Jinrong Yang

Yinhao Li, Han Bao, Zheng Ge et al.

Bounded by the inherent ambiguity of depth perception, contemporary camera-based 3D object detection methods fall into the performance bottleneck. Intuitively, leveraging temporal multi-view stereo (MVS) technology is the natural knowledge for tackling this ambiguity. However, traditional attempts of MVS are flawed in two aspects when applying to 3D object detection scenes: 1) The affinity measurement among all views suffers expensive computation cost; 2) It is difficult to deal with outdoor scenarios where objects are often mobile. To this end, we introduce an effective temporal stereo method to dynamically select the scale of matching candidates, enable to significantly reduce computation overhead. Going one step further, we design an iterative algorithm to update more valuable candidates, making it adaptive to moving candidates. We instantiate our proposed method to multi-view 3D detector, namely BEVStereo. BEVStereo achieves the new state-of-the-art performance (i.e., 52.5% mAP and 61.0% NDS) on the camera-only track of nuScenes dataset. Meanwhile, extensive experiments reflect our method can deal with complex outdoor scenarios better than contemporary MVS approaches. Codes have been released at https://github.com/Megvii-BaseDetection/BEVStereo.

Jinrong Yang, Songtao Liu, Zeming Li et al.

Autonomous driving requires the model to perceive the environment and (re)act within a low latency for safety. While past works ignore the inevitable changes in the environment after processing, streaming perception is proposed to jointly evaluate the latency and accuracy into a single metric for video online perception. In this paper, instead of searching trade-offs between accuracy and speed like previous works, we point out that endowing real-time models with the ability to predict the future is the key to dealing with this problem. We build a simple and effective framework for streaming perception. It equips a novel DualFlow Perception module (DFP), which includes dynamic and static flows to capture the moving trend and basic detection feature for streaming prediction. Further, we introduce a Trend-Aware Loss (TAL) combined with a trend factor to generate adaptive weights for objects with different moving speeds. Our simple method achieves competitive performance on Argoverse-HD dataset and improves the AP by 4.9% compared to the strong baseline, validating its effectiveness. Our code will be made available at https://github.com/yancie-yjr/StreamYOLO.

Jinrong Yang, Lin Song, Songtao Liu et al.

Many point-based 3D detectors adopt point-feature sampling strategies to drop some points for efficient inference. These strategies are typically based on fixed and handcrafted rules, making it difficult to handle complicated scenes. Different from them, we propose a Dynamic Ball Query (DBQ) network to adaptively select a subset of input points according to the input features, and assign the feature transform with a suitable receptive field for each selected point. It can be embedded into some state-of-the-art 3D detectors and trained in an end-to-end manner, which significantly reduces the computational cost. Extensive experiments demonstrate that our method can increase the inference speed by 30%-100% on KITTI, Waymo, and ONCE datasets. Specifically, the inference speed of our detector can reach 162 FPS on KITTI scene, and 30 FPS on Waymo and ONCE scenes without performance degradation. Due to skipping the redundant points, some evaluation metrics show significant improvements. Codes will be released at https://github.com/yancie-yjr/DBQ-SSD.

Jing Hao, Moyun Liu, Jinrong Yang et al.

Detecting glass regions is a challenging task due to the inherent ambiguity in their transparency and reflective characteristics. Current solutions in this field remain rooted in conventional deep learning paradigms, requiring the construction of annotated datasets and the design of network architectures. However, the evident drawback with these mainstream solutions lies in the time-consuming and labor-intensive process of curating datasets, alongside the increasing complexity of model structures. In this paper, we propose to address these issues by fully harnessing the capabilities of two existing vision foundation models (VFMs): Stable Diffusion and Segment Anything Model (SAM). Firstly, we construct a Synthetic but photorealistic large-scale Glass Surface Detection dataset, dubbed S-GSD, without any labour cost via Stable Diffusion. This dataset consists of four different scales, consisting of 168k images totally with precise masks. Besides, based on the powerful segmentation ability of SAM, we devise a simple Glass surface sEgMentor named GEM, which follows the simple query-based encoder-decoder architecture. Comprehensive experiments are conducted on the large-scale glass segmentation dataset GSD-S. Our GEM establishes a new state-of-the-art performance with the help of these two VFMs, surpassing the best-reported method GlassSemNet with an IoU improvement of 2.1%. Additionally, extensive experiments demonstrate that our synthetic dataset S-GSD exhibits remarkable performance in zero-shot and transfer learning settings. Codes, datasets and models are publicly available at: https://github.com/isbrycee/GEM

