Multi-view 3D detection with bird’s eye view (BEV) is crucial for autonomous driving and robotics, but its robustness in real-world is limited as it struggles to predict accurate depth values. A mainstream solution, cross-modal distillation, transfers depth information from LiDAR to camera models but also unintentionally transfers depth-irrelevant information (e.g. LiDAR density). To mitigate this issue, we propose RayD3D, which transfers crucial depth knowledge along the ray: a line projecting from the camera to true location of an object. It is based on the fundamental imaging principle that predicted location of this object can only vary along this ray, which is finally determined by predicted depth value. Therefore, distilling along the ray enables more effective depth information transfer. More specifically, we design two ray-based distillation modules. Ray-based Contrastive Distillation (RCD) incorporates contrastive learning into distillation by sampling along the ray to learn how LiDAR accurately locates objects. Ray-based Weighted Distillation (RWD) adaptively adjusts distillation weight based on the ray to minimize the interference of depth-irrelevant information in LiDAR. For validation, we widely apply RayD3D into three representative types of BEV-based models, including BEVDet, BEVDepth4D, and BEVFormer. Our method is trained on clean NuScenes, and tested on both clean NuScenes and RoboBEV with a variety types of data corruptions. Our method significantly improves the robustness of all the three base models in all scenarios without increasing inference costs, and achieves the best when compared to recently released multi-view and distillation models.

Tumor synthesis can generate challenging cases that AI often misses or over-detects. Training on these cases improves AI performance. However, most existing synthesis methods are either unconditional- generating images from random variables-or conditioned only on tumor shape. As a result, they lack control over clinically important tumor characteristics, such as texture, heterogeneity, boundary, and pathology. The generated tumors are therefore overly similar or duplicates of existing training cases, failing to effectively address AI's weaknesses. We propose a new text-driven tumor synthesis approach, termed TextoMorph, that provides textual control over tumor characteristics in conjunction with mask control. This approach is particularly beneficial for examples that confuse the AI the most, such as early tumor detection (improving Sensitivity by + 6.5%), tumor segmentation for precise radiotherapy (improving NSD by + 3.1%), and classification between benign and malignant tumors (improving Sensitivity by + 8.2%). By incorporating text mined from radiology reports into the synthesis process, we increase the variability and controllability of the synthetic tumors to target AI's failure cases more precisely. Moreover, TextoMorph uses contrastive learning across different texts and CT scans, significantly reducing dependence on scarce image-report pairs (only 141 pairs used in this study) by leveraging a large corpus of 34,035 radiology reports. Finally, we have developed rigorous tests to evaluate synthetic tumors, showing that our synthetic tumors is realistic and diverse in texture, heterogeneity, boundary, and pathology. Code and models are available at https://github.com/MrGiovanni/TextoMorph.

With over 85 million CT scans performed annually in the United States, creating tumor-related reports is a challenging and time-consuming task for radiologists. To address this need, we present RadGPT, an Anatomy-Aware Vision-Language AI Agent for generating detailed reports from CT scans. RadGPT first segments tumors, including benign cysts and malignant tumors, and their surrounding anatomical structures, then transforms this information into both structured reports and narrative reports. These reports provide tumor size, shape, location, attenuation, volume, and interactions with surrounding blood vessels and organs. Extensive evaluation on unseen hospitals shows that RadGPT can produce accurate reports, with high sensitivity/specificity for small tumor (<2 cm) detection: 80/73% for liver tumors, 92/78% for kidney tumors, and 77/77% for pancreatic tumors. For large tumors, sensitivity ranges from 89% to 97%. The results significantly surpass the state-of-the-art in abdominal CT report generation. RadGPT generated reports for 17 public datasets. Through radiologist review and refinement, we have ensured the reports' accuracy, and created the first publicly available image-text 3D medical dataset, comprising over 1.8 million text tokens and 2.7 million images from 9,262 CT scans, including 2,947 tumor scans/reports of 8,562 tumor instances. Our reports can: (1) localize tumors in eight liver sub-segments and three pancreatic sub-segments annotated per-voxel; (2) determine pancreatic tumor stage (T1-T4) in 260 reports; and (3) present individual analyses of multiple tumors--rare in human-made reports. Importantly, 948 of the reports are for early-stage tumors.

This paper presents MedSegFactory, a versatile medical synthesis framework that generates high-quality paired medical images and segmentation masks across modalities and tasks. It aims to serve as an unlimited data repository, supplying image-mask pairs to enhance existing segmentation tools. The core of MedSegFactory is a dual-stream diffusion model, where one stream synthesizes medical images and the other generates corresponding segmentation masks. To ensure precise alignment between image-mask pairs, we introduce Joint Cross-Attention (JCA), enabling a collaborative denoising paradigm by dynamic cross-conditioning between streams. This bidirectional interaction allows both representations to guide each other's generation, enhancing consistency between generated pairs. MedSegFactory unlocks on-demand generation of paired medical images and segmentation masks through user-defined prompts that specify the target labels, imaging modalities, anatomical regions, and pathological conditions, facilitating scalable and high-quality data generation. This new paradigm of medical image synthesis enables seamless integration into diverse medical imaging workflows, enhancing both efficiency and accuracy. Extensive experiments show that MedSegFactory generates data of superior quality and usability, achieving competitive or state-of-the-art performance in 2D and 3D segmentation tasks while addressing data scarcity and regulatory constraints.

