Yifan Zhang

Yifan Zhang, Zhen Dong, Huanrui Yang et al. · berkeley

Multi-view 3D detection based on BEV (bird-eye-view) has recently achieved significant improvements. However, the huge memory consumption of state-of-the-art models makes it hard to deploy them on vehicles, and the non-trivial latency will affect the real-time perception of streaming applications. Despite the wide application of quantization to lighten models, we show in our paper that directly applying quantization in BEV tasks will 1) make the training unstable, and 2) lead to intolerable performance degradation. To solve these issues, our method QD-BEV enables a novel view-guided distillation (VGD) objective, which can stabilize the quantization-aware training (QAT) while enhancing the model performance by leveraging both image features and BEV features. Our experiments show that QD-BEV achieves similar or even better accuracy than previous methods with significant efficiency gains. On the nuScenes datasets, the 4-bit weight and 6-bit activation quantized QD-BEV-Tiny model achieves 37.2% NDS with only 15.8 MB model size, outperforming BevFormer-Tiny by 1.8% with an 8x model compression. On the Small and Base variants, QD-BEV models also perform superbly and achieve 47.9% NDS (28.2 MB) and 50.9% NDS (32.9 MB), respectively.

Yifan Zhang, Qijian Zhang, Junhui Hou et al.

To achieve reliable and precise scene understanding, autonomous vehicles typically incorporate multiple sensing modalities to capitalize on their complementary attributes. However, existing cross-modal 3D detectors do not fully utilize the image domain information to address the bottleneck issues of the LiDAR-based detectors. This paper presents a new cross-modal 3D object detector, namely UPIDet, which aims to unleash the potential of the image branch from two aspects. First, UPIDet introduces a new 2D auxiliary task called normalized local coordinate map estimation. This approach enables the learning of local spatial-aware features from the image modality to supplement sparse point clouds. Second, we discover that the representational capability of the point cloud backbone can be enhanced through the gradients backpropagated from the training objectives of the image branch, utilizing a succinct and effective point-to-pixel module. Extensive experiments and ablation studies validate the effectiveness of our method. Notably, we achieved the top rank in the highly competitive cyclist class of the KITTI benchmark at the time of submission. The source code is available at https://github.com/Eaphan/UPIDet.

Bo Pang, Yifan Zhang, Yaoyi Li et al.

In this paper, we propose a genuine group-level contrastive visual representation learning method whose linear evaluation performance on ImageNet surpasses the vanilla supervised learning. Two mainstream unsupervised learning schemes are the instance-level contrastive framework and clustering-based schemes. The former adopts the extremely fine-grained instance-level discrimination whose supervisory signal is not efficient due to the false negatives. Though the latter solves this, they commonly come with some restrictions affecting the performance. To integrate their advantages, we design the SMoG method. SMoG follows the framework of contrastive learning but replaces the contrastive unit from instance to group, mimicking clustering-based methods. To achieve this, we propose the momentum grouping scheme which synchronously conducts feature grouping with representation learning. In this way, SMoG solves the problem of supervisory signal hysteresis which the clustering-based method usually faces, and reduces the false negatives of instance contrastive methods. We conduct exhaustive experiments to show that SMoG works well on both CNN and Transformer backbones. Results prove that SMoG has surpassed the current SOTA unsupervised representation learning methods. Moreover, its linear evaluation results surpass the performances obtained by vanilla supervised learning and the representation can be well transferred to downstream tasks.

Longhui Yu, Yifan Zhang, Lanqing Hong et al.

