMT can be seamlessly integrated as a plug-and-play module and surpasses previous works, \emph{e.g.} reducing the training phase time by nearly $50\%$ and improving the linear probing accuracy of MAE by $1.3\% \sim 3.9\%$ across various datasets, including CIFAR-10/100, Tiny ImageNet, and ImageNet-1K. FAMT also demonstrates superior performance in downstream detection and segmentation tasks.

Semantic scene completion (SSC) aims to complete a partial 3D scene and predict its semantics simultaneously. Most existing works adopt the voxel representations, thus suffering from the growth of memory and computation cost as the voxel resolution increases. Though a few works attempt to solve SSC from the perspective of 3D point clouds, they have not fully exploited the correlation and complementarity between the two tasks of scene completion and semantic segmentation. In our work, we present CasFusionNet, a novel cascaded network for point cloud semantic scene completion by dense feature fusion. Specifically, we design (i) a global completion module (GCM) to produce an upsampled and completed but coarse point set, (ii) a semantic segmentation module (SSM) to predict the per-point semantic labels of the completed points generated by GCM, and (iii) a local refinement module (LRM) to further refine the coarse completed points and the associated labels from a local perspective. We organize the above three modules via dense feature fusion in each level, and cascade a total of four levels, where we also employ feature fusion between each level for sufficient information usage. Both quantitative and qualitative results on our compiled two point-based datasets validate the effectiveness and superiority of our CasFusionNet compared to state-of-the-art methods in terms of both scene completion and semantic segmentation. The codes and datasets are available at: https://github.com/JinfengX/CasFusionNet.