PanTS is a large-scale, multi-institutional dataset curated to advance research in pancreatic CT analysis. It contains 36,390 CT scans from 145 medical centers, with expert-validated, voxel-wise annotations of over 993,000 anatomical structures, covering pancreatic tumors, pancreas head, body, and tail, and 24 surrounding anatomical structures such as vascular/skeletal structures and abdominal/thoracic organs. Each scan includes metadata such as patient age, sex, diagnosis, contrast phase, in-plane spacing, slice thickness, etc. AI models trained on PanTS achieve significantly better performance in pancreatic tumor detection, localization, and segmentation than those trained on existing public datasets. Our analysis indicates that these gains are directly attributable to the 16× larger-scale tumor annotations and indirectly supported by the 24 additional surrounding anatomical structures. As the largest and most comprehensive resource of its kind, PanTS offers a new benchmark for developing and evaluating AI models in pancreatic CT analysis.

Widely adopted evaluation metrics for sparse-view CT reconstruction, such as Structural Similarity Index Measure and Peak Signal-to-Noise Ratio, prioritize pixel-wise fidelity but often fail to capture the completeness of critical anatomical structures, particularly small or thin regions that are easily missed. To address this limitation, we propose a suite of novel anatomy-aware evaluation metrics designed to assess structural completeness across anatomical structures, including large organs, small organs, intestines, and vessels. Building on these metrics, we introduce CARE, a Completeness-Aware Reconstruction Enhancement framework that incorporates structural penalties during training to encourage anatomical preservation of significant structures. CARE is model-agnostic and can be seamlessly integrated into analytical, implicit, and generative methods. When applied to these methods, CARE substantially improves structural completeness in CT reconstructions, achieving up to **32%** improvement for large organs, **22%** for small organs, **40%** for intestines, and **36%** for vessels.

X-ray, known for its ability to reveal internal structures of objects, is expected to provide richer information for 3D reconstruction than visible light. Yet, existing NeRF algorithms overlook this nature of X-ray, leading to their limitations in capturing structural contents of imaged objects. In this paper, we propose a framework, Structure-Aware X-ray Neural Radiodensity Fields (SAX-NeRF), for sparse-view X-ray 3D reconstruction. Firstly, we design a Line Segment-based Transformer (Lineformer) as the backbone of SAX-NeRF. Linefomer captures internal structures of objects in 3D space by modeling the dependencies within each line segment of an X-ray. Secondly, we present a Masked Local-Global (MLG) ray sampling strategy to extract contextual and geometric information in 2D projection. Plus, we collect a larger-scale dataset X3D covering wider X-ray applications. Experiments on X3D show that SAX-NeRF surpasses previous NeRF-based methods by 12.56 and 2.49 dB on novel view synthesis and CT reconstruction. https://github.com/caiyuanhao1998/SAX-NeRF

Tumor synthesis enables the creation of artificial tumors in medical images, facilitating the training of AI models for tumor detection and segmentation. However, success in tumor synthesis hinges on creating visually realistic tumors that are generalizable across multiple organs and, furthermore, the resulting AI models being capable of detecting real tumors in images sourced from different domains (e.g., hospitals). This paper made a progressive stride toward generalizable tumor synthesis by leveraging a critical observation: early-stage tumors (< 2cm) tend to have similar imaging characteristics in computed tomography (CT), whether they originate in the liver, pancreas, or kidneys. We have ascertained that generative AI models, e.g., Diffusion Models, can create realistic tumors generalized to a range of organs even when trained on a limited number of tumor examples from only one organ. Moreover, we have shown that AI models trained on these synthetic tumors can be generalized to detect and segment real tumors from CT volumes, encompassing a broad spectrum of patient demographics, imaging protocols, and healthcare facilities.

The pre-training and fine-tuning paradigm has become prominent in transfer learning. For example, if the model is pre-trained on ImageNet and then fine-tuned to PASCAL, it can significantly outperform that trained on PASCAL from scratch. While ImageNet pre-training has shown enormous success, it is formed in 2D, and the learned features are for classification tasks; when transferring to more diverse tasks, like 3D image segmentation, its performance is inevitably compromised due to the deviation from the original ImageNet context. A significant challenge lies in the lack of large, annotated 3D datasets rivaling the scale of ImageNet for model pre-training. To overcome this challenge, we make two contributions. Firstly, we construct AbdomenAtlas 1.1 that comprises **9,262** three-dimensional computed tomography (CT) volumes with high-quality, per-voxel annotations of 25 anatomical structures and pseudo annotations of seven tumor types. Secondly, we develop a suite of models that are pre-trained on our AbdomenAtlas 1.1 for transfer learning. Our preliminary analyses indicate that the model trained only with 21 CT volumes, 672 masks, and 40 GPU hours has a transfer learning ability similar to the model trained with 5,050 (unlabeled) CT volumes and 1,152 GPU hours. More importantly, the transfer learning ability of supervised models can further scale up with larger annotated datasets, achieving significantly better performance than preexisting pre-trained models, irrespective of their pre-training methodologies or data sources. We hope this study can facilitate collective efforts in constructing larger 3D medical datasets and more releases of supervised pre-trained models.

