Xulei Yang

Kaixin Xu, Zhe Wang, Chunyun Chen et al.

Vision transformers have emerged as a promising alternative to convolutional neural networks for various image analysis tasks, offering comparable or superior performance. However, one significant drawback of ViTs is their resource-intensive nature, leading to increased memory footprint, computation complexity, and power consumption. To democratize this high-performance technology and make it more environmentally friendly, it is essential to compress ViT models, reducing their resource requirements while maintaining high performance. In this paper, we introduce a new block-structured pruning to address the resource-intensive issue for ViTs, offering a balanced trade-off between accuracy and hardware acceleration. Unlike unstructured pruning or channel-wise structured pruning, block pruning leverages the block-wise structure of linear layers, resulting in more efficient matrix multiplications. To optimize this pruning scheme, our paper proposes a novel hardware-aware learning objective that simultaneously maximizes speedup and minimizes power consumption during inference, tailored to the block sparsity structure. This objective eliminates the need for empirical look-up tables and focuses solely on reducing parametrized layer connections. Moreover, our paper provides a lightweight algorithm to achieve post-training pruning for ViTs, utilizing second-order Taylor approximation and empirical optimization to solve the proposed hardware-aware objective. Extensive experiments on ImageNet are conducted across various ViT architectures, including DeiT-B and DeiT-S, demonstrating competitive performance with other pruning methods and achieving a remarkable balance between accuracy preservation and power savings. Especially, we achieve 3.93x speedup on dedicated hardware and GPUs respectively for DeiT-B, and a power reduction by 1.4x on GPUs. Code released to https://github.com/Akimoto-Cris/LPViT.

Ziyuan Zhao, Kaixin Xu, Huai Zhe Yeo et al.

Domain shift has been a long-standing issue for medical image segmentation. Recently, unsupervised domain adaptation (UDA) methods have achieved promising cross-modality segmentation performance by distilling knowledge from a label-rich source domain to a target domain without labels. In this work, we propose a multi-scale self-ensembling based UDA framework for automatic segmentation of two key brain structures i.e., Vestibular Schwannoma (VS) and Cochlea on high-resolution T2 images. First, a segmentation-enhanced contrastive unpaired image translation module is designed for image-level domain adaptation from source T1 to target T2. Next, multi-scale deep supervision and consistency regularization are introduced to a mean teacher network for self-ensemble learning to further close the domain gap. Furthermore, self-training and intensity augmentation techniques are utilized to mitigate label scarcity and boost cross-modality segmentation performance. Our method demonstrates promising segmentation performance with a mean Dice score of 83.8% and 81.4% and an average asymmetric surface distance (ASSD) of 0.55 mm and 0.26 mm for the VS and Cochlea, respectively in the validation phase of the crossMoDA 2022 challenge.

Yining Pan, Qiongjie Cui, Xulei Yang et al.

LiDAR-based 3D panoptic segmentation often struggles with the inherent sparsity of data from LiDAR sensors, which makes it challenging to accurately recognize distant or small objects. Recently, a few studies have sought to overcome this challenge by integrating LiDAR inputs with camera images, leveraging the rich and dense texture information provided by the latter. While these approaches have shown promising results, they still face challenges, such as misalignment during data augmentation and the reliance on post-processing steps. To address these issues, we propose Image-Assists-LiDAR (IAL), a novel multi-modal 3D panoptic segmentation framework. In IAL, we first introduce a modality-synchronized data augmentation strategy, PieAug, to ensure alignment between LiDAR and image inputs from the start. Next, we adopt a transformer decoder to directly predict panoptic segmentation results. To effectively fuse LiDAR and image features into tokens for the decoder, we design a Geometric-guided Token Fusion (GTF) module. Additionally, we leverage the complementary strengths of each modality as priors for query initialization through a Prior-based Query Generation (PQG) module, enhancing the decoder's ability to generate accurate instance masks. Our IAL framework achieves state-of-the-art performance compared to previous multi-modal 3D panoptic segmentation methods on two widely used benchmarks. Code and models are publicly available at <https://github.com/IMPL-Lab/IAL.git>.

