Long-term time series forecasting (LTSF) has been widely applied in finance, traffic prediction, and other domains. Recently, patch-based transformers have emerged as a promising approach, segmenting data into sub-level patches that serve as input tokens. However, existing methods mostly rely on predetermined patch lengths, necessitating expert knowledge and posing challenges in capturing diverse characteristics across various scales. Moreover, time series data exhibit diverse variations and fluctuations across different temporal scales, which traditional approaches struggle to model effectively. In this paper, we propose a dynamic tokenizer with a dynamic sparse learning algorithm to capture diverse receptive fields and sparse patterns of time series data. In order to build hierarchical receptive fields, we develop a multi-scale Transformer model, coupled with multi-scale sequence extraction, capable of capturing multi-resolution features. Additionally, we introduce a group-aware rotary position encoding technique to enhance intra- and inter-group position awareness among representations across different temporal scales. Our proposed model, named DRFormer, is evaluated on various real-world datasets, and experimental results demonstrate its superiority compared to existing methods. Our code is available at: https://github.com/ruixindingECNU/DRFormer.
More and more end-to-end text spotting methods based on Transformer architecture have demonstrated superior performance. These methods utilize a bipartite graph matching algorithm to perform one-to-one optimal matching between predicted objects and actual objects. However, the instability of bipartite graph matching can lead to inconsistent optimization targets, thereby affecting the training performance of the model. Existing literature applies denoising training to solve the problem of bipartite graph matching instability in object detection tasks. Unfortunately, this denoising training method cannot be directly applied to text spotting tasks, as these tasks need to perform irregular shape detection tasks and more complex text recognition tasks than classification. To address this issue, we propose a novel denoising training method (DNTextSpotter) for arbitrary-shaped text spotting. Specifically, we decompose the queries of the denoising part into noised positional queries and noised content queries. We use the four Bezier control points of the Bezier center curve to generate the noised positional queries. For the noised content queries, considering that the output of the text in a fixed positional order is not conducive to aligning position with content, we employ a masked character sliding method to initialize noised content queries, thereby assisting in the alignment of text content and position. To improve the model's perception of the background, we further utilize an additional loss function for background characters classification in the denoising training part.Although DNTextSpotter is conceptually simple, it outperforms the state-of-the-art methods on four benchmarks (Total-Text, SCUT-CTW1500, ICDAR15, and Inverse-Text), especially yielding an improvement of 11.3% against the best approach in Inverse-Text dataset.
Thermal point clouds integrate thermal radiation and laser point clouds effectively. However, the semantic information for the interpretation of building thermal point clouds can hardly be precisely inferred. Transferring the semantics encapsulated in 3D building models at LoD3 has a potential to fill this gap. In this work, we propose a workflow enriching thermal point clouds with the geo-position and semantics of LoD3 building models, which utilizes features of both modalities: The proposed method can automatically co-register the point clouds from different sources and enrich the thermal point cloud in facade-detailed semantics. The enriched thermal point cloud supports thermal analysis and can facilitate the development of currently scarce deep learning models operating directly on thermal point clouds.
Echocardiography video is a primary modality for diagnosing heart diseases, but the limited data poses challenges for both clinical teaching and machine learning training. Recently, video generative models have emerged as a promising strategy to alleviate this issue. However, previous methods often relied on holistic conditions during generation, hindering the flexible movement control over specific cardiac structures. In this context, we propose an explainable and controllable method for echocardiography video generation, taking an initial frame and a motion curve as guidance. Our contributions are three-fold. First, we extract motion information from each heart substructure to construct motion curves, enabling the diffusion model to synthesize customized echocardiography videos by modifying these curves. Second, we propose the structure-to-motion alignment module, which can map semantic features onto motion curves across cardiac structures. Third, The position-aware attention mechanism is designed to enhance video consistency utilizing Gaussian masks with structural position information. Extensive experiments on three echocardiography datasets show that our method outperforms others regarding fidelity and consistency. The full code will be released at https://github.com/mlmi-2024-72/ECM.
This study presents a comprehensive analysis of the YOLOv5 object detection model, examining its architecture, training methodologies, and performance. Key components, including the Cross Stage Partial backbone and Path Aggregation-Network, are explored in detail. The paper reviews the model's performance across various metrics and hardware platforms. Additionally, the study discusses the transition from Darknet to PyTorch and its impact on model development. Overall, this research provides insights into YOLOv5's capabilities and its position within the broader landscape of object detection and why it is a popular choice for constrained edge deployment scenarios.