How can we test AI performance? This question seems trivial, but it isn't. Standard benchmarks often have problems such as in-distribution and small-size test sets, oversimplified metrics, unfair comparisons, and short-term outcome pressure. As a consequence, good performance on standard benchmarks does not guarantee success in real-world scenarios. To address these problems, we present Touchstone, a large-scale collaborative segmentation benchmark of 9 types of abdominal organs. This benchmark is based on 5,195 training CT scans from 76 hospitals around the world and 5,903 testing CT scans from 11 additional hospitals. This diverse test set enhances the statistical significance of benchmark results and rigorously evaluates AI algorithms across various out-of-distribution scenarios. We invited 14 inventors of 19 AI algorithms to train their algorithms, while our team, as a third party, independently evaluated these algorithms on three test sets. In addition, we also evaluated pre-existing AI frameworks---which, differing from algorithms, are more flexible and can support different algorithms—including MONAI from NVIDIA, nnU-Net from DKFZ, and numerous other open-source frameworks. We are committed to expanding this benchmark to encourage more innovation of AI algorithms for the medical domain.

The tracking-by-detection paradigm currently dominates multiple target tracking algorithms. It usually includes three tasks: target detection, appearance feature embedding, and data association. Carrying out these three tasks successively usually leads to lower tracking efficiency. In this paper, we propose a one-stage anchor-free multiple task learning framework which carries out target detection and appearance feature embedding in parallel to substantially increase the tracking speed. This framework simultaneously predicts a target detection and produces a feature embedding for each location, by sharing a pyramid of feature maps. We propose a deformable local attention module which utilizes the correlations between features at different locations within a target to obtain more discriminative features. We further propose a task-aware prediction module which utilizes deformable convolutions to select the most suitable locations for the different tasks. At the selected locations, classification of samples into foreground or background, appearance feature embedding, and target box regression are carried out. Two effective training strategies, regression range overlapping and sample reweighting, are proposed to reduce missed detections in dense scenes. Ambiguous samples whose identities are difficult to determine are effectively dealt with to obtain more accurate feature embedding of target appearance. An appearance-enhanced non-maximum suppression is proposed to reduce over-suppression of true targets in crowded scenes. Based on the one-stage anchor-free network with the deformable local attention module and the task-aware prediction module, we implement a new online multiple target tracker. Experimental results show that our tracker achieves a very fast speed while maintaining a high tracking accuracy.

Radiography imaging protocols focus on particular body regions, therefore producing images of great similarity and yielding recurrent anatomical structures across patients. Exploiting this structured information could potentially ease the detection of anomalies from radiography images. To this end, we propose a Simple Space-Aware Memory Matrix for In-painting and Detecting anomalies from radiography images (abbreviated as SimSID). We formulate anomaly detection as an image reconstruction task, consisting of a space-aware memory matrix and an in-painting block in the feature space. During the training, SimSID can taxonomize the ingrained anatomical structures into recurrent visual patterns, and in the inference, it can identify anomalies (unseen/modified visual patterns) from the test image. Our SimSID surpasses the state of the arts in unsupervised anomaly detection by +8.0%, +5.0%, and +9.9% AUC scores on ZhangLab, COVIDx, and CheXpert benchmark datasets, respectively.

Large deep learning models have shown great potential with state-of-the-art results in many tasks. However, running these large models is quite challenging on an accelerator (GPU or TPU) because the on-device memory is too limited for the size of these models. Intra-layer model parallelism is an approach to address the issues by partitioning individual layers or operators across multiple devices in a distributed accelerator cluster. But, the data communications generated by intra-layer model parallelism can contribute to a significant proportion of the overall execution time and severely hurt the computational efficiency.

We demonstrate that AI models can accurately segment liver tumors without the need for manual annotation by using synthetic tumors in CT scans. Our synthetic tumors have two intriguing advantages: (I) realistic in shape and texture, which even medical professionals can confuse with real tumors; (II) effective for training AI models, which can perform liver tumor segmentation similarly to the model trained on real tumors—this result is exciting because no existing work, using synthetic tumors only, has thus far reached a similar or even close performance to real tumors. This result also implies that manual efforts for annotating tumors voxel by voxel (which took years to create) can be significantly reduced in the future. Moreover, our synthetic tumors can automatically generate many examples of small (or even tiny) synthetic tumors and have the potential to im-prove the success rate of detecting small liver tumors, which is critical for detecting the early stages of cancer. In addition to enriching the training data, our synthesizing strategy also enables us to rigorously assess the AI robustness.