Kaixin Xu, Qingtian Feng, Hao Chen et al.

Applying deep neural networks to 3D point cloud processing has attracted increasing attention due to its advanced performance in many areas, such as AR/VR, autonomous driving, and robotics. However, as neural network models and 3D point clouds expand in size, it becomes a crucial challenge to reduce the computational and memory overhead to meet latency and energy constraints in real-world applications. Although existing approaches have proposed to reduce both computational cost and memory footprint, most of them only address the spatial redundancy in inputs, i.e. removing the redundancy of background points in 3D data. In this paper, we propose a novel post-training weight pruning scheme for 3D object detection that is (1) orthogonal to all existing point cloud sparsifying methods, which determines redundant parameters in the pretrained model that lead to minimal distortion in both locality and confidence (detection distortion); and (2) a universal plug-and-play pruning framework that works with arbitrary 3D detection model. This framework aims to minimize detection distortion of network output to maximally maintain detection precision, by identifying layer-wise sparsity based on second-order Taylor approximation of the distortion. Albeit utilizing second-order information, we introduced a lightweight scheme to efficiently acquire Hessian information, and subsequently perform dynamic programming to solve the layer-wise sparsity. Extensive experiments on KITTI, Nuscenes and ONCE datasets demonstrate that our approach is able to maintain and even boost the detection precision on pruned model under noticeable computation reduction (FLOPs). Noticeably, we achieve over 3.89x, 3.72x FLOPs reduction on CenterPoint and PVRCNN model, respectively, without mAP decrease, significantly improving the state-of-the-art.

Naiyu Fang, Zheyuan Zhou, Fayao Liu et al.

3D semantic occupancy prediction offers an intuitive and efficient scene understanding and has attracted significant interest in autonomous driving perception. Existing approaches either rely on full supervision, which demands costly voxel-level annotations, or on self-supervision, which provides limited guidance and yields suboptimal performance. To address these challenges, we propose OccLE, a Label-Efficient 3D Semantic Occupancy Prediction that takes images and LiDAR as inputs and maintains high performance with limited voxel annotations. Our intuition is to decouple the semantic and geometric learning tasks and then fuse the learned feature grids from both tasks for the final semantic occupancy prediction. Therefore, the semantic branch distills 2D foundation model to provide aligned pseudo labels for 2D and 3D semantic learning. The geometric branch integrates image and LiDAR inputs in cross-plane synergy based on their inherency, employing semi-supervision to enhance geometry learning. We fuse semantic-geometric feature grids through Dual Mamba and incorporate a scatter-accumulated projection to supervise unannotated prediction with aligned pseudo labels. Experiments show that OccLE achieves competitive performance with only 10\% of voxel annotations on the SemanticKITTI and Occ3D-nuScenes datasets. The code will be publicly released on https://github.com/NerdFNY/OccLE

Jinpeng Lin, Xulei Yang, Tianrui Li et al.

Recognizing 3D point cloud plays a pivotal role in many real-world applications. However, deploying 3D point cloud deep learning model is vulnerable to adversarial attacks. Despite many efforts into developing robust model by adversarial training, they may become less effective against emerging attacks. This limitation motivates the development of adversarial purification which employs generative model to mitigate the impact of adversarial attacks. In this work, we highlight the remaining challenges from two perspectives. First, the purification based method requires retraining the classifier on purified samples which introduces additional computation overhead. Moreover, in a more realistic scenario, testing samples arrives in a streaming fashion and adversarial samples are not isolated from clean samples. These challenges motivates us to explore dynamically update model upon observing testing samples. We proposed a test-time purified self-training strategy to achieve this objective. Adaptive thresholding and feature distribution alignment are introduced to improve the robustness of self-training. Extensive results on different adversarial attacks suggest the proposed method is complementary to purification based method in handling continually changing adversarial attacks on the testing data stream.

Rong Li, Shijie Li, Lingdong Kong et al.