Object detection for autonomous vehicles has received increasing attention in recent years, where labeled data are often expensive while unlabeled data can be collected readily, calling for research on semi-supervised learning for this area. Existing semi-supervised object detection (SSOD) methods usually assume that the labeled and unlabeled data come from the same data distribution. In autonomous driving, however, data are usually collected from different scenarios, such as different weather conditions or different times in a day. Motivated by this, we study a novel but challenging domain inconsistent SSOD problem. It involves two kinds of distribution shifts among different domains, including (1) data distribution discrepancy, and (2) class distribution shifts, making existing SSOD methods suffer from inaccurate pseudo-labels and hurting model performance. To address this problem, we propose a novel method, namely Dual-Curriculum Teacher (DucTeacher). Specifically, DucTeacher consists of two curriculums, i.e., (1) domain evolving curriculum seeks to learn from the data progressively to handle data distribution discrepancy by estimating the similarity between domains, and (2) distribution matching curriculum seeks to estimate the class distribution for each unlabeled domain to handle class distribution shifts. In this way, DucTeacher can calibrate biased pseudo-labels and handle the domain-inconsistent SSOD problem effectively. DucTeacher shows its advantages on SODA10M, the largest public semi-supervised autonomous driving dataset, and COCO, a widely used SSOD benchmark. Experiments show that DucTeacher achieves new state-of-the-art performance on SODA10M with 2.2 mAP improvement and on COCO with 0.8 mAP improvement.

Zhili Liu, Kai Chen, Yifan Zhang et al.

Text-to-image (T2I) diffusion models often inadvertently generate unwanted concepts such as watermarks and unsafe images. These concepts, termed as the "implicit concepts", could be unintentionally learned during training and then be generated uncontrollably during inference. Existing removal methods still struggle to eliminate implicit concepts primarily due to their dependency on the model's ability to recognize concepts it actually can not discern. To address this, we utilize the intrinsic geometric characteristics of implicit concepts and present the Geom-Erasing, a novel concept removal method based on the geometric-driven control. Specifically, once an unwanted implicit concept is identified, we integrate the existence and geometric information of the concept into the text prompts with the help of an accessible classifier or detector model. Subsequently, the model is optimized to identify and disentangle this information, which is then adopted as negative prompts during generation. Moreover, we introduce the Implicit Concept Dataset (ICD), a novel image-text dataset imbued with three typical implicit concepts (i.e., QR codes, watermarks, and text), reflecting real-life situations where implicit concepts are easily injected. Geom-Erasing effectively mitigates the generation of implicit concepts, achieving the state-of-the-art results on the Inappropriate Image Prompts (I2P) and our challenging Implicit Concept Dataset (ICD) benchmarks.

Weiwei Cui, Yaqi Wang, Yilong Li et al.

Accurate tooth volume segmentation is a prerequisite for computer-aided dental analysis. Deep learning-based tooth segmentation methods have achieved satisfying performances but require a large quantity of tooth data with ground truth. The dental data publicly available is limited meaning the existing methods can not be reproduced, evaluated and applied in clinical practice. In this paper, we establish a 3D dental CBCT dataset CTooth+, with 22 fully annotated volumes and 146 unlabeled volumes. We further evaluate several state-of-the-art tooth volume segmentation strategies based on fully-supervised learning, semi-supervised learning and active learning, and define the performance principles. This work provides a new benchmark for the tooth volume segmentation task, and the experiment can serve as the baseline for future AI-based dental imaging research and clinical application development.

Yifan Zhang, Arnav Vaibhav Malawade, Xiaofang Zhang et al.

Autonomous systems (AS) are systems that can adapt and change their behavior in response to unanticipated events and include systems such as aerial drones, autonomous vehicles, and ground/aquatic robots. AS require a wide array of sensors, deep-learning models, and powerful hardware platforms to perceive and safely operate in real-time. However, in many contexts, some sensing modalities negatively impact perception while increasing the system's overall energy consumption. Since AS are often energy-constrained edge devices, energy-efficient sensor fusion methods have been proposed. However, existing methods either fail to adapt to changing scenario conditions or to optimize energy efficiency system-wide. We propose CARMA: a context-aware sensor fusion approach that uses context to dynamically reconfigure the computation flow on a Field-Programmable Gate Array (FPGA) at runtime. By clock-gating unused sensors and model sub-components, CARMA significantly reduces the energy used by a multi-sensory object detector without compromising performance. We use a Deep-learning Processor Unit (DPU) based reconfiguration approach to minimize the latency of model reconfiguration. We evaluate multiple context-identification strategies, propose a novel system-wide energy-performance joint optimization, and evaluate scenario-specific perception performance. Across challenging real-world sensing contexts, CARMA outperforms state-of-the-art methods with up to 1.3x speedup and 73% lower energy consumption.