Robust and reliable semantic segmentation in complex scenes is crucial for many real-life applications such as autonomous safe driving and nighttime rescue. In most approaches, it is typical to make use of RGB images as input. They however work well only in preferred weather conditions; when facing adverse conditions such as rainy, overexposure, or low-light, they often fail to deliver satisfactory results. This has led to the recent investigation into multispectral semantic segmentation, where RGB and thermal infrared (RGBT) images are both utilized as input. This gives rise to significantly more robust segmentation of image objects in complex scenes and under adverse conditions. Nevertheless, the present focus in single RGBT image input restricts existing methods from well addressing dynamic real-world scenes. Motivated by the above observations, in this paper, we set out to address a relatively new task of semantic segmentation of multispectral video input, which we refer to as Multispectral Video Semantic Segmentation, or MVSS in short. An in-house MVSeg dataset is thus curated, consisting of 738 calibrated RGB and thermal videos, accompanied by 3,545 fine-grained pixel-level semantic annotations of 26 categories. Our dataset contains a wide range of challenging urban scenes in both daytime and nighttime. Moreover, we propose an effective MVSS baseline, dubbed MVNet, which is to our knowledge the first model to jointly learn semantic representations from multispectral and temporal contexts. Comprehensive experiments are conducted using various semantic segmentation models on the MVSeg dataset. Empirically, the engagement of multispectral video input is shown to lead to significant improvement in semantic segmentation; the effectiveness of our MVNet baseline has also been verified.

Radiography imaging protocols focus on particular body regions, therefore producing images of great similarity and yielding recurrent anatomical structures across patients. To exploit this structured information, we propose the use of Space-aware Memory Queues for In-painting and Detecting anomalies from radiography images (abbreviated as SQUID). We show that SQUID can taxonomize the ingrained anatomical structures into recurrent patterns; and in the inference, it can identify anomalies (unseen/modified patterns) in the image. SQUID surpasses 13 state-of-the-art methods in unsupervised anomaly detection by at least 5 points on two chest X-ray benchmark datasets measured by the Area Under the Curve (AUC). Additionally, we have created a new dataset (DigitAnatomy), which synthesizes the spatial correlation and consistent shape in chest anatomy. We hope DigitAnatomy can prompt the development, evaluation, and interpretability of anomaly detection methods.

An increasing number of public datasets have shown a marked impact on automated organ segmentation and tumor detection. However, due to the small size and partially labeled problem of each dataset, as well as a limited investigation of diverse types of tumors, the resulting models are often limited to segmenting specific organs/tumors and ignore the semantics of anatomical structures, nor can they be extended to novel domains. To address these issues, we propose the CLIP-Driven Universal Model, which incorporates text embedding learned from Contrastive Language-Image Pre-training (CLIP) to segmentation models. This CLIP-based label encoding captures anatomical relationships, enabling the model to learn a structured feature embedding and segment 25 organs and 6 types of tumors. The proposed model is developed from an assembly of 14 datasets, using a total of 3,410 CT scans for training and then evaluated on 6,162 external CT scans from 3 additional datasets. We rank first on the Medical Segmentation Decathlon (MSD) public leaderboard and achieve state-of-the-art results on Beyond The Cranial Vault (BTCV). Additionally, the Universal Model is computationally more efficient (6× faster) compared with dataset-specific models, generalized better to CT scans from varying sites, and shows stronger transfer learning performance on novel tasks.

In response to innovations in machine learning (ML) models, production workloads changed radically and rapidly. TPU v4 is the fifth Google domain specific architecture (DSA) and its third supercomputer for such ML models. Optical circuit switches (OCSes) dynamically reconfigure its interconnect topology to improve scale, availability, utilization, modularity, deployment, security, power, and performance; users can pick a twisted 3D torus topology if desired. Much cheaper, lower power, and faster than Infiniband, OCSes and underlying optical components are <5% of system cost and <3% of system power. Each TPU v4 includes SparseCores, dataflow processors that accelerate models that rely on embeddings by 5x--7x yet use only 5% of die area and power. Deployed since 2020, TPU v4 outperforms TPU v3 by 2.1x and improves performance/Watt by 2.7x. The TPU v4 supercomputer is 4x larger at 4096 chips and thus nearly 10x faster overall, which along with OCS flexibility and availability allows a large language model to train at an average of ~60% of peak FLOPS/second. For similar sized systems, it is ~4.3x--4.5x faster than the Graphcore IPU Bow and is 1.2x--1.7x faster and uses 1.3x--1.9x less power than the Nvidia A100. TPU v4s inside the energy-optimized warehouse scale computers of Google Cloud use ~2--6x less energy and produce ~20x less CO2e than contemporary DSAs in typical on-premise data centers.

Annotating medical images, particularly for organ segmentation, is laborious and time-consuming. For example, annotating an abdominal organ requires an estimated rate of 30-60 minutes per CT volume based on the expertise of an annotator and the size, visibility, and complexity of the organ. Therefore, publicly available datasets for multi-organ segmentation are often limited in data size and organ diversity. This paper proposes an active learning procedure to expedite the annotation process for organ segmentation and creates the largest multi-organ dataset (by far) with the spleen, liver, kidneys, stomach, gallbladder, pancreas, aorta, and IVC annotated in 8,448 CT volumes, equating to 3.2 million slices. The conventional annotation methods would take an experienced annotator up to 1,600 weeks (or roughly 30.8 years) to complete this task. In contrast, our annotation procedure has accomplished this task in three weeks (based on an 8-hour workday, five days a week) while maintaining a similar or even better annotation quality. This achievement is attributed to three unique properties of our method: (1) label bias reduction using multiple pre-trained segmentation models, (2) effective error detection in the model predictions, and (3) attention guidance for annotators to make corrections on the most salient errors. Furthermore, we summarize the taxonomy of common errors made by AI algorithms and annotators. This allows for continuous improvement of AI and annotations, significantly reducing the annotation costs required to create large-scale datasets for a wider variety of medical imaging tasks. Code and dataset are available at https://github.com/MrGiovanni/AbdomenAtlas