3D Visual Grounding (3DVG) aims to locate objects in 3D scenes based on textual descriptions, essential for applications like augmented reality and robotics. Traditional 3DVG approaches rely on annotated 3D datasets and predefined object categories, limiting scalability and adaptability. To overcome these limitations, we introduce SeeGround, a zero-shot 3DVG framework leveraging 2D Vision-Language Models (VLMs) trained on large-scale 2D data. SeeGround represents 3D scenes as a hybrid of query-aligned rendered images and spatially enriched text descriptions, bridging the gap between 3D data and 2D-VLMs input formats. We propose two modules: the Perspective Adaptation Module, which dynamically selects viewpoints for query-relevant image rendering, and the Fusion Alignment Module, which integrates 2D images with 3D spatial descriptions to enhance object localization. Extensive experiments on ScanRefer and Nr3D demonstrate that our approach outperforms existing zero-shot methods by large margins. Notably, we exceed weakly supervised methods and rival some fully supervised ones, outperforming previous SOTA by 7.7% on ScanRefer and 7.1% on Nr3D, showcasing its effectiveness in complex 3DVG tasks.

Xiaofeng Yang, Yiwen Chen, Cheng Chen et al.

We propose a unified framework aimed at enhancing the diffusion priors for 3D generation tasks. Despite the critical importance of these tasks, existing methodologies often struggle to generate high-caliber results. We begin by examining the inherent limitations in previous diffusion priors. We identify a divergence between the diffusion priors and the training procedures of diffusion models that substantially impairs the quality of 3D generation. To address this issue, we propose a novel, unified framework that iteratively optimizes both the 3D model and the diffusion prior. Leveraging the different learnable parameters of the diffusion prior, our approach offers multiple configurations, affording various trade-offs between performance and implementation complexity. Notably, our experimental results demonstrate that our method markedly surpasses existing techniques, establishing new state-of-the-art in the realm of text-to-3D generation. Furthermore, our approach exhibits impressive performance on both NeRF and the newly introduced 3D Gaussian Splatting backbones. Additionally, our framework yields insightful contributions to the understanding of recent score distillation methods, such as the VSD and DDS loss.

Mingda Xu, Peisheng Qian, Ziyuan Zhao et al.

Protein-protein interactions (PPIs) play key roles in a broad range of biological processes. Numerous strategies have been proposed for predicting PPIs, and among them, graph-based methods have demonstrated promising outcomes owing to the inherent graph structure of PPI networks. This paper reviews various graph-based methodologies, and discusses their applications in PPI prediction. We classify these approaches into two primary groups based on their model structures. The first category employs Graph Neural Networks (GNN) or Graph Convolutional Networks (GCN), while the second category utilizes Graph Attention Networks (GAT), Graph Auto-Encoders and Graph-BERT. We highlight the distinctive methodologies of each approach in managing the graph-structured data inherent in PPI networks and anticipate future research directions in this domain.

Rong Li, Shijie Li, Lingdong Kong et al.

3D Visual Grounding (3DVG) seeks to locate target objects in 3D scenes using natural language descriptions, enabling downstream applications such as augmented reality and robotics. Existing approaches typically rely on labeled 3D data and predefined categories, limiting scalability to open-world settings. We present SeeGround, a zero-shot 3DVG framework that leverages 2D Vision-Language Models (VLMs) to bypass the need for 3D-specific training. To bridge the modality gap, we introduce a hybrid input format that pairs query-aligned rendered views with spatially enriched textual descriptions. Our framework incorporates two core components: a Perspective Adaptation Module that dynamically selects optimal viewpoints based on the query, and a Fusion Alignment Module that integrates visual and spatial signals to enhance localization precision. Extensive evaluations on ScanRefer and Nr3D confirm that SeeGround achieves substantial improvements over existing zero-shot baselines -- outperforming them by 7.7% and 7.1%, respectively -- and even rivals fully supervised alternatives, demonstrating strong generalization under challenging conditions.

