Due to their ability of follow natural language instructions, vision-language-action (VLA) models are increasingly preva- lent in the embodied AI arena, following the widespread suc- cess of their precursors—LLMs and VLMs. In this paper, we discuss 10 principal milestones in the ongoing develop- ment of VLA models—multimodality, reasoning, data, eval- uation, cross-robkot action generalization, efficiency, whole- body coordination, safety, agents, and coordination with hu- mans. Furthermore, we discuss the emerging trends of us- ing spatial understanding, modeling world dynamics, post training, and data synthesis—all aiming to reach these mile- stones. Through these discussions, we hope to bring attention to the research avenues that may accelerate the development of VLA models into wider acceptability.

Reliable object handover between humans and robots represents a fundamental capability for collaborative robotic systems. However, the diversity of human hand poses and object properties often leads to unstable grasping and unsafe interactions, posing significant challenges for robust human-robot collaboration. To address these, we propose Bi-Handover, a novel paradigm that enables bidirectional, reliable object transfer by constructing stable, safe intermediate handover states. Our framework maps human hand postures to parallel gripper grasping configurations with equivalent grasping capabilities. The diverse grasp patterns employed by both the object giver and receiver generate multiple intermediate handover states that critically determine task success. We evaluate the robustness of these intermediate states using an integrated methodology that combines grasp stability prediction with safety quantification, ultimately selecting the optimal state to ensure reliable handover performance. Bi-Handover is the first approach to achieve reliable bidirectional handover of arbitrary objects, demonstrating substantial performance gains over existing baselines through extensive experimental validation.

Autonomous grasping in cluttered environments represents one of the most challenging problems in robotic manipulation. The presence of occlusions and densely packed objects significantly complicates perception and substantially increases the risk of collisions. To address these challenges, we present a novel multiscale perception-based framework for robust robotic grasping in dense, cluttered environments. Our methodology adopts a multiscale progressive perception architecture: First, a global context awareness module analyzes the distribution of viable grasping opportunities, assesses collision risks, and evaluates spatial optimization potential to systematically identify optimal grasping regions. Second, a regional collision prediction module provides intermediate-scale analysis, effectively reducing collision incidents through enhanced spatial awareness. Finally, a local grasp evaluation module refines grasp selection by optimizing stability metrics and predicting the probability of grasp success. Comprehensive experiments across simulated environments and real-world scenarios demonstrate that our approach achieves substantially superior performance compared to existing baselines, confirming its effectiveness and applicability.

In this paper, we tackle the open-set temporal action segmentation task, which aims to identify unknown frames while ensuring accurate segmentation of known actions in the temporal domain. Existing open-set methods struggle with identifying unknown frames due to their indistinguishability against ambiguous known frames during action transitions, resulting in significant performance degradation. To address this, we propose the action distribution flow, which models transitions between action sequences to capture the inherent feature discrepancies between unknown and known frames. Specifically, our method first models the distributions of known actions using the training data, and then interpolates these distributions along the optimal transport path for consecutive actions in the testing videos. By evaluating the likelihood of testing frames against the modeled action distribution flow, our approach effectively identifies unknown frames without requiring additional training or prior knowledge of the unknown data. Extensive experiments on open-set versions of the GTEA, 50Salads, and Breakfast datasets demonstrate the superiority of the proposed method across all evaluation metrics.

In this paper, we propose an efficient multi-level convolution architecture for 3D visual grounding. Conventional methods are difficult to meet the requirements of real-time inference due to the two-stage or point-based architecture. Inspired by the success of multi-level fully sparse convolutional architecture in 3D object detection, we aim to build a new 3D visual grounding framework following this technical route. However, as in 3D visual grounding task the 3D scene representation should be deeply interacted with text features, sparse convolution-based architecture is inefficient for this interaction due to the large amount of voxel features. To this end, we propose text-guided pruning (TGP) and completion-based addition (CBA) to deeply fuse 3D scene representation and text features in an efficient way by gradual region pruning and target completion. Specifically, TGP iteratively sparsifies the 3D scene representation and thus efficiently interacts the voxel features with text features by cross-attention. To mitigate the affect of pruning on delicate geometric information, CBA adaptively fixes the over-pruned region by voxel completion with negligible computational overhead. Compared with previous single-stage methods, our method achieves top inference speed and surpasses previous fastest method by 100% FPS. Our method also achieves state-of-the-art accuracy even compared with two-stage methods, with +1.13 lead of [email protected] on ScanRefer, and +2.6 and +3.2 leads on NR3D and SR3D respectively. The code is available at https://github.com/GWxuan/TSP3D.