Research in 3D mapping is crucial for smart city applications, yet the cost of acquiring 3D data often hinders progress. Visual localization, particularly monocular camera position estimation, offers a solution by determining the camera's pose solely through visual cues. However, this task is challenging due to limited data from a single camera. To tackle these challenges, we organized the AISG-SLA Visual Localization Challenge (VLC) at IJCAI 2023 to explore how AI can accurately extract camera pose data from 2D images in 3D space. The challenge attracted over 300 participants worldwide, forming 50+ teams. Winning teams achieved high accuracy in pose estimation using images from a car-mounted camera with low frame rates. The VLC dataset is available for research purposes upon request via [email omitted].
This paper investigates the estimation problem of the pose (orientation and position) and linear velocity of a rigid body, as well as the landmark positions, using an inertial measurement unit (IMU) and a monocular camera. First, we propose a globally exponentially stable (GES) linear time-varying (LTV) observer for the estimation of body-frame landmark positions and velocity, using IMU and monocular bearing measurements. Thereafter, using the gyro measurements, some landmarks known in the inertial frame and the estimates from the LTV observer, we propose a nonlinear pose observer on $\SO(3)\times \mathbb{R}^3$. The overall estimation system is shown to be almost globally asymptotically stable (AGAS) using the notion of almost global input-to-state stability (ISS). Interestingly, we show that with the knowledge (in the inertial frame) of a small number of landmarks, we can recover (under some conditions) the unknown positions (in the inertial frame) of a large number of landmarks. Numerical simulation results are presented to illustrate the performance of the proposed estimation scheme.
This research presents a novel, analytical, Koopman Operator based formulation for position and attitude dynamics which can be used to derive control strategies for underactuated systems. Compared to data driven Koopman based techniques, the analytical approach presented in this work is model based and allows for an exact linear representation of the original nonlinear position and attitude dynamics. In fact, the resulting infinite dimensional model, defined in the lifted state space, is linear in the autonomous component and state dependent in the control. A boundary study is carried on to define the range of validity of the finite truncation of the Koopman based model followed by a controllability and stabilizability analysis to show the feasibility of employing the derived model for control system design. Compared to existing literature formulation, the presented model results in a better approximation of the original dyanmics using a more compact truncation of the lifted state space. Moreover, the model is derived using the Koopman approach on the entirety of the dynamics and does not require the need of angular velocity dynamic compensation. A case study involving an underactuated quadrotor unmanned aerial vehicle (UAV) is provided to show that, for practical use, a truncated subset of the infinite dimensional model, embeds most of the original nonlinear dynamics and can be used to design linear control strategies in the lifted space which results in nonlinear controllers in the original state space. The main advantages of the presented approach reside in the effective use of linear control strategies for nonlinear plats and the solution of the underactuation problem employing a single control loop.
Current methods commonly utilize three-branch structures of inversion, reconstruction, and editing, to tackle consistent image editing task. However, these methods lack control over the generation position of the edited object and have issues with background preservation. To overcome these limitations, we propose a tuning-free method with only two branches: inversion and editing. This approach allows users to simultaneously edit the object's action and control the generation position of the edited object. Additionally, it achieves improved background preservation. Specifically, we transfer the edited object information to the target area and repair or preserve the background of other areas during the inversion process at a specific time step. In the editing stage, we use the image features in self-attention to query the key and value of the corresponding time step in the inversion to achieve consistent image editing. Impressive image editing results and quantitative evaluation demonstrate the effectiveness of our method. The code is available at https://github.com/mobiushy/move-act.
Sparse query-based paradigms have achieved significant success in multi-view 3D detection for autonomous vehicles. Current research faces challenges in balancing between enlarging receptive fields and reducing interference when aggregating multi-view features. Moreover, different poses of cameras present challenges in training global attention models. To address these problems, this paper proposes a divided view method, in which features are modeled globally via the visibility crossattention mechanism, but interact only with partial features in a divided local virtual space. This effectively reduces interference from other irrelevant features and alleviates the training difficulties of the transformer by decoupling the position embedding from camera poses. Additionally, 2D historical RoI features are incorporated into the object-centric temporal modeling to utilize highlevel visual semantic information. The model is trained using a one-to-many assignment strategy to facilitate stability. Our framework, named DVPE, achieves state-of-the-art performance (57.2% mAP and 64.5% NDS) on the nuScenes test set. Codes will be available at https://github.com/dop0/DVPE.