Yifan Zhang, Bryan Hooi

Diffusion models have revolutionized text-to-image generation, but their real-world applications are hampered by the extensive time needed for hundreds of diffusion steps. Although progressive distillation has been proposed to speed up diffusion sampling to 2-8 steps, it still falls short in one-step generation, and necessitates training multiple student models, which is highly parameter-extensive and time-consuming. To overcome these limitations, we introduce High-frequency-Promoting Adaptation (HiPA), a parameter-efficient approach to enable one-step text-to-image diffusion. Grounded in the insight that high-frequency information is essential but highly lacking in one-step diffusion, HiPA focuses on training one-step, low-rank adaptors to specifically enhance the under-represented high-frequency abilities of advanced diffusion models. The learned adaptors empower these diffusion models to generate high-quality images in just a single step. Compared with progressive distillation, HiPA achieves much better performance in one-step text-to-image generation (37.3 $\rightarrow$ 23.8 in FID-5k on MS-COCO 2017) and 28.6x training speed-up (108.8 $\rightarrow$ 3.8 A100 GPU days), requiring only 0.04% training parameters (7,740 million $\rightarrow$ 3.3 million). We also demonstrate HiPA's effectiveness in text-guided image editing, inpainting and super-resolution tasks, where our adapted models consistently deliver high-quality outputs in just one diffusion step. The source code will be released.

Zhiwen Chen, Zhiyu Zhu, Yifan Zhang et al.

In this paper, we delve into the nuanced challenge of tailoring the Segment Anything Models (SAMs) for integration with event data, with the overarching objective of attaining robust and universal object segmentation within the event-centric domain. One pivotal issue at the heart of this endeavor is the precise alignment and calibration of embeddings derived from event-centric data such that they harmoniously coincide with those originating from RGB imagery. Capitalizing on the vast repositories of datasets with paired events and RGB images, our proposition is to harness and extrapolate the profound knowledge encapsulated within the pre-trained SAM framework. As a cornerstone to achieving this, we introduce a multi-scale feature distillation methodology. This methodology rigorously optimizes the alignment of token embeddings originating from event data with their RGB image counterparts, thereby preserving and enhancing the robustness of the overall architecture. Considering the distinct significance that token embeddings from intermediate layers hold for higher-level embeddings, our strategy is centered on accurately calibrating the pivotal token embeddings. This targeted calibration is aimed at effectively managing the discrepancies in high-level embeddings originating from both the event and image domains. Extensive experiments on different datasets demonstrate the effectiveness of the proposed distillation method. Code in http://github.com/happychenpipi/EventSAM.

Yifan Zhang, Junhui Hou, Siyu Ren et al.

This paper introduces a novel self-supervised learning framework for enhancing 3D perception in autonomous driving scenes. Specifically, our approach, namely NCLR, focuses on 2D-3D neural calibration, a novel pretext task that estimates the rigid pose aligning camera and LiDAR coordinate systems. First, we propose the learnable transformation alignment to bridge the domain gap between image and point cloud data, converting features into a unified representation space for effective comparison and matching. Second, we identify the overlapping area between the image and point cloud with the fused features. Third, we establish dense 2D-3D correspondences to estimate the rigid pose. The framework not only learns fine-grained matching from points to pixels but also achieves alignment of the image and point cloud at a holistic level, understanding the LiDAR-to-camera extrinsic parameters. We demonstrate the efficacy of NCLR by applying the pre-trained backbone to downstream tasks, such as LiDAR-based 3D semantic segmentation, object detection, and panoptic segmentation. Comprehensive experiments on various datasets illustrate the superiority of NCLR over existing self-supervised methods. The results confirm that joint learning from different modalities significantly enhances the network's understanding abilities and effectiveness of learned representation. The code is publicly available at https://github.com/Eaphan/NCLR.

Jun Ma, Feifei Li, Sumin Kim et al.