While multimodal large language models demonstrate strong performance in complex reasoning tasks, they pose significant challenges related to model complexity during deployment, especially for resource-limited devices. In this paper, we propose an automatic pruning method for large vision-language models to enhance the efficiency of multimodal reasoning. Conventional methods rely on the training data of the original model to select the proper pruning ratio for different network components. However, these methods are impractical for large vision-language models due to the unaffordable search costs caused by web-scale training corpus. In contrast, our approach only leverages a small number of samples to search for the desired pruning policy by maximizing its generalization ability on unknown training data while maintaining the model accuracy, which enables the achievement of an optimal trade-off between accuracy and efficiency for large visual language models. Specifically, we formulate the generalization gap of the pruning strategy using the structural risk minimization principle. Based on both task performance and generalization capability, we iteratively search for the optimal pruning policy within a given search space and optimize the vision projector to evolve the search space with higher upper bound of performance. We conduct extensive experiments on the ScienceQA, Vizwiz, MM-vet, and LLaVA-Bench datasets for the task of visual question answering. Using only 64 samples for pruning policy search, EfficientLLaVA achieves an accuracy of 83.05% on ScienceQA, along with a ×1.8 speedup compared to the dense LLaVA-v1.5-7B model. The code is available at: https://github.com/liangynTHU/EfficientLLaVA

In this paper, we propose a general framework for universal zero-shot goal-oriented navigation. Existing zero-shot methods build inference framework upon large language models (LLM) for specific tasks, which differs a lot in overall pipeline and fails to generalize across different types of goal. Towards the aim of universal zero-shot navigation, we propose a uniform graph representation to unify different goals, including object category, instance image and text description. We also convert the observation of agent into an online maintained scene graph. With this consistent scene and goal representation, we preserve most structural information compared with pure text and are able to leverage LLM for explicit graph-based reasoning. Specifically, we conduct graph matching between the scene graph and goal graph at each time instant and propose different strategies to generate long-term goal of exploration according to different matching states. The agent first iteratively searches subgraph of goal when zero-matched. With partial matching, the agent then utilizes coordinate projection and anchor pair alignment to infer the goal location. Finally scene graph correction and goal verification are applied for perfect matching. We also present a blacklist mechanism to enable robust switch between stages. Extensive experiments on several benchmarks show that our UniGoal achieves state-of-the-art zero-shot performance on three studied navigation tasks with a single model, even outperforming task-specific zero-shot methods and supervised universal methods. Project Page.

Visual navigation with an image as goal is a fundamental and challenging problem. Conventional methods either rely on end-to-end RL learning or modular-based policy with topological graph or BEV map as memory, which cannot fully model the geometric relationship between the explored 3D environment and the goal image. In order to efficiently and accurately localize the goal image in 3D space, we build our navigation system upon the renderable 3D gaussian (3DGS) representation. However, due to the computational intensity of 3DGS optimization and the large search space of 6-DoF camera pose, directly leveraging 3DGS for image localization during agent exploration process is prohibitively inefficient. To this end, we propose IGL-Nav, an Incremental 3D Gaussian Localization framework for efficient and 3D-aware image-goal navigation. Specifically, we incrementally update the scene representation as new images arrive with feed-forward monocular prediction. Then we coarsely localize the goal by leveraging the geometric information for discrete space matching, which can be equivalent to efficient 3D convolution. When the agent is close to the goal, we finally solve the fine target pose with optimization via differentiable rendering. The proposed IGL-Nav outperforms existing state-of-the-art methods by a large margin across diverse experimental configurations. It can also handle the more challenging free-view image-goal setting and be deployed on real-world robotic platform using a cellphone to capture goal image at arbitrary pose. Project page: https://gwxuan.github.io/IGL-Nav/.

Embodied Instruction Following (EIF) requires agents to complete human instruction by interacting objects in complicated surrounding environments. Conventional methods directly consider the sparse human instruction to generate action plans for agents, which usually fail to achieve human goals because of the instruction incoherence in action descriptions. On the contrary, we propose ThinkBot that reasons the thought chain in human instruction to recover the missing action descriptions, so that the agent can successfully complete human goals by following the coherent instruction. Specifically, we first design an instruction completer based on large language models to recover the missing actions with interacted objects between consecutive human instruction, where the perceived surrounding environments and the completed sub-goals are considered for instruction completion. Based on the partially observed scene semantic maps, we present an object localizer to infer the position of interacted objects and the related Bayesian uncertainty for close-loop planning. Extensive experiments in the simulated environment show that our ThinkBot outperforms the state-of-the-art EIF methods by a sizable margin in both success rate and execution efficiency. Project page: https://guanxinglu.github.io/thinkbot/.

