Fan Yang

Ningxin Zheng, Huiqiang Jiang, Quanlu Zhang et al. · microsoft-research

Dynamic sparsity, where the sparsity patterns are unknown until runtime, poses a significant challenge to deep learning. The state-of-the-art sparsity-aware deep learning solutions are restricted to pre-defined, static sparsity patterns due to significant overheads associated with preprocessing. Efficient execution of dynamic sparse computation often faces the misalignment between the GPU-friendly tile configuration for efficient execution and the sparsity-aware tile shape that minimizes coverage wastes (non-zero values in tensor). In this paper, we propose PIT, a deep-learning compiler for dynamic sparsity. PIT proposes a novel tiling mechanism that leverages Permutation Invariant Transformation (PIT), a mathematically proven property, to transform multiple sparsely located micro-tiles into a GPU-efficient dense tile without changing the computation results, thus achieving both high GPU utilization and low coverage waste. Given a model, PIT first finds feasible PIT rules for all its operators and generates efficient GPU kernels accordingly. At runtime, with the novel SRead and SWrite primitives, PIT rules can be executed extremely fast to support dynamic sparsity in an online manner. Extensive evaluation on diverse models shows that PIT can accelerate dynamic sparsity computation by up to 5.9x (average 2.43x) over state-of-the-art compilers.

Hongxiang Huang, Daihui Yang, Gang Dai et al.

The study of ancient writings has great value for archaeology and philology. Essential forms of material are photographic characters, but manual photographic character recognition is extremely time-consuming and expertise-dependent. Automatic classification is therefore greatly desired. However, the current performance is limited due to the lack of annotated data. Data generation is an inexpensive but useful solution for data scarcity. Nevertheless, the diverse glyph shapes and complex background textures of photographic ancient characters make the generation task difficult, leading to the unsatisfactory results of existing methods. In this paper, we propose an unsupervised generative adversarial network called AGTGAN. By the explicit global and local glyph shape style modeling followed by the stroke-aware texture transfer, as well as an associate adversarial learning mechanism, our method can generate characters with diverse glyphs and realistic textures. We evaluate our approach on the photographic ancient character datasets, e.g., OBC306 and CSDD. Our method outperforms the state-of-the-art approaches in various metrics and performs much better in terms of the diversity and authenticity of generated samples. With our generated images, experiments on the largest photographic oracle bone character dataset show that our method can achieve a significant increase in classification accuracy, up to 16.34%.

Mancheng Meng, Ziyan Wu, Terrence Chen et al.

Predicting the future trajectory of a person remains a challenging problem, due to randomness and subjectivity of human movement. However, the moving patterns of human in a constrained scenario typically conform to a limited number of regularities to a certain extent, because of the scenario restrictions and person-person or person-object interactivity. Thus, an individual person in this scenario should follow one of the regularities as well. In other words, a person's subsequent trajectory has likely been traveled by others. Based on this hypothesis, we propose to forecast a person's future trajectory by learning from the implicit scene regularities. We call the regularities, inherently derived from the past dynamics of the people and the environment in the scene, scene history. We categorize scene history information into two types: historical group trajectory and individual-surroundings interaction. To exploit these two types of information for trajectory prediction, we propose a novel framework Scene History Excavating Network (SHENet), where the scene history is leveraged in a simple yet effective approach. In particular, we design two components: the group trajectory bank module to extract representative group trajectories as the candidate for future path, and the cross-modal interaction module to model the interaction between individual past trajectory and its surroundings for trajectory refinement. In addition, to mitigate the uncertainty in ground-truth trajectory, caused by the aforementioned randomness and subjectivity of human movement, we propose to include smoothness into the training process and evaluation metrics. We conduct extensive evaluations to validate the efficacy of our proposed framework on ETH, UCY, as well as a new, challenging benchmark dataset PAV, demonstrating superior performance compared to state-of-the-art methods.

Nimrod Dvir, Elaine Friedman, Suraj Commuri et al.

This study introduces and empirically tests a novel predictive model for digital information engagement (IE) - the READ model, an acronym for the four pivotal attributes of engaging information: Representativeness, Ease-of-use, Affect, and Distribution. Conceptualized within the theoretical framework of Cumulative Prospect Theory, the model integrates key cognitive biases with computational linguistics and natural language processing to develop a multidimensional perspective on information engagement. A rigorous testing protocol was implemented, involving 50 randomly selected pairs of synonymous words (100 words in total) from the WordNet database. These words' engagement levels were evaluated through a large-scale online survey (n = 80,500) to derive empirical IE metrics. The READ attributes for each word were then computed and their predictive efficacy examined. The findings affirm the READ model's robustness, accurately predicting a word's IE level and distinguishing the more engaging word from a pair of synonyms with an 84% accuracy rate. The READ model's potential extends across various domains, including business, education, government, and healthcare, where it could enhance content engagement and inform AI language model development and generative text work. Future research should address the model's scalability and adaptability across different domains and languages, thereby broadening its applicability and efficacy.

Zhuyun Zhou, Zongwei Wu, Florian Bolli et al.