Temporal information is crucial for detecting occluded instances. Existing temporal representations have progressed from BEV or PV features to more compact query features. Compared to these aforementioned features, predictions offer the highest level of abstraction, providing explicit information. In the context of online vectorized HD map construction, this unique characteristic of predictions is potentially advantageous for long-term temporal modeling and the integration of map priors. This paper introduces PrevPredMap, a pioneering temporal modeling framework that leverages previous predictions for constructing online vectorized HD maps. We have meticulously crafted two essential modules for PrevPredMap: the previous-predictions-based query generator and the dynamic-position-query decoder. Specifically, the previous-predictions-based query generator is designed to separately encode different types of information from previous predictions, which are then effectively utilized by the dynamic-position-query decoder to generate current predictions. Furthermore, we have developed a dual-mode strategy to ensure PrevPredMap's robust performance across both single-frame and temporal modes. Extensive experiments demonstrate that PrevPredMap achieves state-of-the-art performance on the nuScenes and Argoverse2 datasets. Code will be available at https://github.com/pnnnnnnn/PrevPredMap.
We prove the availability of inner product form position encoding in the state space dual algorithm and study the effectiveness of different position embeddings in the hybrid quadratic causal self-attention and state space dual algorithms. We propose inner function attention with dynamic mask, which can improve the expressiveness of the attention algorithm and avoid the sequence noise significantly affecting the accuracy of the attention score. We also design cross domain mixture of experts, which can improve the granularity of the sparse activation feedforward network while maintaining the efficiency of parameter utilization and retrieval. The combination of these methods constitutes our foundation model architecture: Wonderful Matrices. We conduct experiments on the language modeling task and find that Wonderful Matrices are more efficient and effective in handling complex language tasks.
In this paper, we present TAPTRv2, a Transformer-based approach built upon TAPTR for solving the Tracking Any Point (TAP) task. TAPTR borrows designs from DEtection TRansformer (DETR) and formulates each tracking point as a point query, making it possible to leverage well-studied operations in DETR-like algorithms. TAPTRv2 improves TAPTR by addressing a critical issue regarding its reliance on cost-volume,which contaminates the point query\'s content feature and negatively impacts both visibility prediction and cost-volume computation. In TAPTRv2, we propose a novel attention-based position update (APU) operation and use key-aware deformable attention to realize. For each query, this operation uses key-aware attention weights to combine their corresponding deformable sampling positions to predict a new query position. This design is based on the observation that local attention is essentially the same as cost-volume, both of which are computed by dot-production between a query and its surrounding features. By introducing this new operation, TAPTRv2 not only removes the extra burden of cost-volume computation, but also leads to a substantial performance improvement. TAPTRv2 surpasses TAPTR and achieves state-of-the-art performance on many challenging datasets, demonstrating the superiority
Contemporary medical contrastive learning faces challenges from inconsistent semantics and sample pair morphology, leading to dispersed and converging semantic shifts. The variability in text reports, due to multiple authors, complicates semantic consistency. To tackle these issues, we propose a two-step approach. Initially, text reports are converted into a standardized triplet format, laying the groundwork for our novel concept of ``observations'' and ``verdicts''. This approach refines the {Entity, Position, Exist} triplet into binary questions, guiding towards a clear ``verdict''. We also innovate in visual pre-training with a Meijering-based masking, focusing on features representative of medical images' local context. By integrating this with our text conversion method, our model advances cross-modal representation in a multimodal contrastive learning framework, setting new benchmarks in medical image analysis.
The recent work of Horikawa and Nakagawa (2024) claims that under a complete market admitting statistical arbitrage, the difference between the hedging position provided by deep hedging and that of the replicating portfolio is a statistical arbitrage. This raises concerns as it entails that deep hedging can include a speculative component aimed simply at exploiting the structure of the risk measure guiding the hedging optimisation problem. We test whether such finding remains true in a GARCH-based market model, which is an illustrative case departing from complete market dynamics. We observe that the difference between deep hedging and delta hedging is a speculative overlay if the risk measure considered does not put sufficient relative weight on adverse outcomes. Nevertheless, a suitable choice of risk measure can prevent the deep hedging agent from engaging in speculation.