Vision Transformer (ViT) has become one of the most popular neural architectures due to its great scalability, computational efficiency, and compelling performance in many vision tasks. However, ViT has shown inferior performance to Convolutional Neural Network (CNN) on medical tasks due to its data-hungry nature and the lack of an-notated medical data. In this paper, we pre-train ViTs on 266,340 chest X-rays using Masked Autoencoders (MAE) which reconstruct missing pixels from a small part of each image. For comparison, CNNs are also pre-trained on the same 266,340 X-rays using advanced self-supervised methods (e.g. MoCo v2). The results show that our pre-trained ViT performs comparably (sometimes better) to the state-of-the-art CNN (DenseNet-121) for multi-label thorax dis-ease classification. This performance is attributed to the strong recipes extracted from our empirical studies for pre-training and fine-tuning ViT. The pre-training recipe signifies that medical reconstruction requires a much smaller proportion of an image (10% vs. 25%) and a more moderate random resized crop range (0.5∼1.0 vs. 0.2∼1.0) compared with natural imaging. Furthermore, we remark that in-domain transfer learning is preferred whenever possible. The fine-tuning recipe discloses that layer-wise LR decay, RandAug magnitude, and DropPath rate are significant factors to consider. We hope that this study can direct future research on the application of Transformers to a larger variety of medical imaging tasks.

Large language models have been shown to achieve remarkable performance across a variety of natural language tasks using few-shot learning, which drastically reduces the number of task-specific training examples needed to adapt the model to a particular application. To further our understanding of the impact of scale on few-shot learning, we trained a 540-billion parameter, densely activated, Transformer language model, which we call Pathways Language Model (PaLM). We trained PaLM on 6144 TPU v4 chips using Pathways, a new ML system which enables highly efficient training across multiple TPU Pods. We demonstrate continued benefits of scaling by achieving state-of-the-art few-shot learning results on hundreds of language understanding and generation benchmarks. On a number of these tasks, PaLM 540B achieves breakthrough performance, outperforming the finetuned state-of-the-art on a suite of multi-step reasoning tasks, and outperforming average human performance on the recently released BIG-bench benchmark. A significant number of BIG-bench tasks showed discontinuous improvements from model scale, meaning that performance steeply increased as we scaled to our largest model. PaLM also has strong capabilities in multilingual tasks and source code generation, which we demonstrate on a wide array of benchmarks. We additionally provide a comprehensive analysis on bias and toxicity, and study the extent of training data memorization with respect to model scale. Finally, we discuss the ethical considerations related to large language models and discuss potential mitigation strategies.

Semi-supervised video action recognition tends to enable deep neural networks to achieve remarkable performance even with very limited labeled data. However, existing methods are mainly transferred from current image-based methods (e.g., FixMatch). Without specifically utilizing the temporal dynamics and inherent multimodal attributes, their results could be suboptimal. To better leverage the encoded temporal information in videos, we introduce temporal gradient as an additional modality for more attentive feature extraction in this paper. To be specific, our method explicitly distills the fine-grained motion representations from temporal gradient (TG) and imposes consistency across different modalities (i.e., RGB and TG). The performance of semi-supervised action recognition is significantly improved without additional computation or parameters during inference. Our method achieves the state-of-the-art performance on three video action recognition benchmarks (i.e., Kinetics-400, UCF-101, and HMDB-51) under several typical semi-supervised settings (i.e., different ratios of labeled data). Code is made available at https://github.com/lambert-x/video-semisup.

Scaling language models with more data, compute and parameters has driven significant progress in natural language processing. For example, thanks to scaling, GPT-3 was able to achieve strong results on in-context learning tasks. However, training these large dense models requires significant amounts of computing resources. In this paper, we propose and develop a family of language models named \glam (\textbf{G}eneralist \textbf{La}nguage \textbf{M}odel), which uses a sparsely activated mixture-of-experts architecture to scale the model capacity while also incurring substantially less training cost compared to dense variants. The largest \glam has 1.2 trillion parameters, which is approximately 7x larger than GPT-3. It consumes only 1/3 of the energy used to train GPT-3 and requires half of the computation flops for inference, while still achieving better overall fewshot performance across 29 NLP tasks.

Google deployed several TPU generations since 2015, teaching us lessons that changed our views: semi-conductor technology advances unequally; compiler compatibility trumps binary compatibility, especially for VLIW domain-specific architectures (DSA); target total cost of ownership vs initial cost; support multi-tenancy; deep neural networks (DNN) grow 1.5X annually; DNN advances evolve workloads; some inference tasks require floating point; inference DSAs need air-cooling; apps limit latency, not batch size; and backwards ML compatibility helps deploy DNNs quickly. These lessons molded TPUv4i, an inference DSA deployed since 2020.

