Document pre-trained models and grid-based models have proven to be very effective on various tasks in Document AI. However, for the document layout analysis (DLA) task, existing document pre-trained models, even those pre-trained in a multi-modal fashion, usually rely on either textual features or visual features. Grid-based models for DLA are multi-modality but largely neglect the effect of pre-training. To fully leverage multi-modal information and exploit pre-training techniques to learn better representation for DLA, in this paper, we present VGT, a two-stream Vision Grid Transformer, in which Grid Transformer (GiT) is proposed and pre-trained for 2D token-level and segment-level semantic understanding. Furthermore, a new dataset named D^4LA, which is so far the most diverse and detailed manually-annotated benchmark for document layout analysis, is curated and released. Experiment results have illustrated that the proposed VGT model achieves new state-of-the-art results on DLA tasks, e.g. PubLayNet (95.7% to 96.2%), DocBank (79.6% to 84.1%), and D^4LA (67.7% to 68.8%). The code and models as well as the D4LA dataset will be made publicly available.
论文检索
输入标题、作者或关键词，从 2,156 篇学术成果中精准定位
In this paper, we present a method for converting a given scene image into a sketch using different types and multiple levels of abstraction. We distinguish between two types of abstraction. The first considers the fidelity of the sketch, varying its representation from a more precise portrayal of the input to a looser depiction. The second is defined by the visual simplicity of the sketch, moving from a detailed depiction to a sparse sketch. Using an explicit disentanglement into two abstraction axes --- and multiple levels for each one --- provides users additional control over selecting the desired sketch based on their personal goals and preferences. To form a sketch at a given level of fidelity and simplification, we train two MLP networks. The first network learns the desired placement of strokes, while the second network learns to gradually remove strokes from the sketch without harming its recognizability and semantics. Our approach is able to generate sketches of complex scenes including those with complex backgrounds (e.g. natural and urban settings) and subjects (e.g. animals and people) while depicting gradual abstractions of the input scene in terms of fidelity and simplicity.
Safety-critical applications such as autonomous vehicles and social robots require fast computation and accurate probability density estimation on trajectory prediction. To address both requirements, this paper presents a new normalizing flow-based trajectory prediction model named FlowChain. FlowChain is a stack of conditional continuously-indexed flows (CIFs) that are expressive and allow analytical probability density computation. This analytical computation is faster than the generative models that need additional approximations such as kernel density estimation. Moreover, FlowChain is more accurate than the Gaussian mixture-based models due to fewer assumptions on the estimated density. FlowChain also allows a rapid update of estimated probability densities. This update is achieved by adopting the newest observed position and reusing the flow transformations and its log-det-jacobians that represent the motion trend. This update is completed in less than one millisecond because this reuse greatly omits the computational cost. Experimental results showed our FlowChain achieved state-of-the-art trajectory prediction accuracy compared to previous methods. Furthermore, our FlowChain demonstrated superiority in the accuracy and speed of density estimation. Our code is available at https://github.com/meaten/FlowChain-ICCV2023
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.
While large-scale image-text pretrained models such as CLIP have been used for multiple video-level tasks on trimmed videos, their use for temporal localization in untrimmed videos is still a relatively unexplored task. We design a new approach for this called UnLoc, which uses pretrained image and text towers, and feeds tokens to a video-text fusion model. The output of the fusion module are then used to construct a feature pyramid in which each level connects to a head to predict a per-frame relevancy score and start/end time displacements. Unlike previous works, our architecture enables Moment Retrieval, Temporal Localization, and Action Segmentation with a single stage model, without the need for action proposals, motion based pretrained features or representation masking. Unlike specialized models, we achieve state of the art results on all three different localization tasks with a unified approach. Code is available at: https://github.com/google-research/scenic.