Runpei Dong, Chunrui Han, Yuang Peng et al. · tsinghua

This paper presents DreamLLM, a learning framework that first achieves versatile Multimodal Large Language Models (MLLMs) empowered with frequently overlooked synergy between multimodal comprehension and creation. DreamLLM operates on two fundamental principles. The first focuses on the generative modeling of both language and image posteriors by direct sampling in the raw multimodal space. This approach circumvents the limitations and information loss inherent to external feature extractors like CLIP, and a more thorough multimodal understanding is obtained. Second, DreamLLM fosters the generation of raw, interleaved documents, modeling both text and image contents, along with unstructured layouts. This allows DreamLLM to learn all conditional, marginal, and joint multimodal distributions effectively. As a result, DreamLLM is the first MLLM capable of generating free-form interleaved content. Comprehensive experiments highlight DreamLLM's superior performance as a zero-shot multimodal generalist, reaping from the enhanced learning synergy. Project page: https://dreamllm.github.io.

Yinhao Li, Zheng Ge, Guanyi Yu et al.

In this research, we propose a new 3D object detector with a trustworthy depth estimation, dubbed BEVDepth, for camera-based Bird's-Eye-View (BEV) 3D object detection. Our work is based on a key observation -- depth estimation in recent approaches is surprisingly inadequate given the fact that depth is essential to camera 3D detection. Our BEVDepth resolves this by leveraging explicit depth supervision. A camera-awareness depth estimation module is also introduced to facilitate the depth predicting capability. Besides, we design a novel Depth Refinement Module to counter the side effects carried by imprecise feature unprojection. Aided by customized Efficient Voxel Pooling and multi-frame mechanism, BEVDepth achieves the new state-of-the-art 60.9% NDS on the challenging nuScenes test set while maintaining high efficiency. For the first time, the NDS score of a camera model reaches 60%.

Chunrui Han, Jinrong Yang, Jianjian Sun et al. · tsinghua

Long-term temporal fusion is a crucial but often overlooked technique in camera-based Bird's-Eye-View (BEV) 3D perception. Existing methods are mostly in a parallel manner. While parallel fusion can benefit from long-term information, it suffers from increasing computational and memory overheads as the fusion window size grows. Alternatively, BEVFormer adopts a recurrent fusion pipeline so that history information can be efficiently integrated, yet it fails to benefit from longer temporal frames. In this paper, we explore an embarrassingly simple long-term recurrent fusion strategy built upon the LSS-based methods and find it already able to enjoy the merits from both sides, i.e., rich long-term information and efficient fusion pipeline. A temporal embedding module is further proposed to improve the model's robustness against occasionally missed frames in practical scenarios. We name this simple but effective fusing pipeline VideoBEV. Experimental results on the nuScenes benchmark show that VideoBEV obtains strong performance on various camera-based 3D perception tasks, including object detection (55.4\% mAP and 62.9\% NDS), segmentation (48.6\% vehicle mIoU), tracking (54.8\% AMOTA), and motion prediction (0.80m minADE and 0.463 EPA).

Liang Zhao, En Yu, Zheng Ge et al. · tsinghua

Human-AI interactivity is a critical aspect that reflects the usability of multimodal large language models (MLLMs). However, existing end-to-end MLLMs only allow users to interact with them through language instructions, leading to the limitation of the interactive accuracy and efficiency. In this study, we present precise referring instructions that utilize diverse reference representations such as points and boxes as referring prompts to refer to the special region. This enables MLLMs to focus on the region of interest and achieve finer-grained interaction. Based on precise referring instruction, we propose ChatSpot, a unified end-to-end multimodal large language model that supports diverse forms of interactivity including mouse clicks, drag-and-drop, and drawing boxes, which provides a more flexible and seamless interactive experience. We also construct a multi-grained vision-language instruction-following dataset based on existing datasets and GPT-4 generating. Furthermore, we design a series of evaluation tasks to assess the effectiveness of region recognition and interaction. Experimental results showcase ChatSpot's promising performance.

Zhuoling Li, Chunrui Han, Zheng Ge et al.