Autonomous racing has rapidly gained research attention. Traditionally, racing cars rely on 2D LiDAR as their primary visual system. In this work, we explore the integration of an event camera with the existing system to provide enhanced temporal information. Our goal is to fuse the 2D LiDAR data with event data in an end-to-end learning framework for steering prediction, which is crucial for autonomous racing. To the best of our knowledge, this is the first study addressing this challenging research topic. We start by creating a multisensor dataset specifically for steering prediction. Using this dataset, we establish a benchmark by evaluating various SOTA fusion methods. Our observations reveal that existing methods often incur substantial computational costs. To address this, we apply low-rank techniques to propose a novel, efficient, and effective fusion design. We introduce a new fusion learning policy to guide the fusion process, enhancing robustness against misalignment. Our fusion architecture provides better steering prediction than LiDAR alone, significantly reducing the RMSE from 7.72 to 1.28. Compared to the second-best fusion method, our work represents only 11% of the learnable parameters while achieving better accuracy. The source code, dataset, and benchmark will be released to promote future research.

Fan Yang, Xiaofei Wang

Using deep learning methods to detect students' classroom behavior automatically is a promising approach for analyzing their class performance and improving teaching effectiveness. However, the lack of publicly available spatio-temporal datasets on student behavior, as well as the high cost of manually labeling such datasets, pose significant challenges for researchers in this field. To address this issue, we proposed a method for extending the spatio-temporal behavior dataset in Student Classroom Scenarios (SCB-ST-Dataset4) through image dataset. Our SCB-ST-Dataset4 comprises 757265 images with 25810 labels, focusing on 3 behaviors: hand-raising, reading, writing. Our proposed method can rapidly generate spatio-temporal behavior datasets without requiring extra manual labeling. Furthermore, we proposed a Behavior Similarity Index (BSI) to explore the similarity of behaviors. We evaluated the dataset using the YOLOv5, YOLOv7, YOLOv8, and SlowFast algorithms, achieving a mean average precision (map) of up to 82.3%. Last, we fused multiple models to generate student behavior-related data from various perspectives. The experiment further demonstrates the effectiveness of our method. And SCB-ST-Dataset4 provides a robust foundation for future research in student behavior detection, potentially contributing to advancements in this field. The SCB-ST-Dataset4 is available for download at: https://github.com/Whiffe/SCB-dataset.

Yu Qi, Fan Yang, Yousong Zhu et al.

Autoregressive language modeling (ALM) have been successfully used in self-supervised pre-training in Natural language processing (NLP). However, this paradigm has not achieved comparable results with other self-supervised approach in computer vision (e.g., contrastive learning, mask image modeling). In this paper, we try to find the reason why autoregressive modeling does not work well on vision tasks. To tackle this problem, we fully analyze the limitation of visual autoregressive methods and proposed a novel stochastic autoregressive image modeling (named SAIM) by the two simple designs. First, we employ stochastic permutation strategy to generate effective and robust image context which is critical for vision tasks. Second, we create a parallel encoder-decoder training process in which the encoder serves a similar role to the standard vision transformer focus on learning the whole contextual information, and meanwhile the decoder predicts the content of the current position, so that the encoder and decoder can reinforce each other. By introducing stochastic prediction and the parallel encoder-decoder, SAIM significantly improve the performance of autoregressive image modeling. Our method achieves the best accuracy (83.9%) on the vanilla ViT-Base model among methods using only ImageNet-1K data. Transfer performance in downstream tasks also show that our model achieves competitive performance.

Jiageng Zhong, Ming Li, Yinliang Chen et al.

For intelligent quadcopter UAVs, a robust and reliable autonomous planning system is crucial. Most current trajectory planning methods for UAVs are suitable for static environments but struggle to handle dynamic obstacles, which can pose challenges and even dangers to flight. To address this issue, this paper proposes a vision-based planning system that combines tracking and trajectory prediction of dynamic obstacles to achieve efficient and reliable autonomous flight. We use a lightweight object detection algorithm to identify dynamic obstacles and then use Kalman Filtering to track and estimate their motion states. During the planning phase, we not only consider static obstacles but also account for the potential movements of dynamic obstacles. For trajectory generation, we use a B-spline-based trajectory search algorithm, which is further optimized with various constraints to enhance safety and alignment with the UAV's motion characteristics. We conduct experiments in both simulation and real-world environments, and the results indicate that our approach can successfully detect and avoid obstacles in dynamic environments in real-time, offering greater reliability compared to existing approaches. Furthermore, with the advancements in Natural Language Processing (NLP) technology demonstrating exceptional zero-shot generalization capabilities, more user-friendly human-machine interactions have become feasible, and this study also explores the integration of autonomous planning systems with Large Language Models (LLMs).

Senmao Cheng, Fan Yang, Zhi Chen et al.