Recent implicit neural representations have shown great results for novel view synthesis. However, existing methods require expensive per-scene optimization from many views hence limiting their application to real-world unbounded urban settings where the objects of interest or backgrounds are observed from very few views. To mitigate this challenge, we introduce a new approach called NeO 360, Neural fields for sparse view synthesis of outdoor scenes. NeO 360 is a generalizable method that reconstructs 360deg scenes from a single or a few posed RGB images. The essence of our approach is in capturing the distribution of complex real-world outdoor 3D scenes and using a hybrid image-conditional triplanar representation that can be queried from any world point. Our representation combines the best of both voxel-based and bird's-eye-view (BEV) representations and is more effective and expressive than each. NeO 360's representation allows us to learn from a large collection of unbounded 3D scenes while offering generalizability to new views and novel scenes from as few as a single image during inference. We demonstrate our approach on the proposed challenging 360deg unbounded dataset, called NeRDS 360, and show that NeO 360 outperforms state-of-the-art generalizable methods for novel view synthesis while also offering editing and composition capabilities. Project page: zubair-irshad.github.io/projects/neo360.html
DiffFit: Unlocking Transferability of Large Diffusion Models via Simple Parameter-efficient Fine-Tuning
PDF ↗Diffusion models have proven to be highly effective in generating high-quality images. However, adapting large pre-trained diffusion models to new domains remains an open challenge, which is critical for real-world applications. This paper proposes DiffFit, a parameter-efficient strategy to fine-tune large pre-trained diffusion models that enable fast adaptation to new domains. DiffFit is embarrassingly simple that only fine-tunes the bias term and newly-added scaling factors in specific layers, yet resulting in significant training speed-up and reduced model storage costs. Compared with full fine-tuning, DiffFit achieves 2x training speed-up and only needs to store approximately 0.12% of the total model parameters. Intuitive theoretical analysis has been provided to justify the efficacy of scaling factors on fast adaptation. On 8 downstream datasets, DiffFit achieves superior or competitive performances compared to the full fine-tuning while being more efficient. Remarkably, we show that DiffFit can adapt a pre-trained low-resolution generative model to a high-resolution one by adding minimal cost. Among diffusion-based methods, DiffFit sets a new state-of-the-art FID of 3.02 on ImageNet 512x512 benchmark by fine-tuning only 25 epochs from a public pre-trained ImageNet 256x256 checkpoint while being 30x more training efficient than the closest competitor.
Conditional generative models typically demand large annotated training sets to achieve high-quality synthesis. As a result, there has been significant interest in designing models that perform plug-and-play generation, i.e., to use a predefined or pretrained model, which is not explicitly trained on the generative task, to guide the generative process (e.g., using language). However, such guidance is typically useful only towards synthesizing high-level semantics rather than editing fine-grained details as in image-to-image translation tasks. To this end, and capitalizing on the powerful fine-grained generative control offered by the recent diffusion-based generative models, we introduce Steered Diffusion, a generalized framework for photorealistic zero-shot conditional image generation using a diffusion model trained for unconditional generation. The key idea is to steer the image generation of the diffusion model at inference time via designing a loss using a pre-trained inverse model that characterizes the conditional task. This loss modulates the sampling trajectory of the diffusion process. Our framework allows for easy incorporation of multiple conditions during inference. We present experiments using steered diffusion on several tasks including inpainting, colorization, text-guided semantic editing, and image super-resolution. Our results demonstrate clear qualitative and quantitative improvements over state-of-the-art diffusion-based plug-and-play models while adding negligible additional computational cost.
PODIA-3D: Domain Adaptation of 3D Generative Model Across Large Domain Gap Using Pose-Preserved Text-to-Image Diffusion
PDF ↗Recently, significant advancements have been made in 3D generative models, however training these models across diverse domains is challenging and requires an huge amount of training data and knowledge of pose distribution. Text-guided domain adaptation methods have allowed the generator to be adapted to the target domains using text prompts, thereby obviating the need for assembling numerous data. Recently, DATID-3D presents impressive quality of samples in text-guided domain, preserving diversity in text by leveraging text-to-image diffusion. However, adapting 3D generators to domains with significant domain gaps from the source domain still remains challenging due to issues in current text-to-image diffusion models as following: 1) shape-pose trade-off in diffusion-based translation, 2) pose bias, and 3) instance bias in the target domain, resulting in inferior 3D shapes, low text-image correspondence, and low intra-domain diversity in the generated samples. To address these issues, we propose a novel pipeline called PODIA-3D, which uses pose-preserved text-to-image diffusion-based domain adaptation for 3D generative models. We construct a pose-preserved text-to-image diffusion model that allows the use of extremely high-level noise for significant domain changes. We also propose specialized-to-general sampling strategies to improve the details of the generated samples. Moreover, to overcome the instance bias, we introduce a text-guided debiasing method that improves intra-domain diversity. Consequently, our method successfully adapts 3D generators across significant domain gaps. Our qualitative results and user study demonstrate that our approach outperforms existing 3D text-guided domain adaptation methods in terms of text-image correspondence, realism, diversity of rendered images, and sense of depth of 3D shapes in the generated samples.