Embodied tasks require the agent to fully understand 3D scenes simultaneously with its exploration, so an online, real-time, fine-grained and highly-generalized 3D perception model is desperately needed. Since high-quality 3D data is limited, directly training such a model in 3D is infeasible. Meanwhile, vision foundation models (VFM) has revolutionized the field of 2D computer vision with superior performance, which makes the use of VFM to assist embodied 3D perception a promising direction. However, most existing VFM-assisted 3D perception methods are either offline or too slow that cannot be applied in practical embodied tasks. In this paper, we aim to leverage Segment Anything Model (SAM) for real-time 3D instance segmentation in an online setting. This is a challenging problem since future frames are not available in the input streaming RGB-D video, and an instance may be observed in several frames so efficient object matching between frames is required. To address these challenges, we first propose a geometric-aware query lifting module to represent the 2D masks generated by SAM by 3D-aware queries, which is then iteratively refined by a dual-level query decoder. In this way, the 2D masks are transferred to fine-grained shapes on 3D point clouds. Benefit from the query representation for 3D masks, we can compute the similarity matrix between the 3D masks from different views by efficient matrix operation, which enables real-time inference. Experiments on ScanNet, ScanNet200, SceneNN and 3RScan show our method achieves state-of-the-art performance among online 3D perception models, even outperforming offline VFM-assisted 3D instance segmentation methods by a large margin. Our method also demonstrates great generalization ability in several zero-shot dataset transferring experiments and show great potential in data-efficient setting.

In this paper, we propose a novel network framework for indoor 3D object detection to handle variable input frame numbers in practical scenarios. Existing methods only consider fixed frames of input data for a single detector, such as monocular RGB-D images or point clouds reconstructed from dense multi-view RGB-D images. While in practical application scenes such as robot navigation and manipulation, the raw input to the 3D detectors is the RGB-D images with variable frame numbers instead of the reconstructed scene point cloud. However, the previous approaches can only handle fixed frame input data and have poor performance with variable frame input. In order to facilitate 3D object detection methods suitable for practical tasks, we present a novel 3D detection framework named AnyView for our practical applications, which generalizes well across different numbers of input frames with a single model. To be specific, we propose a geometric learner to mine the local geometric features of each input RGB-D image frame and implement local-global feature interaction through a designed spatial mixture module. Meanwhile, we further utilize a dynamic token strategy to adaptively adjust the number of extracted features for each frame, which ensures consistent global feature density and further enhances the generalization after fusion. Extensive experiments on the ScanNet dataset show our method achieves both great generalizability and high detection accuracy with a simple and clean architecture containing a similar amount of parameters with the baselines.

Multi-task robotic bimanual manipulation is becoming increasingly popular as it enables sophisticated tasks that require diverse dual-arm collaboration patterns. Compared to unimanual manipulation, bimanual tasks pose challenges to understanding the multi-body spatiotemporal dynamics. An existing method ManiGaussian [30] pioneers encoding the spatiotemporal dynamics into the visual representation via Gaussian world model for single-arm settings, which ignores the interaction of multiple embodiments for dual-arm systems with significant performance drop. In this paper, we propose ManiGaussian++, an extension of ManiGaussian framework that improves multi-task bimanual manipulation by digesting multi-body scene dynamics through a hierarchical Gaussian world model. To be specific, we first generate task-oriented Gaussian Splatting from intermediate visual features, which aims to differentiate acting and stabilizing arms for multi-body spatiotemporal dynamics modeling. We then build a hierarchical Gaussian world model with the leader-follower architecture, where the multi-body spatiotemporal dynamics is mined for intermediate visual representation via future scene prediction. The leader predicts Gaussian Splatting deformation caused by motions of the stabilizing arm, through which the follower generates the physical consequences resulted from the movement of the acting arm. As a result, our method significantly outperforms the current state-of-the-art bimanual manipulation techniques by an improvement of 20.2% in 10 simulated tasks, and achieves 60% success rate on average in 9 challenging real-world tasks. Our code is available at https://github.com/April-Yz/ManiGaussian_Bimanual.

In this paper, we propose an autonomous robot packing system named RoboPacker designed to tightly store cluttered general objects into shipping boxes with high space utilization, which is a fundamental process in numerous industrial applications. However, achieving tight packaging for general objects often demands significant labor from human packers, particularly in high-throughput scenes. Compared to existing robot packing approaches, RoboPacker effectively overcomes challenges such as diverse object appearances, severe occlusion, and crowded packing spaces. Specifically, we propose an open-vocabulary shape estimation method to reconstruct complete point clouds for cluttered objects. We also design effective interactions with object clutter to gather informative visual clues for shape estimation under high uncertainty. Additionally, we introduce a hierarchical reinforcement learning framework to optimize packing order, location, and orientation for maximum space utilization. The robotic packing system integrates these techniques with feasible manipulation methods for real-world implementation. In this way, RoboPacker achieves efficient packing of novel and irregular objects, which is more suitable for real deployment environments. The Real-world experiments demonstrate RoboPacker can tightly pack 20 densely cluttered everyday objects from 8 seen and 4 novel classes into the 40×40×20 cm shipping box with a 73.3% success rate. The demonstration video can be found at https://gary3410.github.io/RoboPacker/.