Image stitching is to construct panoramic images with wider field of vision (FOV) from some images captured from different viewing positions. To solve the problem of fusion ghosting in the stitched image, seam-driven methods avoid the misalignment area to fuse images by predicting the best seam. Currently, as standard tools of the OpenCV library, dynamic programming (DP) and GraphCut (GC) are still the only commonly used seam prediction methods despite the fact that they were both proposed two decades ago. However, GC can get excellent seam quality but poor real-time performance while DP method has good efficiency but poor seam quality. In this paper, we propose a deep learning based seam prediction method (DSeam) for the sake of high seam quality with high efficiency. To overcome the difficulty of the seam description in network and no GroundTruth for training we design a selective consistency loss combining the seam shape constraint and seam quality constraint to supervise the network learning. By the constraint of the selection of consistency loss, we implicitly defined the mask boundaries as seams and transform seam prediction into mask prediction. To our knowledge, the proposed DSeam is the first deep learning based seam prediction method for image stitching. Extensive experimental results well demonstrate the superior performance of our proposed Dseam method which is 15 times faster than the classic GC seam prediction method in OpenCV 2.4.9 with similar seam quality.

Fan Yang, Tao Wang, Xiaofei Wang

Accurately detecting student behavior in classroom videos can aid in analyzing their classroom performance and improving teaching effectiveness. However, the current accuracy rate in behavior detection is low. To address this challenge, we propose the Student Classroom Behavior Detection system based on based on YOLOv7-BRA (YOLOv7 with Bi-level Routing Attention ). We identified eight different behavior patterns, including standing, sitting, speaking, listening, walking, raising hands, reading, and writing. We constructed a dataset, which contained 11,248 labels and 4,001 images, with an emphasis on the common behavior of raising hands in a classroom setting (Student Classroom Behavior dataset, SCB-Dataset). To improve detection accuracy, we added the biformer attention module to the YOLOv7 network. Finally, we fused the results from YOLOv7 CrowdHuman, SlowFast, and DeepSort models to obtain student classroom behavior data. We conducted experiments on the SCB-Dataset, and YOLOv7-BRA achieved an mAP@0.5 of 87.1%, resulting in a 2.2% improvement over previous results. Our SCB-dataset can be downloaded from: https://github.com/Whiffe/SCB-datase

Xuan He, Fan Yang, Kailun Yang et al.

Transformer-based methods have demonstrated superior performance for monocular 3D object detection recently, which aims at predicting 3D attributes from a single 2D image. Most existing transformer-based methods leverage both visual and depth representations to explore valuable query points on objects, and the quality of the learned query points has a great impact on detection accuracy. Unfortunately, existing unsupervised attention mechanisms in transformers are prone to generate low-quality query features due to inaccurate receptive fields, especially on hard objects. To tackle this problem, this paper proposes a novel "Supervised Scale-aware Deformable Attention" (SSDA) for monocular 3D object detection. Specifically, SSDA presets several masks with different scales and utilizes depth and visual features to adaptively learn a scale-aware filter for object query augmentation. Imposing the scale awareness, SSDA could well predict the accurate receptive field of an object query to support robust query feature generation. Aside from this, SSDA is assigned with a Weighted Scale Matching (WSM) loss to supervise scale prediction, which presents more confident results as compared to the unsupervised attention mechanisms. Extensive experiments on the KITTI and Waymo Open datasets demonstrate that SSDA significantly improves the detection accuracy, especially on moderate and hard objects, yielding state-of-the-art performance as compared to the existing approaches. Our code will be made publicly available at https://github.com/mikasa3lili/SSD-MonoDETR.

Zhi Chen, Fan Yang, Wenbing Tao

Point cloud registration is a fundamental step for many tasks. In this paper, we propose a neural network named DetarNet to decouple the translation $t$ and rotation $R$, so as to overcome the performance degradation due to their mutual interference in point cloud registration. First, a Siamese Network based Progressive and Coherent Feature Drift (PCFD) module is proposed to align the source and target points in high-dimensional feature space, and accurately recover translation from the alignment process. Then we propose a Consensus Encoding Unit (CEU) to construct more distinguishable features for a set of putative correspondences. After that, a Spatial and Channel Attention (SCA) block is adopted to build a classification network for finding good correspondences. Finally, the rotation is obtained by Singular Value Decomposition (SVD). In this way, the proposed network decouples the estimation of translation and rotation, resulting in better performance for both of them. Experimental results demonstrate that the proposed DetarNet improves registration performance on both indoor and outdoor scenes. Our code will be available in \url{https://github.com/ZhiChen902/DetarNet}.

Ao Luo, Fan Yang, Kunming Luo et al.

Estimating per-pixel motion between video frames, known as optical flow, is a long-standing problem in video understanding and analysis. Most contemporary optical flow techniques largely focus on addressing the cross-image matching with feature similarity, with few methods considering how to explicitly reason over the given scene for achieving a holistic motion understanding. In this work, taking a fresh perspective, we introduce a novel graph-based approach, called adaptive graph reasoning for optical flow (AGFlow), to emphasize the value of scene/context information in optical flow. Our key idea is to decouple the context reasoning from the matching procedure, and exploit scene information to effectively assist motion estimation by learning to reason over the adaptive graph. The proposed AGFlow can effectively exploit the context information and incorporate it within the matching procedure, producing more robust and accurate results. On both Sintel clean and final passes, our AGFlow achieves the best accuracy with EPE of 1.43 and 2.47 pixels, outperforming state-of-the-art approaches by 11.2% and 13.6%, respectively.