PlankAssembly: Robust 3D Reconstruction from Three Orthographic Views with Learnt Shape Programs
PDF ↗In this paper, we develop a new method to automatically convert 2D line drawings from three orthographic views into 3D CAD models. Existing methods for this problem reconstruct 3D models by back-projecting the 2D observations into 3D space while maintaining explicit correspondence between the input and output. Such methods are sensitive to errors and noises in the input, thus often fail in practice where the input drawings created by human designers are imperfect. To overcome this difficulty, we leverage the attention mechanism in a Transformer-based sequence generation model to learn flexible mappings between the input and output. Further, we design shape programs which are suitable for generating the objects of interest to boost the reconstruction accuracy and facilitate CAD modeling applications. Experiments on a new benchmark dataset show that our method significantly outperforms existing ones when the inputs are noisy or incomplete.
Federated learning enables multiple decentralized clients to learn collaboratively without sharing local data. However, the expensive annotation cost on local clients remains an obstacle in utilizing local data. In this paper, we propose a federated active learning paradigm to efficiently learn a global model with a limited annotation budget while protecting data privacy in a decentralized learning manner. The main challenge faced by federated active learning is the mismatch between the active sampling goal of the global model on the server and that of the asynchronous local clients. This becomes even more significant when data is distributed non-IID across local clients. To address the aforementioned challenge, we propose Knowledge-Aware Federated Active Learning (KAFAL), which consists of Knowledge-Specialized Active Sampling (KSAS) and Knowledge-Compensatory Federated Update (KCFU). Specifically, KSAS is a novel active sampling method tailored for the federated active learning problem, aiming to deal with the mismatch challenge by sampling actively based on the discrepancies between local and global models. KSAS intensifies specialized knowledge in local clients, ensuring the sampled data is informative for both the local clients and the global model. Meanwhile, KCFU deals with the client heterogeneity caused by limited data and non-IID data distributions by compensating for each client's ability in weak classes with the assistance of the global model. Extensive experiments and analyses are conducted to show the superiority of KAFAL over recent state-of-the-art active learning methods. Code is available at https://github.com/ycao5602/KAFAL.
This paper proposes a shape anchor guided learning strategy (AncLearn) for robust holistic indoor scene understanding. We observe that the search space constructed by current methods for proposal feature grouping and instance point sampling often introduces massive noise to instance detection and mesh reconstruction. Accordingly, we develop AncLearn to generate anchors that dynamically fit instance surfaces to (i) unmix noise and target-related features for offering reliable proposals at the detection stage, and (ii) reduce outliers in object point sampling for directly providing well-structured geometry priors without segmentation during reconstruction. We embed AncLearn into a reconstruction-from-detection learning system (AncRec) to generate high-quality semantic scene models in a purely instance-oriented manner. Experiments conducted on the ScanNetv2 dataset (with ground truths from Scan2CAD and SceneCAD) demonstrate that our shape anchor-based method consistently achieves state-of-the-art performance in terms of 3D object detection, layout estimation, and shape reconstruction.