This paper introduces a new concept called "transferable visual words" (TransVW), aiming to achieve annotation efficiency for deep learning in medical image analysis. Medical imaging-focusing on particular parts of the body for defined clinical purposes-generates images of great similarity in anatomy across patients and yields sophisticated anatomical patterns across images, which are associated with rich semantics about human anatomy and which are natural visual words. We show that these visual words can be automatically harvested according to anatomical consistency via self-discovery, and that the self-discovered visual words can serve as strong yet free supervision signals for deep models to learn semantics-enriched generic image representation via self-supervision (self-classification and self-restoration). Our extensive experiments demonstrate the annotation efficiency of TransVW by offering higher performance and faster convergence with reduced annotation cost in several applications. Our TransVW has several important advantages, including (1) TransVW is a fully autodidactic scheme, which exploits the semantics of visual words for self-supervised learning, requiring no expert annotation; (2) visual word learning is an add-on strategy, which complements existing self-supervised methods, boosting their performance; and (3) the learned image representation is semantics-enriched models, which have proven to be more robust and generalizable, saving annotation efforts for a variety of applications through transfer learning. Our code, pre-trained models, and curated visual words are available at https://github.com/JLiangLab/TransVW.

Abstract The multi-dimensional assignment problem is universal for data association analysis such as data association-based visual multi-object tracking and multi-graph matching. In this paper, multi-dimensional assignment is formulated as a rank-1 tensor approximation problem. A dualL1-normalized context/hyper-context aware tensor power iteration optimization method is proposed. The method is applied to multi-object tracking and multi-graph matching. In the optimization method, tensor power iteration with the dual unit norm enables the capture of information across multiple sample sets. Interactions between sample associations are modeled as contexts or hyper-contexts which are combined with the global affinity into a unified optimization. The optimization is flexible for accommodating various types of contextual models. In multi-object tracking, the global affinity is defined according to the appearance similarity between objects detected in different frames. Interactions between objects are modeled as motion contexts which are encoded into the global association optimization. The tracking method integrates high order motion information and high order appearance variation. The multi-graph matching method carries out matching over graph vertices and structure matching over graph edges simultaneously. The matching consistency across multi-graphs is based on the high-order tensor optimization. Various types of vertex affinities and edge/hyper-edge affinities are flexibly integrated. Experiments on several public datasets, such as the MOT16 challenge benchmark, validate the effectiveness of the proposed methods.

The state-of-the-art models for medical image segmentation are variants of U-Net and fully convolutional networks (FCN). Despite their success, these models have two limitations: (1) their optimal depth is apriori unknown, requiring extensive architecture search or inefficient ensemble of models of varying depths; and (2) their skip connections impose an unnecessarily restrictive fusion scheme, forcing aggregation only at the same-scale feature maps of the encoder and decoder sub-networks. To overcome these two limitations, we propose UNet++, a new neural architecture for semantic and instance segmentation, by (1) alleviating the unknown network depth with an efficient ensemble of U-Nets of varying depths, which partially share an encoder and co-learn simultaneously using deep supervision; (2) redesigning skip connections to aggregate features of varying semantic scales at the decoder sub-networks, leading to a highly flexible feature fusion scheme; and (3) devising a pruning scheme to accelerate the inference speed of UNet++. We have evaluated UNet++ using six different medical image segmentation datasets, covering multiple imaging modalities such as computed tomography (CT), magnetic resonance imaging (MRI), and electron microscopy (EM), and demonstrating that (1) UNet++ consistently outperforms the baseline models for the task of semantic segmentation across different datasets and backbone architectures; (2) UNet++ enhances segmentation quality of varying-size objects-an improvement over the fixed-depth U-Net; (3) Mask RCNN++ (Mask R-CNN with UNet++ design) outperforms the original Mask R-CNN for the task of instance segmentation; and (4) pruned UNet++ models achieve significant speedup while showing only modest performance degradation. Our implementation and pre-trained models are available at https://github.com/MrGiovanni/UNetPlusPlus.

Person Re-identification is a very challenging task due to inter-class ambiguity caused by similar appearances, and large intra-class diversity caused by viewpoints, illuminations, and poses. To address these challenges, in this article, a graph convolution network based model for person re-identification is proposed to learn more discriminative feature embeddings, where a graph-structured relationship between person images and person parts are together integrated. Graph convolution networks extract common characteristics of the same person, while pyramid feature embedding exploits parts relations and learns stable representation with each person image. We achieve a very competitive performance respectively on three widely used datasets, indicating that the proposed approach significantly outperforms the baseline methods and achieves the state-of-the-art performance.

Generative adversarial networks (GANs) have ushered in a revolution in image-to-image translation. The development and proliferation of GANs raises an interesting question: can we train a GAN to remove an object, if present, from an image while otherwise preserving the image? Specifically, can a GAN "virtually heal" anyone by turning his medical image, with an unknown health status (diseased or healthy), into a healthy one, so that diseased regions could be revealed by subtracting those two images? Such a task requires a GAN to identify a minimal subset of target pixels for domain translation, an ability that we call fixed-point translation, which no GAN is equipped with yet. Therefore, we propose a new GAN, called Fixed-Point GAN, trained by (1) supervising same-domain translation through a conditional identity loss, and (2) regularizing cross-domain translation through revised adversarial, domain classification, and cycle consistency loss. Based on fixed-point translation, we further derive a novel framework for disease detection and localization using only image-level annotation. Qualitative and quantitative evaluations demonstrate that the proposed method outperforms the state of the art in multi-domain image-to-image translation and that it surpasses predominant weakly-supervised localization methods in both disease detection and localization. Implementation is available at https://github.com/jlianglab/Fixed-Point-GAN.