Efficiency is quite important for 3D lane detection due to practical deployment demand. In this work, we propose a simple, fast, and end-to-end detector that still maintains high detection precision. Specifically, we devise a set of fully convolutional heads based on row-wise classification. In contrast to previous counterparts, ours supports recognizing both vertical and horizontal lanes. Besides, our method is the first one to perform row-wise classification in bird-eye-view. In the heads, we split feature into multiple groups and every group of feature corresponds to a lane instance. During training, the predictions are associated with lane labels using the proposed single-win one-to-one matching to compute loss, and no post-processing operation is demanded for inference. In this way, our proposed fully convolutional detector, GroupLane, realizes end-to-end detection like DETR. Evaluated on 3 real world 3D lane benchmarks, OpenLane, Once-3DLanes, and OpenLane-Huawei, GroupLane adopting ConvNext-Base as the backbone outperforms the published state-of-the-art PersFormer by 13.6% F1 score in the OpenLane validation set. Besides, GroupLane with ResNet18 still surpasses PersFormer by 4.9% F1 score, while the inference speed is nearly 7x faster and the FLOPs is only 13.3% of it.

En Yu, Liang Zhao, Yana Wei et al.

Humans possess the remarkable ability to foresee the future to a certain extent based on present observations, a skill we term as foresight minds. However, this capability remains largely under explored within existing Multimodal Large Language Models (MLLMs), hindering their capacity to learn the fundamental principles of how things operate and the intentions behind the observed subjects. To address this issue, we introduce the integration of future modeling into the existing learning frameworks of MLLMs. By utilizing the subject trajectory, a highly structured representation of a consecutive frame sequence, as a learning objective, we aim to bridge the gap between the past and the future. We propose two innovative methods to empower MLLMs with foresight minds, Foresight Pre-Training (FPT) and Foresight Instruction-Tuning (FIT), which are inspired by the modern learning paradigm of LLMs. Specifically, FPT jointly training various tasks centered on trajectories, enabling MLLMs to learn how to attend and predict entire trajectories from a given initial observation. Then, FIT requires MLLMs to first predict trajectories of related objects and then reason about potential future events based on them. Aided by FPT and FIT, we build a novel and unified MLLM named Merlin that supports multi-images input and analysis about potential actions of multiple objects for the future reasoning. Experimental results show Merlin powerful foresight minds with impressive performance on both future reasoning and visual comprehension tasks.

En Yu, Songtao Liu, Zhuoling Li et al.

Although existing multi-object tracking (MOT) algorithms have obtained competitive performance on various benchmarks, almost all of them train and validate models on the same domain. The domain generalization problem of MOT is hardly studied. To bridge this gap, we first draw the observation that the high-level information contained in natural language is domain invariant to different tracking domains. Based on this observation, we propose to introduce natural language representation into visual MOT models for boosting the domain generalization ability. However, it is infeasible to label every tracking target with a textual description. To tackle this problem, we design two modules, namely visual context prompting (VCP) and visual-language mixing (VLM). Specifically, VCP generates visual prompts based on the input frames. VLM joints the information in the generated visual prompts and the textual prompts from a pre-defined Trackbook to obtain instance-level pseudo textual description, which is domain invariant to different tracking scenes. Through training models on MOT17 and validating them on MOT20, we observe that the pseudo textual descriptions generated by our proposed modules improve the generalization performance of query-based trackers by large margins.

Jinrong Yang, En Yu, Zeming Li et al.

3D Multi-Object Tracking (MOT) has achieved tremendous achievement thanks to the rapid development of 3D object detection and 2D MOT. Recent advanced works generally employ a series of object attributes, e.g., position, size, velocity, and appearance, to provide the clues for the association in 3D MOT. However, these cues may not be reliable due to some visual noise, such as occlusion and blur, leading to tracking performance bottleneck. To reveal the dilemma, we conduct extensive empirical analysis to expose the key bottleneck of each clue and how they correlate with each other. The analysis results motivate us to efficiently absorb the merits among all cues, and adaptively produce an optimal tacking manner. Specifically, we present Location and Velocity Quality Learning, which efficiently guides the network to estimate the quality of predicted object attributes. Based on these quality estimations, we propose a quality-aware object association (QOA) strategy to leverage the quality score as an important reference factor for achieving robust association. Despite its simplicity, extensive experiments indicate that the proposed strategy significantly boosts tracking performance by 2.2% AMOTA and our method outperforms all existing state-of-the-art works on nuScenes by a large margin. Moreover, QTrack achieves 48.0% and 51.1% AMOTA tracking performance on the nuScenes validation and test sets, which significantly reduces the performance gap between pure camera and LiDAR based trackers.