In this technical report, we detail our first-place solution for the 2024 Waymo Open Dataset Challenge's semantic segmentation track. We significantly enhanced the performance of Point Transformer V3 on the Waymo benchmark by implementing cutting-edge, plug-and-play training and inference technologies. Notably, our advanced version, Point Transformer V3 Extreme, leverages multi-frame training and a no-clipping-point policy, achieving substantial gains over the original PTv3 performance. Additionally, employing a straightforward model ensemble strategy further boosted our results. This approach secured us the top position on the Waymo Open Dataset semantic segmentation leaderboard, markedly outperforming other entries.
The long-context capability of the Large Language Models (LLM) has made significant breakthroughs, but the maximum supported context length remains a critical bottleneck limiting their practical applications. The constraint of context length in LLMs arises from the self-attention mechanism, which cannot effectively and efficiently capture the semantic relationships within infinitely long contexts via the limited pre-trained positional information and attention scope. In this work, we propose \textbf{ReAttention}, a training-free approach enabling LLM based on the self-attention mechanism to support an infinite context with a finite attention scope under sufficient memory resources. ReAttention performs the position-agnostic top-$k$ attention before the ordinary position-aware self-attention, freeing LLMs from the length extrapolation issue. We validate the performance of ReAttention on the LongBench, L-Eval, and InfiniteBench and demonstrate that it is on par with traditional methods. Furthermore, we also apply ReAttention on mainstream LLMs, including LLaMA3.1-8B and Mistral-v0.3-7B, enabling them to support context lengths of at least 1M and even expanding the context length of LLaMA3.2-3B-chat by 128$\times$ to 4M without any further training in Needle-In-A-Haystack tests. We also improve the efficiency of ReAttention with Triton and achieve an efficient extrapolation without additional overhead.
3D visual grounding aims to identify objects in 3D point cloud scenes that match specific natural language descriptions. This requires the model to not only focus on the target object itself but also to consider the surrounding environment to determine whether the descriptions are met. Most previous works attempt to accomplish both tasks within the same module, which can easily lead to a distraction of attention. To this end, we propose PD-APE, a dual-branch decoding framework that separately decodes target object attributes and surrounding layouts. Specifically, in the target object branch, the decoder processes text tokens that describe features of the target object (e.g., category and color), guiding the queries to pay attention to the target object itself. In the surrounding branch, the queries align with other text tokens that carry surrounding environment information, making the attention maps accurately capture the layout described in the text. Benefiting from the proposed dual-branch design, the queries are allowed to focus on points relevant to each branch's specific objective. Moreover, we design an adaptive position encoding method for each branch respectively. In the target object branch, the position encoding relies on the relative positions between seed points and predicted 3D boxes. In the surrounding branch, the attention map is additionally guided by the confidence between visual and text features, enabling the queries to focus on points that have valuable layout information. Extensive experiments demonstrate that we surpass the state-of-the-art on two widely adopted 3D visual grounding datasets, ScanRefer and Nr3D.
How well will current interpretability techniques generalize to future models? A relevant case study is Mamba, a recent recurrent architecture with scaling comparable to Transformers. We adapt pre-Mamba techniques to Mamba and partially reverse-engineer the circuit responsible for the Indirect Object Identification (IOI) task. Our techniques provide evidence that 1) Layer 39 is a key bottleneck, 2) Convolutions in layer 39 shift names one position forward, and 3) The name entities are stored linearly in Layer 39's SSM. Finally, we adapt an automatic circuit discovery tool, positional Edge Attribution Patching, to identify a Mamba IOI circuit. Our contributions provide initial evidence that circuit-based mechanistic interpretability tools work well for the Mamba architecture.
Decentralized Finance (DeFi) has reshaped the possibilities of reserve banking in the form of the Collateralized Debt Position (CDP). Key to the safety of CDPs is the money supply architecture that enables issued debt to maintain its value. In traditional markets, and with respect to the United States Dollar system, interest rates are set by the Federal Reserve in an attempt to influence the effects of excessive inflation. DeFi enables a more transparent approach that typically relies on interest rates or other debt recovery mechanisms being directly informed by asset price. This research investigates contemporary DeFi money supply and debt management strategies and their limitations. Furthermore, this paper introduces a time-weighted approach to interest rate management that implements a Proportional-Integral-Derivative control system to constantly adapt to market activities and protect the value of issued currency, while addressing observed limitations.