Vision transformers have achieved leading performance on various visual tasks yet still suffer from high computational complexity. The situation deteriorates in dense prediction tasks like semantic segmentation, as high-resolution inputs and outputs usually imply more tokens involved in computations. Directly removing the less attentive tokens has been discussed for the image classification task but can not be extended to semantic segmentation since a dense prediction is required for every patch. To this end, this work introduces a Dynamic Token Pruning (DToP) method based on the early exit of tokens for semantic segmentation. Motivated by the coarse-to-fine segmentation process by humans, we naturally split the widely adopted auxiliary-loss-based network architecture into several stages, where each auxiliary block grades every token's difficulty level. We can finalize the prediction of easy tokens in advance without completing the entire forward pass. Moreover, we keep k highest confidence tokens for each semantic category to uphold the representative context information. Thus, computational complexity will change with the difficulty of the input, akin to the way humans do segmentation. Experiments suggest that the proposed DToP architecture reduces on average 20% 35% of computational cost for current semantic segmentation methods based on plain vision transformers without accuracy degradation. The code is available through the following link: https://github.com/zbwxp/Dynamic-Token-Pruning.
Mitigating Adversarial Vulnerability through Causal Parameter Estimation by Adversarial Double Machine Learning
PDF ↗Adversarial examples derived from deliberately crafted perturbations on visual inputs can easily harm decision process of deep neural networks. To prevent potential threats, various adversarial training-based defense methods have grown rapidly and become a de facto standard approach for robustness. Despite recent competitive achievements, we observe that adversarial vulnerability varies across targets and certain vulnerabilities remain prevalent. Intriguingly, such peculiar phenomenon cannot be relieved even with deeper architectures and advanced defense methods. To address this issue, in this paper, we introduce a causal approach called Adversarial Double Machine Learning (ADML), which allows us to quantify the degree of adversarial vulnerability for network predictions and capture the effect of treatments on outcome of interests. ADML can directly estimate causal parameter of adversarial perturbations per se and mitigate negative effects that can potentially damage robustness, bridging a causal perspective into the adversarial vulnerability. Through extensive experiments on various CNN and Transformer architectures, we corroborate that ADML improves adversarial robustness with large margins and relieve the empirical observation.
Temporal-Coded Spiking Neural Networks with Dynamic Firing Threshold: Learning with Event-Driven Backpropagation
PDF ↗Spiking Neural Networks (SNNs) offer a highly promising computing paradigm due to their biological plausibility, exceptional spatiotemporal information processing capability and low power consumption. As a temporal encoding scheme for SNNs, Time-To-First-Spike (TTFS) encodes information using the timing of a single spike, which allows spiking neurons to transmit information through sparse spike trains and results in lower power consumption and higher computational efficiency compared to traditional rate-based encoding counterparts. However, despite the advantages of the TTFS encoding scheme, the effective and efficient training of TTFS-based deep SNNs remains a significant and open research problem. In this work, we first examine the factors underlying the limitations of applying existing TTFS-based learning algorithms to deep SNNs. Specifically, we investigate issues related to over-sparsity of spikes and the complexity of finding the `causal set'. We then propose a simple yet efficient dynamic firing threshold (DFT) mechanism for spiking neurons to address these issues. Building upon the proposed DFT mechanism, we further introduce a novel direct training algorithm for TTFS-based deep SNNs, called DTA-TTFS. This method utilizes event-driven processing and spike timing to enable efficient learning of deep SNNs. The proposed training method was validated on the image classification task and experimental results clearly demonstrate that our proposed method achieves state-of-the-art accuracy in comparison to existing TTFS-based learning algorithms, while maintaining high levels of sparsity and energy efficiency on neuromorphic inference accelerator.
Part-prototype networks (e.g., ProtoPNet, ProtoTree, and ProtoPool) have attracted broad research interest for their intrinsic interpretability and comparable accuracy to non-interpretable counterparts. However, recent works find that the interpretability from prototypes is fragile, due to the semantic gap between the similarities in the feature space and that in the input space. In this work, we strive to address this challenge by making the first attempt to quantitatively and objectively evaluate the interpretability of the part-prototype networks. Specifically, we propose two evaluation metrics, termed as "consistency score" and "stability score", to evaluate the explanation consistency across images and the explanation robustness against perturbations, respectively, both of which are essential for explanations taken into practice. Furthermore, we propose an elaborated part-prototype network with a shallow-deep feature alignment (SDFA) module and a score aggregation (SA) module to improve the interpretability of prototypes. We conduct systematical evaluation experiments and provide substantial discussions to uncover the interpretability of existing part-prototype networks. Experiments on three benchmarks across nine architectures demonstrate that our model achieves significantly superior performance to the state of the art, in both the accuracy and interpretability. Our code is available at https://github.com/hqhQAQ/EvalProtoPNet.