Jinrong Yang, Songtao Liu, Zeming Li et al.

The perceptive models of autonomous driving require fast inference within a low latency for safety. While existing works ignore the inevitable environmental changes after processing, streaming perception jointly evaluates the latency and accuracy into a single metric for video online perception, guiding the previous works to search trade-offs between accuracy and speed. In this paper, we explore the performance of real time models on this metric and endow the models with the capacity of predicting the future, significantly improving the results for streaming perception. Specifically, we build a simple framework with two effective modules. One is a Dual Flow Perception module (DFP). It consists of dynamic flow and static flow in parallel to capture moving tendency and basic detection feature, respectively. Trend Aware Loss (TAL) is the other module which adaptively generates loss weight for each object with its moving speed. Realistically, we consider multiple velocities driving scene and further propose Velocity-awared streaming AP (VsAP) to jointly evaluate the accuracy. In this realistic setting, we design a efficient mix-velocity training strategy to guide detector perceive any velocities. Our simple method achieves the state-of-the-art performance on Argoverse-HD dataset and improves the sAP and VsAP by 4.7% and 8.2% respectively compared to the strong baseline, validating its effectiveness.

Jinrong Yang, Tiancai Wang, Zheng Ge et al.

The great progress of 3D object detectors relies on large-scale data and 3D annotations. The annotation cost for 3D bounding boxes is extremely expensive while the 2D ones are easier and cheaper to collect. In this paper, we introduce a hybrid training framework, enabling us to learn a visual 3D object detector with massive 2D (pseudo) labels, even without 3D annotations. To break through the information bottleneck of 2D clues, we explore a new perspective: Temporal 2D Supervision. We propose a temporal 2D transformation to bridge the 3D predictions with temporal 2D labels. Two steps, including homography wraping and 2D box deduction, are taken to transform the 3D predictions into 2D ones for supervision. Experiments conducted on the nuScenes dataset show strong results (nearly 90% of its fully-supervised performance) with only 25% 3D annotations. We hope our findings can provide new insights for using a large number of 2D annotations for 3D perception.

Weixin Mao, Jinrong Yang, Zheng Ge et al.

Depth perception is a crucial component of monoc-ular 3D detection tasks that typically involve ill-posed problems. In light of the success of sample mining techniques in 2D object detection, we propose a simple yet effective mining strategy for improving depth perception in 3D object detection. Concretely, we introduce a plain metric to evaluate the quality of depth predictions, which chooses the mined sample for the model. Moreover, we propose a Gradient-aware and Model-perceive Mining strategy (GMM) for depth learning, which exploits the predicted depth quality for better depth learning through easy mining. GMM is a general strategy that can be readily applied to several state-of-the-art monocular 3D detectors, improving the accuracy of depth prediction. Extensive experiments on the nuScenes dataset demonstrate that the proposed methods significantly improve the performance of 3D object detection while outperforming other state-of-the-art sample mining techniques by a considerable margin. On the nuScenes benchmark, GMM achieved the state-of-the-art (42.1% mAP and 47.3% NDS) performance in monocular object detection.

Yinhao Li, Jinrong Yang, Jianjian Sun et al.

Bounded by the inherent ambiguity of depth perception, contemporary multi-view 3D object detection methods fall into the performance bottleneck. Intuitively, leveraging temporal multi-view stereo (MVS) technology is the natural knowledge for tackling this ambiguity. However, traditional attempts of MVS has two limitations when applying to 3D object detection scenes: 1) The affinity measurement among all views suffers expensive computational cost; 2) It is difficult to deal with outdoor scenarios where objects are often mobile. To this end, we propose BEVStereo++: by introducing a dynamic temporal stereo strategy, BEVStereo++ is able to cut down the harm that is brought by introducing temporal stereo when dealing with those two scenarios. Going one step further, we apply Motion Compensation Module and long sequence Frame Fusion to BEVStereo++, which shows further performance boosting and error reduction. Without bells and whistles, BEVStereo++ achieves state-of-the-art(SOTA) on both Waymo and nuScenes dataset.

Haoran Wei, Lingyu Kong, Jinyue Chen et al.