Zhe Wang, Xulei Yang

There has been huge progress on video action recognition in recent years. However, many works focus on tweaking existing 2D backbones due to the reliance of ImageNet pretraining, which restrains the models from achieving higher efficiency for video recognition. In this work we introduce a simple and very efficient 3D convolutional neural network for video action recognition. The design of the building block consists of a channel-wise convolution, followed by a spatial group convolution, and finally a temporal convolution. We evaluate the performance and efficiency of our proposed network on several video action recognition datasets by directly training on the target dataset without relying on pertaining. On Something-Something-V1&V2, Kinetics-400 and Multi-Moments in Time, our network can match or even surpass the performance of other models which are several times larger. On the fine-grained action recognition dataset FineGym, we beat the previous state-of-the-art accuracy achieved with 2-stream methods by more than 5% using only RGB input.

Gabriel Tjio, Jie Zhang, Xulei Yang et al.

Test-time adaptation enables models to adapt to evolving domains. However, balancing the tradeoff between preserving knowledge and adapting to domain shifts remains challenging for model adaptation methods, since adapting to domain shifts can induce forgetting of task-relevant knowledge. To address this problem, we propose FOCUS, a novel frequency-based conditioning approach within a diffusion-driven input-adaptation framework. Utilising learned, spatially adaptive frequency priors, our approach conditions the reverse steps during diffusion-driven denoising to preserve task-relevant semantic information for dense prediction. FOCUS leverages a trained, lightweight, Y-shaped Frequency Prediction Network (Y-FPN) that disentangles high and low frequency information from noisy images. This minimizes the computational costs involved in implementing our approach in a diffusion-driven framework. We train Y-FPN with FrequencyMix, a novel data augmentation method that perturbs the images across diverse frequency bands, which improves the robustness of our approach to diverse corruptions. We demonstrate the effectiveness of FOCUS for semantic segmentation and monocular depth estimation across 15 corruption types and three datasets, achieving state-of-the-art averaged performance. In addition to improving standalone performance, FOCUS complements existing model adaptation methods since we can derive pseudo labels from FOCUS-denoised images for additional supervision. Even under limited, intermittent supervision with the pseudo labels derived from the FOCUS denoised images, we show that FOCUS mitigates catastrophic forgetting for recent model adaptation methods.

Yuzhe Zhu, Lile Cai, Kangkang Lu et al.

Neural Radiance Field (NeRF) models are implicit neural scene representation methods that offer unprecedented capabilities in novel view synthesis. Semantically-aware NeRFs not only capture the shape and radiance of a scene, but also encode semantic information of the scene. The training of semantically-aware NeRFs typically requires pixel-level class labels, which can be prohibitively expensive to collect. In this work, we explore active learning as a potential solution to alleviate the annotation burden. We investigate various design choices for active learning of semantically-aware NeRF, including selection granularity and selection strategies. We further propose a novel active learning strategy that takes into account 3D geometric constraints in sample selection. Our experiments demonstrate that active learning can effectively reduce the annotation cost of training semantically-aware NeRF, achieving more than 2X reduction in annotation cost compared to random sampling.

Yifan Liu, Xun Xu, Shijie Li et al.

Multi-camera systems provide richer contextual information for industrial anomaly detection. However, traditional methods process each view independently, disregarding the complementary information across viewpoints. Existing multi-view anomaly detection approaches typically employ data-driven cross-view attention for feature fusion but fail to leverage the unique geometric properties of multi-camera setups. In this work, we introduce an epipolar geometry-constrained attention module to guide cross-view fusion, ensuring more effective information aggregation. To further enhance the potential of cross-view attention, we propose a pretraining strategy inspired by memory bank-based anomaly detection. This approach encourages normal feature representations to form multiple local clusters and incorporate multi-view aware negative sample synthesis to regularize pretraining. We demonstrate that our epipolar guided multi-view anomaly detection framework outperforms existing methods on the state-of-the-art multi-view anomaly detection dataset.

Qiyu Wei, Xun Xu, Zeng Zeng et al.