Promptable segmentation foundation models have emerged as a transformative approach to addressing the diverse needs in medical images, but most existing models require expensive computing, posing a big barrier to their adoption in clinical practice. In this work, we organized the first international competition dedicated to promptable medical image segmentation, featuring a large-scale dataset spanning nine common imaging modalities from over 20 different institutions. The top teams developed lightweight segmentation foundation models and implemented an efficient inference pipeline that substantially reduced computational requirements while maintaining state-of-the-art segmentation accuracy. Moreover, the post-challenge phase advanced the algorithms through the design of performance booster and reproducibility tasks, resulting in improved algorithms and validated reproducibility of the winning solution. Furthermore, the best-performing algorithms have been incorporated into the open-source software with a user-friendly interface to facilitate clinical adoption. The data and code are publicly available to foster the further development of medical image segmentation foundation models and pave the way for impactful real-world applications.

Ye Qiao, Haocheng Xu, Yifan Zhang et al.

Neural Architecture Search (NAS) has proven effective in discovering new Convolutional Neural Network (CNN) architectures, particularly for scenarios with well-defined accuracy optimization goals. However, previous approaches often involve time-consuming training on super networks or intensive architecture sampling and evaluations. Although various zero-cost proxies correlated with CNN model accuracy have been proposed for efficient architecture search without training, their lack of hardware consideration makes it challenging to target highly resource-constrained edge devices such as microcontroller units (MCUs). To address these challenges, we introduce MONAS, a novel hardware-aware zero-shot NAS framework specifically designed for MCUs in edge computing. MONAS incorporates hardware optimality considerations into the search process through our proposed MCU hardware latency estimation model. By combining this with specialized performance indicators (proxies), MONAS identifies optimal neural architectures without incurring heavy training and evaluation costs, optimizing for both hardware latency and accuracy under resource constraints. MONAS achieves up to a 1104x improvement in search efficiency over previous work targeting MCUs and can discover CNN models with over 3.23x faster inference on MCUs while maintaining similar accuracy compared to more general NAS approaches.

Ye Qiao, Haocheng Xu, Yifan Zhang et al.

Neural Architecture Search (NAS) effectively discovers new Convolutional Neural Network (CNN) architectures, particularly for accuracy optimization. However, prior approaches often require resource-intensive training on super networks or extensive architecture evaluations, limiting practical applications. To address these challenges, we propose MicroNAS, a hardware-aware zero-shot NAS framework designed for microcontroller units (MCUs) in edge computing. MicroNAS considers target hardware optimality during the search, utilizing specialized performance indicators to identify optimal neural architectures without high computational costs. Compared to previous works, MicroNAS achieves up to 1104x improvement in search efficiency and discovers models with over 3.23x faster MCU inference while maintaining similar accuracy

Yifan Zhang, Xinkui Zhao, Zuxin Wang et al.

The rapid advancement of large language models has unlocked remarkable capabilities across a diverse array of natural language processing tasks. However, the considerable differences among available LLMs-in terms of cost, performance, and computational demands-pose significant challenges for users aiming to identify the most suitable model for specific tasks. In this work, we present LightRouter, a novel framework designed to systematically select and integrate a small subset of LLMs from a larger pool, with the objective of jointly optimizing both task performance and cost efficiency. LightRouter leverages an adaptive selection mechanism to identify models that require only a minimal number of boot tokens, thereby reducing costs, and further employs an effective integration strategy to combine their outputs. Extensive experiments across multiple benchmarks demonstrate that LightRouter matches or outperforms widely-used ensemble baselines, achieving up to a 25% improvement in accuracy. Compared with leading high-performing models, LightRouter achieves comparable performance while reducing inference costs by up to 27%. Importantly, our framework operates without any prior knowledge of individual models and relies exclusively on inexpensive, lightweight models. This work introduces a practical approach for efficient LLM selection and provides valuable insights into optimal strategies for model combination.

Shuaicheng Niu, Guohao Chen, Deyu Chen et al.