Self-supervised monocular scene flow estimation, aiming to understand both 3D structures and 3D motions from two temporally consecutive monocular images, has received increasing attention for its simple and economical sensor setup. However, the accuracy of current methods suffers from the bottleneck of less-efficient network architecture and lack of motion rigidity for regularization. In this paper, we propose a superior model named EMR-MSF by borrowing the advantages of network architecture design under the scope of supervised learning. We further impose explicit and robust geometric constraints with an elaborately constructed ego-motion aggregation module where a rigidity soft mask is proposed to filter out dynamic regions for stable ego-motion estimation using static regions. Moreover, we propose a motion consistency loss along with a mask regularization loss to fully exploit static regions. Several efficient training strategies are integrated including a gradient detachment technique and an enhanced view synthesis process for better performance. Our proposed method outperforms the previous self-supervised works by a large margin and catches up to the performance of supervised methods. On the KITTI scene flow benchmark, our approach improves the SF-all metric of the state-of-the-art self-supervised monocular method by 44% and demonstrates superior performance across sub-tasks including depth and visual odometry, amongst other self-supervised single-task or multi-task methods.
We address the problem of network calibration adjusting miscalibrated confidences of deep neural networks. Many approaches to network calibration adopt a regularization-based method that exploits a regularization term to smooth the miscalibrated confidences. Although these approaches have shown the effectiveness on calibrating the networks, there is still a lack of understanding on the underlying principles of regularization in terms of network calibration. We present in this paper an in-depth analysis of existing regularization-based methods, providing a better understanding on how they affect to network calibration. Specifically, we have observed that 1) the regularization-based methods can be interpreted as variants of label smoothing, and 2) they do not always behave desirably. Based on the analysis, we introduce a novel loss function, dubbed ACLS, that unifies the merits of existing regularization methods, while avoiding the limitations. We show extensive experimental results for image classification and semantic segmentation on standard benchmarks, including CIFAR10, Tiny-ImageNet, ImageNet, and PASCAL VOC, demonstrating the effectiveness of our loss function.
Object localization in general environments is a fundamental part of vision systems. While dominating on the COCO benchmark, recent Transformer-based detection methods are not competitive in diverse domains. Moreover, these methods still struggle to very accurately estimate the object bounding boxes in complex environments. We introduce Cascade-DETR for high-quality universal object detection. We jointly tackle the generalization to diverse domains and localization accuracy by proposing the Cascade Attention layer, which explicitly integrates object-centric information into the detection decoder by limiting the attention to the previous box prediction. To further enhance accuracy, we also revisit the scoring of queries. Instead of relying on classification scores, we predict the expected IoU of the query, leading to substantially more well-calibrated confidences. Lastly, we introduce a universal object detection benchmark, UDB10, that contains 10 datasets from diverse domains. While also advancing the state-of-the-art on COCO, Cascade-DETR substantially improves DETR-based detectors on all datasets in UDB10, even by over 10 mAP in some cases. The improvements under stringent quality requirements are even more pronounced. Our code and pretrained models are at https://github.com/SysCV/cascade-detr.
Knowledge distillation (KD) is a promising approach that facilitates the compact student model to learn dark knowledge from the huge teacher model for better results. Although KD methods are well explored in the 2D detection task, existing approaches are not suitable for 3D monocular detection without considering spatial cues. Motivated by the potential of depth information, we propose a novel distillation framework that validly improves the performance of the student model without extra depth labels. Specifically, we first put forward a perspective-induced feature imitation, which utilizes the perspective principle (the farther the smaller) to facilitate the student to imitate more features of farther objects from the teacher model. Moreover, we construct a depth-guided matrix by the predicted depth gap of teacher and student to facilitate the model to learn more knowledge of farther objects in prediction level distillation. The proposed method is available for advanced monocular detectors with various backbones, which also brings no extra inference time. Extensive experiments on the KITTI and nuScenes benchmarks with diverse settings demonstrate that the proposed method outperforms the state-of-the-art KD methods.