The patterns on wafer maps play a crucial role in helping engineers identify the causes of production issues during semiconductor manufacturing. In order to reduce costs and improve accuracy, automation technology is essential, and recent developments in deep learning have led to impressive results in wafer map pattern recognition. In this context, inspired by the effectiveness of semi-supervised learning and contrastive learning methods, we introduce an innovative approach that integrates the Mean Teacher framework with the supervised contrastive learning loss for enhanced wafer map pattern recognition. Our methodology not only addresses the nuances of wafer patterns but also tackles challenges arising from limited labeled data. To further refine the process, we address data imbalance in the wafer dataset by employing SMOTE and under-sampling techniques. We conduct a comprehensive analysis of our proposed method and demonstrate its effectiveness through experiments using real-world dataset WM811K obtained from semiconductor manufacturers. Compared to the baseline method, our method has achieved 5.46%, 6.68%, 5.42%, and 4.53% improvements in Accuracy, Precision, Recall, and F1 score, respectively.

Xue Geng, Zhe Wang, Chunyun Chen et al.

Deep neural networks (DNNs) have been widely used in many artificial intelligence (AI) tasks. However, deploying them brings significant challenges due to the huge cost of memory, energy, and computation. To address these challenges, researchers have developed various model compression techniques such as model quantization and model pruning. Recently, there has been a surge in research of compression methods to achieve model efficiency while retaining the performance. Furthermore, more and more works focus on customizing the DNN hardware accelerators to better leverage the model compression techniques. In addition to efficiency, preserving security and privacy is critical for deploying DNNs. However, the vast and diverse body of related works can be overwhelming. This inspires us to conduct a comprehensive survey on recent research toward the goal of high-performance, cost-efficient, and safe deployment of DNNs. Our survey first covers the mainstream model compression techniques such as model quantization, model pruning, knowledge distillation, and optimizations of non-linear operations. We then introduce recent advances in designing hardware accelerators that can adapt to efficient model compression approaches. Additionally, we discuss how homomorphic encryption can be integrated to secure DNN deployment. Finally, we discuss several issues, such as hardware evaluation, generalization, and integration of various compression approaches. Overall, we aim to provide a big picture of efficient DNNs, from algorithm to hardware accelerators and security perspectives.

Ziyuan Zhao, Peisheng Qian, Xulei Yang et al.

Protein-protein interactions (PPIs) are crucial in various biological processes and their study has significant implications for drug development and disease diagnosis. Existing deep learning methods suffer from significant performance degradation under complex real-world scenarios due to various factors, e.g., label scarcity and domain shift. In this paper, we propose a self-ensembling multigraph neural network (SemiGNN-PPI) that can effectively predict PPIs while being both efficient and generalizable. In SemiGNN-PPI, we not only model the protein correlations but explore the label dependencies by constructing and processing multiple graphs from the perspectives of both features and labels in the graph learning process. We further marry GNN with Mean Teacher to effectively leverage unlabeled graph-structured PPI data for self-ensemble graph learning. We also design multiple graph consistency constraints to align the student and teacher graphs in the feature embedding space, enabling the student model to better learn from the teacher model by incorporating more relationships. Extensive experiments on PPI datasets of different scales with different evaluation settings demonstrate that SemiGNN-PPI outperforms state-of-the-art PPI prediction methods, particularly in challenging scenarios such as training with limited annotations and testing on unseen data.

Way Tan, Bihan Wen, Xulei Yang

StyleGAN is one of the state-of-the-art image generators which is well-known for synthesizing high-resolution and hyper-realistic face images. Though images generated by vanilla StyleGAN model are visually appealing, they sometimes contain prominent circular artifacts which severely degrade the quality of generated images. In this work, we provide a systematic investigation on how those circular artifacts are formed by studying the functionalities of different stages of vanilla StyleGAN architecture, with both mechanism analysis and extensive experiments. The key modules of vanilla StyleGAN that promote such undesired artifacts are highlighted. Our investigation also explains why the artifacts are usually circular, relatively small and rarely split into 2 or more parts. Besides, we propose a simple yet effective solution to remove the prominent circular artifacts for vanilla StyleGAN, by applying a novel pixel-instance normalization (PIN) layer.