In this paper we formulate multi-target tracking (MTT) as a high-order graph matching problem and propose a l1-norm tensor power iteration solution. Concretely, the search for trajectory-observation correspondences in MTT task is cast as a hypergraph matching problem to maximize a multi-linear objective function over all permutations of the associations. This function is defined by a tensor representing the affinity between association tuples where pair-wise similarities, motion consistency and spatial structural information can be embedded expediently. To solve the matching problem, a dual-direction unit l1-norm constrained tensor power iteration algorithm is proposed. Additionally, as measuring the appearance affinity with features extracted from the rectangle patch, which is adopted in most methods, has a weak discrimination when bounding boxes overlap each other heavily, we present a deep pair-wise appearance similarity metric based on object mask in this paper where just the features from true target region are utilized. Experimental evaluation shows that our approach achieves an accuracy comparable to state-of-the-art online trackers. The source code of the proposed approach will be released to facilitate further studies on the MTT problem.

Intense interest in applying convolutional neural networks (CNNs) in biomedical image analysis is wide spread, but its success is impeded by the lack of large annotated datasets in biomedical imaging. Annotating biomedical images is not only tedious and time consuming, but also demanding of costly, specialty-oriented knowledge and skills, which are not easily accessible. To dramatically reduce annotation cost, this paper presents a novel method called AIFT (active, incremental fine-tuning) to naturally integrate active learning and transfer learning into a single framework. AIFT starts directly with a pre-trained CNN to seek "worthy" samples from the unannotated for annotation, and the (fine-tuned) CNN is further fine-tuned continuously by incorporating newly annotated samples in each iteration to enhance the CNN's performance incrementally. We have evaluated our method in three different biomedical imaging applications, demonstrating that the cost of annotation can be cut by at least half. This performance is attributed to the several advantages derived from the advanced active and incremental capability of our AIFT method.

Object detection plays a critical role for automatic video analysis in many vision applications. Background subtraction has been the mainstream in the field of moving objects detection. However, most of state‐of‐the‐art techniques of background subtraction operate on each pixel independently ignoring the global features of images. A motion detection method based on subspace update of background is proposed in this study. This method uses a subspace spanned by the principal components of background sequence to characterise the background and integrates the regional continuity of objects to segment the foreground. To deal with changes in the background geometry, a learning factor is introduced into the authors’ model to update the subspace timely. Additionally, to reduce computational complexity, they use two‐dimension principal component analysis (PCA) rather than traditional PCA to obtain the principal components of background. Experiments demonstrate that the update policy is effective and in most cases this proposed method can achieve better results than others compared in this study.

Low-rank matrix completion aims to recover a matrix from a small subset of its entries and has received much attention in the field of computer vision. Most existing methods formulate the task as a low-rank matrix approximation problem. A truncated nuclear norm has recently been proposed as a better approximation to the rank of matrix than a nuclear norm. The corresponding optimization method, truncated nuclear norm regularization (TNNR), converges better than the nuclear norm minimization-based methods. However, it is not robust to the number of subtracted singular values and requires a large number of iterations to converge. In this paper, a TNNR method based on weighted residual error (TNNR-WRE) for matrix completion and its extension model (ETNNR-WRE) are proposed. TNNR-WRE assigns different weights to the rows of the residual error matrix in an augmented Lagrange function to accelerate the convergence of the TNNR method. The ETNNR-WRE is much more robust to the number of subtracted singular values than the TNNR-WRE, TNNR alternating direction method of multipliers, and TNNR accelerated proximal gradient with Line search methods. Experimental results using both synthetic and real visual data sets show that the proposed TNNR-WRE and ETNNR-WRE methods perform better than TNNR and Iteratively Reweighted Nuclear Norm (IRNN) methods.

A trusted display service assures the confidentiality and authenticity of content output by a security-sensitive application and thus prevents a compromised commodity operating system or application from surreptitiously reading or modifying the displayed output. Past approaches have failed to provide trusted display on commodity platforms that use modern graphics processing units (GPUs). For example, full GPU virtualization encourages the sharing of GPU address space with multiple virtual machines {\em without} providing adequate hardware protection mechanisms; e.g., address-space separation and instruction execution control. This paper proposes a new trusted display service that has a minimal trusted code base and maintains full compatibility with commodity computing platforms. The service relies on a GPU separation kernel that (1) defines different types of GPU objects, (2) mediates access to security-sensitive objects, and (3) emulates object whenever required by commodity-platform compatibility. The separation kernel employs a new address-space separation mechanism that avoids the challenging problem of GPU instruction verification without adequate hardware support. The implementation of the trusted-display service has a code base that is two orders of magnitude smaller than other similar services, such as those based on full GPU virtualization. Performance measurements show that the trusted-display overhead added over and above that of the underlying trusted system is fairly modest.