Modern Large Vision-Language Models (LVLMs) enjoy the same vision vocabulary -- CLIP, which can cover most common vision tasks. However, for some special vision task that needs dense and fine-grained vision perception, e.g., document-level OCR or chart understanding, especially in non-English scenarios, the CLIP-style vocabulary may encounter low efficiency in tokenizing the vision knowledge and even suffer out-of-vocabulary problem. Accordingly, we propose Vary, an efficient and effective method to scale up the vision vocabulary of LVLMs. The procedures of Vary are naturally divided into two folds: the generation and integration of a new vision vocabulary. In the first phase, we devise a vocabulary network along with a tiny decoder-only transformer to produce the desired vocabulary via autoregression. In the next, we scale up the vanilla vision vocabulary by merging the new one with the original one (CLIP), enabling the LVLMs can quickly garner new features. Compared to the popular BLIP-2, MiniGPT4, and LLaVA, Vary can maintain its vanilla capabilities while enjoying more excellent fine-grained perception and understanding ability. Specifically, Vary is competent in new document parsing features (OCR or markdown conversion) while achieving 78.2% ANLS in DocVQA and 36.2% in MMVet. Our code will be publicly available on the homepage.

Pan Wang, Liangliang Ren, Shengkai Wu et al.

The point cloud based 3D single object tracking has drawn increasing attention. Although many breakthroughs have been achieved, we also reveal two severe issues. By extensive analysis, we find the prediction manner of current approaches is non-robust, i.e., exposing a misalignment gap between prediction score and actually localization accuracy. Another issue is the sparse point returns will damage the feature matching procedure of the SOT task. Based on these insights, we introduce two novel modules, i.e., Adaptive Refine Prediction (ARP) and Target Knowledge Transfer (TKT), to tackle them, respectively. To this end, we first design a strong pipeline to extract discriminative features and conduct the matching with the attention mechanism. Then, ARP module is proposed to tackle the misalignment issue by aggregating all predicted candidates with valuable clues. Finally, TKT module is designed to effectively overcome incomplete point cloud due to sparse and occlusion issues. We call our overall framework PCET. By conducting extensive experiments on the KITTI and Waymo Open Dataset, our model achieves state-of-the-art performance while maintaining a lower computational cost.

Lijun Gou, Jinrong Yang, Hangcheng Yu et al.

Unsupervised domain adaptation is critical in various computer vision tasks, such as object detection, instance segmentation, etc. They attempt to reduce domain bias-induced performance degradation while also promoting model application speed. Previous works in domain adaptation object detection attempt to align image-level and instance-level shifts to eventually minimize the domain discrepancy, but they may align single-class features to mixed-class features in image-level domain adaptation because each image in the object detection task may be more than one class and object. In order to achieve single-class with single-class alignment and mixed-class with mixed-class alignment, we treat the mixed-class of the feature as a new class and propose a mixed-classes $H-divergence$ for object detection to achieve homogenous feature alignment and reduce negative transfer. Then, a Semantic Consistency Feature Alignment Model (SCFAM) based on mixed-classes $H-divergence$ was also presented. To improve single-class and mixed-class semantic information and accomplish semantic separation, the SCFAM model proposes Semantic Prediction Models (SPM) and Semantic Bridging Components (SBC). And the weight of the pix domain discriminator loss is then changed based on the SPM result to reduce sample imbalance. Extensive unsupervised domain adaption experiments on widely used datasets illustrate our proposed approach's robust object detection in domain bias settings.

Jing Hao, Yuxuan Fan, Yanpeng Sun et al.

Recent advances in large vision-language models (LVLMs) have demonstrated strong performance on general-purpose medical tasks. However, their effectiveness in specialized domains such as dentistry remains underexplored. In particular, panoramic X-rays, a widely used imaging modality in oral radiology, pose interpretative challenges due to dense anatomical structures and subtle pathological cues, which are not captured by existing medical benchmarks or instruction datasets. To this end, we introduce MMOral, the first large-scale multimodal instruction dataset and benchmark tailored for panoramic X-ray interpretation. MMOral consists of 20,563 annotated images paired with 1.3 million instruction-following instances across diverse task types, including attribute extraction, report generation, visual question answering, and image-grounded dialogue. In addition, we present MMOral-Bench, a comprehensive evaluation suite covering five key diagnostic dimensions in dentistry. We evaluate 64 LVLMs on MMOral-Bench and find that even the best-performing model, i.e., GPT-4o, only achieves 41.45% accuracy, revealing significant limitations of current models in this domain. To promote the progress of this specific domain, we also propose OralGPT, which conducts supervised fine-tuning (SFT) upon Qwen2.5-VL-7B with our meticulously curated MMOral instruction dataset. Remarkably, a single epoch of SFT yields substantial performance enhancements for LVLMs, e.g., OralGPT demonstrates a 24.73% improvement. Both MMOral and OralGPT hold significant potential as a critical foundation for intelligent dentistry and enable more clinically impactful multimodal AI systems in the dental field. The dataset, model, benchmark, and evaluation suite are available at https://github.com/isbrycee/OralGPT.