Balagopal Unnikrishnan, Pranshu Ranjan Singh, Xulei Yang et al.

Automated heart sounds classification is a much-required diagnostic tool in the view of increasing incidences of heart related diseases worldwide. In this study, we conduct a comprehensive study of heart sounds classification by using various supervised, semi-supervised and unsupervised approaches on the PhysioNet/CinC 2016 Challenge dataset. Supervised approaches, including deep learning and machine learning methods, require large amounts of labelled data to train the models, which are challenging to obtain in most practical scenarios. In view of the need to reduce the labelling burden for clinical practices, where human labelling is both expensive and time-consuming, semi-supervised or even unsupervised approaches in restricted data setting are desirable. A GAN based semi-supervised method is therefore proposed, which allows the usage of unlabelled data samples to boost the learning of data distribution. It achieves a better performance in terms of AUROC over the supervised baseline when limited data samples exist. Furthermore, several unsupervised methods are explored as an alternative approach by considering the given problem as an anomaly detection scenario. In particular, the unsupervised feature extraction using 1D CNN Autoencoder coupled with one-class SVM obtains good performance without any data labelling. The potential of the proposed semi-supervised and unsupervised methods may lead to a workflow tool in the future for the creation of higher quality datasets.

Gabriel Tjio, Xulei Yang, Jia Mei Hong et al.

Manual annotation of pathology slides for cancer diagnosis is laborious and repetitive. Therefore, much effort has been devoted to develop computer vision solutions. Our approach, (FLASH), is based on a Deep Convolutional Neural Network (DCNN) architecture. It reduces computational costs and is faster than typical deep learning approaches by two orders of magnitude, making high throughput processing a possibility. In computer vision approaches using deep learning methods, the input image is subdivided into patches which are separately passed through the neural network. Features extracted from these patches are used by the classifier to annotate the corresponding region. Our approach aggregates all the extracted features into a single matrix before passing them to the classifier. Previously, the features are extracted from overlapping patches. Aggregating the features eliminates the need for processing overlapping patches, which reduces the computations required. DCCN and FLASH demonstrate high sensitivity (~ 0.96), good precision (~0.78) and high F1 scores (~0.84). The average time taken to process each sample for FLASH and DCNN is 96.6 seconds and 9489.20 seconds, respectively. Our approach was approximately 100 times faster than the original DCNN approach while simultaneously preserving high accuracy and precision.

Ziyuan Zhao, Xiaoman Zhang, Cen Chen et al.

Segmentation stands at the forefront of many high-level vision tasks. In this study, we focus on segmenting finger bones within a newly introduced semi-supervised self-taught deep learning framework which consists of a student network and a stand-alone teacher module. The whole system is boosted in a life-long learning manner wherein each step the teacher module provides a refinement for the student network to learn with newly unlabeled data. Experimental results demonstrate the superiority of the proposed method over conventional supervised deep learning methods.

Xulei Yang, Zeng Zeng, Si Yong Yeo et al.

In this study, a multi-task deep neural network is proposed for skin lesion analysis. The proposed multi-task learning model solves different tasks (e.g., lesion segmentation and two independent binary lesion classifications) at the same time by exploiting commonalities and differences across tasks. This results in improved learning efficiency and potential prediction accuracy for the task-specific models, when compared to training the individual models separately. The proposed multi-task deep learning model is trained and evaluated on the dermoscopic image sets from the International Skin Imaging Collaboration (ISIC) 2017 Challenge - Skin Lesion Analysis towards Melanoma Detection, which consists of 2000 training samples and 150 evaluation samples. The experimental results show that the proposed multi-task deep learning model achieves promising performances on skin lesion segmentation and classification. The average value of Jaccard index for lesion segmentation is 0.724, while the average values of area under the receiver operating characteristic curve (AUC) on two individual lesion classifications are 0.880 and 0.972, respectively.