To be trustworthy, security-sensitive applications must be formally verified and hence small and simple, i.e., wimpy. Thus, they cannot include a variety of basic services available only in large and untrustworthy commodity systems, i.e., in giants. Hence, wimps must securely compose with giants to survive on commodity systems, i.e., rely on giants' services but only after efficiently verifying their results. This paper presents a security architecture based on a wimpy kernel that provides on-demand isolated I/O channels for wimp applications, without bloating the underlying trusted computing base. The size and complexity of the wimpy kernel are minimized by safely outsourcing I/O subsystem functions to an untrusted commodity operating system and exporting driver and I/O subsystem code to wimp applications. Using the USB subsystem as a case study, this paper illustrates the dramatic reduction of wimpy-kernel size and complexity, e.g., over 99% of the USB code base is removed. Performance measurements indicate that the wimpy-kernel architecture exhibits the desired execution efficiency.

A trusted path is a protected channel that assures the secrecy and authenticity of data transfers between a user's input/output (I/O) device and a program trusted by that user. We argue that, despite its incontestable necessity, current commodity systems do not support trusted path with any significant assurance. This paper presents a hyper visor-based design that enables a trusted path to bypass an untrusted operating-system, applications, and I/O devices, with a minimal Trusted Computing Base (TCB). We also suggest concrete I/O architectural changes that will simplify future trusted-path system design. Our system enables users to verify the states and configurations of one or more trusted-paths using a simple, secret less, hand-held device. We implement a simple user-oriented trusted path as a case study.

Clear evidence indicates the existence of compromised routers in ISP and enterprise networks. Fault localization (FL) protocols enable a network to localize specific links of compromised routers sabotaging network data delivery and are recognized as an essential means to enhancing network availability in the face of targeted attacks. However, theoretically proven lower bounds have shown that secure FL protocols in the current network infrastructure inevitably incur prohibitive overhead. We observe the current limits are due to a lack of trust relationships among network nodes. We demonstrate that we can achieve much higher FL efficiency by leveraging trusted computing technology to design a trusted network-layer architecture, Tru eN et, with a small Trusted Computing Base (TCB). We intend Tru e N e t to serve as a case study that demonstrates trusted computing's ability in yielding tangible and measurable benefits for secure network protocol designs.

With the continual growth of network speed and the increasing sophistication of network applications, keeping network operations efficient and secure becomes more challenging. Pattern matching is one of the key technologies for content-ware network processing, such as traffic classification, application identification and intrusion prevention. In this paper, we propose a hybrid pattern matching algorithm optimized for multi-core network processing platforms. As a system-level solution, our scheme focuses on both performance stability and hardware/software co-design. To verify the effectiveness of our design, the proposed algorithm is implemented on a state-of-art 16-MIPS-core network processing platform and evaluated with real-life data sets. Experimental results show that, when compared with the traditional Aho-Corasick algorithm, our hybrid solution saves 60~95% memory space while guarantees stable performance on large pattern sets and against adverse test traffic.

An important security challenge is to protect the execution of security-sensitive code on legacy systems from malware that may infect the OS, applications, or system devices. Prior work experienced a tradeoff between the level of security achieved and efficiency. In this work, we leverage the features of modern processors from AMD and Intel to overcome the tradeoff to simultaneously achieve a high level of security and high performance. We present TrustVisor, a special-purpose hypervisor that provides code integrity as well as data integrity and secrecy for selected portions of an application. TrustVisor achieves a high level of security, first because it can protect sensitive code at a very fine granularity, and second because it has a very small code base (only around 6K lines of code) that makes verification feasible. TrustVisor can also attest the existence of isolated execution to an external entity. We have implemented TrustVisor to protect security-sensitive code blocks while imposing less than 7% overhead on the legacy OS and its applications in the common case.

To build high-performance network devices with holistic security protection, a large number of algorithms have been proposed. However, multi-core implementation of the existing algorithms suffers from three limitations: performance instability, data-structure heterogeneity, and hardware dependency. In this paper, we propose three principles for effective network processing on multi-core network processors. To verify the effectiveness of these principles, algorithms for two typical network processing tasks are redesigned and implemented on the Cavium Octeon3860 network processor. Test results show that our schemes achieve superior performance in comparison with existing best-known algorithms.

The security gateways today are required not only to block unauthorized accesses by authenticating packet headers, but also by inspecting connection states to defend against malicious intrusions. Hence session creation rate plays a key role in determining the overall performance of stateful intrusion prevention systems. In this paper, we propose a high-speed session creation scheme optimized for network processors. Main contribution includes: a) A high-performance flow classification algorithm on network processors; b) An efficient TCP three-way handshake scheme designed for fast-path processing using a two-stage intelligent hashing. Experimental results show that: a) The presented parallel optimized flow classification algorithm, Parallel Search Cross-Producting, outperforms the original Cross-Producting and Binary Search Cross-Producting algorithms with 300% and 60% increase of classification speed; b) The proposed fast path three-way handshake scheme, IntelliHash, achieves a TCP connection creation rate over 2M connections per second.