Lijun Gou, Shengkai Wu, Jinrong Yang et al.

One major impediment in rapidly deploying object detection models for industrial applications is the lack of large annotated datasets. We currently have presented the Sacked Carton Dataset(SCD) that contains carton images from three scenarios, such as comprehensive pharmaceutical logistics company(CPLC), e-commerce logistics company(ECLC), fruit market(FM). However, due to domain shift, the model trained with one of the three scenarios in SCD has poor generalization ability when applied to the rest scenarios. To solve this problem, a novel image synthesis method is proposed to replace the foreground texture of the source datasets with the texture of the target datasets. Our method can keep the context relationship of foreground objects and backgrounds unchanged and greatly augment the target datasets. We firstly propose a surface segmentation algorithm to achieve texture decoupling of each instance. Secondly, a contour reconstruction algorithm is proposed to keep the occlusion and truncation relationship of the instance unchanged. Finally, the Gaussian fusion algorithm is used to replace the foreground texture from the source datasets with the texture from the target datasets. The novel image synthesis method can largely boost AP by at least 4.3%~6.5% on RetinaNet and 3.4%~6.8% on Faster R-CNN for the target domain. Code is available at https://github.com/hustgetlijun/RCAN.

Shengkai Wu, Jinrong Yang, Wenqiu Luo et al.

Imitation learning for robotic grasping is often plagued by the multimodal problem: when a scene contains multiple valid targets, demonstrations of grasping different objects create conflicting training signals. Standard imitation learning policies fail by averaging these distinct actions into a single, invalid action. In this paper, we introduce SAM2Grasp, a novel framework that resolves this issue by reformulating the task as a uni-modal, prompt-conditioned prediction problem. Our method leverages the frozen SAM2 model to use its powerful visual temporal tracking capability and introduces a lightweight, trainable action head that operates in parallel with its native segmentation head. This design allows for training only the small action head on pre-computed temporal-visual features from SAM2. During inference, an initial prompt, such as a bounding box provided by an upstream object detection model, designates the specific object to be grasped. This prompt conditions the action head to predict a unique, unambiguous grasp trajectory for that object alone. In all subsequent video frames, SAM2's built-in temporal tracking capability automatically maintains stable tracking of the selected object, enabling our model to continuously predict the grasp trajectory from the video stream without further external guidance. This temporal-prompted approach effectively eliminates ambiguity from the visuomotor policy. We demonstrate through extensive experiments that SAM2Grasp achieves state-of-the-art performance in cluttered, multi-object grasping tasks.

Jinrong Yang, Kexun Chen, Zhuoling Li et al.

Imitation learning (IL) with human demonstrations is a promising method for robotic manipulation tasks. While minimal demonstrations enable robotic action execution, achieving high success rates and generalization requires high cost, e.g., continuously adding data or incrementally conducting human-in-loop processes with complex hardware/software systems. In this paper, we rethink the state/action space of the data collection pipeline as well as the underlying factors responsible for the prediction of non-robust actions. To this end, we introduce a Hierarchical Data Collection Space (HD-Space) for robotic imitation learning, a simple data collection scheme, endowing the model to train with proactive and high-quality data. Specifically, We segment the fine manipulation task into multiple key atomic tasks from a high-level perspective and design atomic state/action spaces for human demonstrations, aiming to generate robust IL data. We conduct empirical evaluations across two simulated and five real-world long-horizon manipulation tasks and demonstrate that IL policy training with HD-Space-based data can achieve significantly enhanced policy performance. HD-Space allows the use of a small amount of demonstration data to train a more powerful policy, particularly for long-horizon manipulation tasks. We aim for HD-Space to offer insights into optimizing data quality and guiding data scaling. project page: https://hd-space-robotics.github.io.

Xu Luo, Longhui Wei, Liangjian Wen et al.