Test-time adaptation (TTA) may fail to improve or even harm the model performance when test data have: 1) mixed distribution shifts, 2) small batch sizes, 3) online imbalanced label distribution shifts. This is often a key obstacle preventing existing TTA methods from being deployed in the real world. In this paper, we investigate the unstable reasons and find that the batch norm layer is a crucial factor hindering TTA stability. Conversely, TTA can perform more stably with batch-agnostic norm layers, i.e., group or layer norm. However, we observe that TTA with group and layer norms does not always succeed and still suffers many failure cases, i.e., the model collapses into trivial solutions by assigning the same class label for all samples. By digging into this, we find that, during the collapse process: 1) the model gradients often undergo an initial explosion followed by rapid degradation, suggesting that certain noisy test samples with large gradients may disrupt adaptation; and 2) the model representations tend to exhibit high correlations and classification bias. To address this, we first propose a sharpness-aware and reliable entropy minimization method, called SAR, for stabilizing TTA from two aspects: 1) remove partial noisy samples with large gradients, 2) encourage model weights to go to a flat minimum so that the model is robust to the remaining noisy samples. Based on SAR, we further introduce SAR^2 to prevent representation collapse with two regularizers: 1) a redundancy regularizer to reduce inter-dimensional correlations among centroid-invariant features; and 2) an inequity regularizer to maximize the prediction entropy of a prototype centroid, thereby penalizing biased representations toward any specific class. Promising results demonstrate that our methods perform more stably over prior methods and are computationally efficient under the above wild test scenarios.

Ye Qiao, Zhiheng Chen, Yian Wang et al.

Transformer-based models have demonstrated superior performance in various fields, including natural language processing and computer vision. However, their enormous model size and high demands in computation, memory, and communication limit their deployment to edge platforms for local, secure inference. Binary transformers offer a compact, low-complexity solution for edge deployment with reduced bandwidth needs and acceptable accuracy. However, existing binary transformers perform inefficiently on current hardware due to the lack of binary specific optimizations. To address this, we introduce COBRA, an algorithm-architecture co-optimized binary Transformer accelerator for edge computing. COBRA features a real 1-bit binary multiplication unit, enabling matrix operations with -1, 0, and +1 values, surpassing ternary methods. With further hardware-friendly optimizations in the attention block, COBRA achieves up to 3,894.7 GOPS throughput and 448.7 GOPS/Watt energy efficiency on edge FPGAs, delivering a 311x energy efficiency improvement over GPUs and a 3.5x throughput improvement over the state-of-the-art binary accelerator, with only negligible inference accuracy degradation.

Guoquan Xu, Hezhi Cao, Yifan Zhang et al.

Attention mechanism plays a more and more important role in point cloud analysis and channel attention is one of the hotspots. With so much channel information, it is difficult for neural networks to screen useful channel information. Thus, an adaptive channel encoding mechanism is proposed to capture channel relationships in this paper. It improves the quality of the representation generated by the network by explicitly encoding the interdependence between the channels of its features. Specifically, a channel-wise convolution (Channel-Conv) is proposed to adaptively learn the relationship between coordinates and features, so as to encode the channel. Different from the popular attention weight schemes, the Channel-Conv proposed in this paper realizes adaptability in convolution operation, rather than simply assigning different weights for channels. Extensive experiments on existing benchmarks verify our method achieves the state of the arts.

Ge Zhang, Shaohui Mei, Mingyang Ma et al.

Spectral unmixing (SU) expresses the mixed pixels existed in hyperspectral images as the product of endmember and abundance, which has been widely used in hyperspectral imagery analysis. However, the influence of light, acquisition conditions and the inherent properties of materials, results in that the identified endmembers can vary spectrally within a given image (construed as spectral variability). To address this issue, recent methods usually use a priori obtained spectral library to represent multiple characteristic spectra of the same object, but few of them extracted the spectral variability explicitly. In this paper, a spectral variability augmented sparse unmixing model (SVASU) is proposed, in which the spectral variability is extracted for the first time. The variable spectra are divided into two parts of intrinsic spectrum and spectral variability for spectral reconstruction, and modeled synchronously in the SU model adding the regular terms restricting the sparsity of abundance and the generalization of the variability coefficient. It is noted that the spectral variability library and the intrinsic spectral library are all constructed from the In-situ observed image. Experimental results over both synthetic and real-world data sets demonstrate that the augmented decomposition by spectral variability significantly improves the unmixing performance than the decomposition only by spectral library, as well as compared to state-of-the-art algorithms.