The category gap between training and evaluation has been characterised as one of the main obstacles to the success of Few-Shot Learning (FSL). In this paper, we for the first time empirically identify image background, common in realistic images, as a shortcut knowledge helpful for in-class classification but ungeneralizable beyond training categories in FSL. A novel framework, COSOC, is designed to tackle this problem by extracting foreground objects in images at both training and evaluation without any extra supervision. Extensive experiments carried on inductive FSL tasks demonstrate the effectiveness of our approaches.

Shengkai Wu, Jinrong Yang, Hangcheng Yu et al.

For most of the anchor-based detectors, Intersection over Union(IoU) is widely utilized to assign targets for the anchors during training. However, IoU pays insufficient attention to the closeness of the anchor's center to the truth box's center. This results in two problems: (1) only one anchor is assigned to most of the slender objects which leads to insufficient supervision information for the slender objects during training and the performance on the slender objects is hurt; (2) IoU can not accurately represent the alignment degree between the receptive field of the feature at the anchor's center and the object. Thus during training, some features whose receptive field aligns better with objects are missing while some features whose receptive field aligns worse with objects are adopted. This hurts the localization accuracy of models. To solve these problems, we firstly design Gaussian Guided IoU(GGIoU) which focuses more attention on the closeness of the anchor's center to the truth box's center. Then we propose GGIoU-balanced learning method including GGIoU-guided assignment strategy and GGIoU-balanced localization loss. The method can assign multiple anchors for each slender object and bias the training process to the features well-aligned with objects. Extensive experiments on the popular benchmarks such as PASCAL VOC and MS COCO demonstrate GGIoU-balanced learning can solve the above problems and substantially improve the performance of the object detection model, especially in the localization accuracy.

Jinrong Yang, Shengkai Wu, Lijun Gou et al.

Carton detection is an important technique in the automatic logistics system and can be applied to many applications such as the stacking and unstacking of cartons, the unloading of cartons in the containers. However, there is no public large-scale carton dataset for the research community to train and evaluate the carton detection models up to now, which hinders the development of carton detection. In this paper, we present a large-scale carton dataset named Stacked Carton Dataset(SCD) with the goal of advancing the state-of-the-art in carton detection. Images are collected from the internet and several warehourses, and objects are labeled using per-instance segmentation for precise localization. There are totally 250,000 instance masks from 16,136 images. In addition, we design a carton detector based on RetinaNet by embedding Offset Prediction between Classification and Localization module(OPCL) and Boundary Guided Supervision module(BGS). OPCL alleviates the imbalance problem between classification and localization quality which boosts AP by 3.1% - 4.7% on SCD while BGS guides the detector to pay more attention to boundary information of cartons and decouple repeated carton textures. To demonstrate the generalization of OPCL to other datasets, we conduct extensive experiments on MS COCO and PASCAL VOC. The improvement of AP on MS COCO and PASCAL VOC is 1.8% - 2.2% and 3.4% - 4.3% respectively.

Shengkai Wu, Jinrong Yang, Xinggang Wang et al.

Single-stage object detectors have been widely applied in computer vision applications due to their high efficiency. However, we find that the loss functions adopted by single-stage object detectors hurt the localization accuracy seriously. Firstly, the standard cross-entropy loss for classification is independent of the localization task and drives all the positive examples to learn as high classification scores as possible regardless of localization accuracy during training. As a result, there will be many detections that have high classification scores but low IoU or detections that have low classification scores but high IoU. Secondly, for the standard smooth L1 loss, the gradient is dominated by the outliers that have poor localization accuracy during training. The above two problems will decrease the localization accuracy of single-stage detectors. In this work, IoU-balanced loss functions that consist of IoU-balanced classification loss and IoU-balanced localization loss are proposed to solve the above problems. The IoU-balanced classification loss pays more attention to positive examples with high IoU and can enhance the correlation between classification and localization tasks. The IoU-balanced localization loss decreases the gradient of examples with low IoU and increases the gradient of examples with high IoU, which can improve the localization accuracy of models. Extensive experiments on challenging public datasets such as MS COCO, PASCAL VOC and Cityscapes demonstrate that both IoU-balanced losses can bring substantial improvement for the popular single-stage detectors, especially for the localization accuracy. On COCO test-dev, the proposed methods can substantially improve AP by $1.0\%\sim1.7\%$ and AP75 by $1.0\%\sim2.4\%$. On PASCAL VOC, it can also substantially improve AP by $1.3\%\sim1.5\%$ and AP80, AP90 by $1.6\%\sim3.9\%$.