Yi-Fan Zhang, Zhang Zhang, Da Li et al.

Generalizable person Re-Identification (ReID) has attracted growing attention in recent computer vision community. In this work, we construct a structural causal model among identity labels, identity-specific factors (clothes/shoes color etc), and domain-specific factors (background, viewpoints etc). According to the causal analysis, we propose a novel Domain Invariant Representation Learning for generalizable person Re-Identification (DIR-ReID) framework. Specifically, we first propose to disentangle the identity-specific and domain-specific feature spaces, based on which we propose an effective algorithmic implementation for backdoor adjustment, essentially serving as a causal intervention towards the SCM. Extensive experiments have been conducted, showing that DIR-ReID outperforms state-of-the-art methods on large-scale domain generalization ReID benchmarks.

Yunxiang Li, Yifan Zhang, Yaqi Wang et al.

As the most economical and routine auxiliary examination in the diagnosis of root canal treatment, oral X-ray has been widely used by stomatologists. It is still challenging to segment the tooth root with a blurry boundary for the traditional image segmentation method. To this end, we propose a model for high-resolution segmentation based on polynomial curve fitting with landmark detection (HS-PCL). It is based on detecting multiple landmarks evenly distributed on the edge of the tooth root to fit a smooth polynomial curve as the segmentation of the tooth root, thereby solving the problem of fuzzy edge. In our model, a maximum number of the shortest distances algorithm (MNSDA) is proposed to automatically reduce the negative influence of the wrong landmarks which are detected incorrectly and deviate from the tooth root on the fitting result. Our numerical experiments demonstrate that the proposed approach not only reduces Hausdorff95 (HD95) by 33.9% and Average Surface Distance (ASD) by 42.1% compared with the state-of-the-art method, but it also achieves excellent results on the minute quantity of datasets, which greatly improves the feasibility of automatic root canal therapy evaluation by medical image computing.

Yi-Fan Zhang, Weiqiang Ren, Zhang Zhang et al.

In object detection, bounding box regression (BBR) is a crucial step that determines the object localization performance. However, we find that most previous loss functions for BBR have two main drawbacks: (i) Both $\ell_n$-norm and IOU-based loss functions are inefficient to depict the objective of BBR, which leads to slow convergence and inaccurate regression results. (ii) Most of the loss functions ignore the imbalance problem in BBR that the large number of anchor boxes which have small overlaps with the target boxes contribute most to the optimization of BBR. To mitigate the adverse effects caused thereby, we perform thorough studies to exploit the potential of BBR losses in this paper. Firstly, an Efficient Intersection over Union (EIOU) loss is proposed, which explicitly measures the discrepancies of three geometric factors in BBR, i.e., the overlap area, the central point and the side length. After that, we state the Effective Example Mining (EEM) problem and propose a regression version of focal loss to make the regression process focus on high-quality anchor boxes. Finally, the above two parts are combined to obtain a new loss function, namely Focal-EIOU loss. Extensive experiments on both synthetic and real datasets are performed. Notable superiorities on both the convergence speed and the localization accuracy can be achieved over other BBR losses.

Bin Fang, Xingming Long, Yifan Zhang et al.

This paper introduces a new type of system for fabric defect detection with the tactile inspection system. Different from existed visual inspection systems, the proposed system implements a vision-based tactile sensor. The tactile sensor, which mainly consists of a camera, four LEDs, and an elastic sensing layer, captures detailed information about fabric surface structure and ignores the color and pattern. Thus, the ambiguity between a defect and image background related to fabric color and pattern is avoided. To utilize the tactile sensor for fabric inspection, we employ intensity adjustment for image preprocessing, Residual Network with ensemble learning for detecting defects, and uniformity measurement for selecting ideal dataset for model training. An experiment is conducted to verify the performance of the proposed tactile system. The experimental results have demonstrated the feasibility of the proposed system, which performs well in detecting structural defects for various types of fabrics. In addition, the system does not require external light sources, which skips the process of setting up and tuning a lighting environment.