This work addresses the certification of the local robustness of vision-based two-stage 6D object pose estimation. The two-stage method for object pose estimation achieves superior accuracy over the single-stage approach by first employing deep neural network-driven keypoint regression and then applying a Perspective-n-Point (PnP) technique. Despite advancements, the certification of these methods’ robustness, especially in safety-critical scenarios, remains scarce. This research aims to fill this gap with a focus on their local robustness on the system level—the capacity to maintain robust estimations amidst semantic input perturbations. The core idea is to transform the certification of local robustness into a process of neural network verification for classification tasks. The challenge is to develop model, input, and output specifications that align with off-the-shelf verification tools. To facilitate verification, we modify the keypoint detection model by substituting non-linear operations with those more amenable to the verification processes. Instead of merely injecting random noise into images, as is common, we employ a convex hull representation of images as input specifications to more accurately depict semantic perturbations. Furthermore, by conducting a sensitivity analysis, we propagate the robustness criteria from pose estimation to keypoint accuracy, and then formulating an optimal error threshold allocation problem that allows for the setting of a maximally permissible keypoint deviation thresholds. Viewing each pixel as an individual class, these thresholds result in linear, classification-akin output specifications. Under certain conditions, we demonstrate that the main components of our certification framework are both sound and complete, and validate its effects through extensive evaluations on realistic perturbations. To our knowledge, this is the first study to certify the robustness of large-scale, keypoint-based pose estimation given images in real-world scenarios.

In this paper, we propose an accurate post-training quantization framework of diffusion models (APQ-DM) for efficient image generation. Conventional quantization frameworks learn shared quantization functions for tensor discretization regardless of the generation timesteps in diffusion models, while the activation distribution differs significantly across various timesteps. Meanwhile, the calibration images are acquired in random timesteps which fail to provide sufficient information for generalizable quantization function learning. Both issues cause sizable quantization errors with obvious image generation performance degradation. On the contrary, we design distribution-aware quantization functions for activation discretization in different timesteps and search the optimal timesteps for informative calibration image generation, so that our quantized diffusion model can reduce the discretization errors with negligible computational overhead. Specifically, we partition various timestep quantization functions into different groups according to the importance weights, which are optimized by differentiable search algorithms. We also extend structural risk minimization principle for informative calibration image generation to enhance the generalization ability in the deployment of quantized diffusion model. Extensive experimental results show that our method outperforms the state-of-the-art post-training quantization of diffusion model by a sizable margin with similar computational cost11Code is available at https://github.com/ChangyuanWang17/APQ-DM.

In this paper, we propose a new framework for online 3D scene perception. Conventional 3D scene perception methods are offline, i.e., take an already reconstructed 3D scene geometry as input, which is not applicable in robotic applications where the input data is streaming RGB-D videos rather than a complete 3D scene reconstructed from precollected RGB-D videos. To deal with online 3D scene perception tasks where data collection and perception should be performed simultaneously, the model should be able to process 3D scenes frame by frame and make use of the temporal information. To this end, we propose an adapter-based plug-and-play module for the backbone of 3D scene perception model, which constructs memory to cache and aggregate the extracted RGB-D features to empower offline models with temporal learning ability. Specifically, we propose a queued memory mechanism to cache the supporting point cloud and image features. Then we devise aggregation modules which directly perform on the memory and pass temporal information to current frame. We further propose 3D-to-2D adapter to enhance image features with strong global context. Our adapters can be easily inserted into mainstream offline architectures of different tasks and significantly boost their performance on online tasks. Extensive experiments on ScanNet and SceneNN datasets demonstrate our approach achieves leading performance on three 3D scene perception tasks compared with state-of-the-art online methods by simply finetuning existing offline models, without any model and task-specific designs. Project page.

In this paper, we present a tensor decomposition and low-rank recovery approach (LowRankOcc) for vision-based 3D semantic occupancy prediction. Conventional methods model outdoor scenes with fine-grained 3D grids, but the sparsity of non-empty voxels introduces consider-able spatial redundancy, leading to potential overfitting risks. In contrast, our approach leverages the intrinsic low-rank property of 3D occupancy data, factorizing voxel representations into low-rank components to efficiently mitigate spatial redundancy without sacrificing performance. Specifically, we present the Vertical-Horizontal (VH) de-composition block factorizes 3D tensors into vertical vectors and horizontal matrices. With our “decomposition-encoding-recovery” framework, we encode 3D contexts with only 1/2D convolutions and poolings, and subsequently recover the encoded compact yet informative context features back to voxel representations. Experimental results demonstrate that LowRankOcc achieves state-of-the-art performances in semantic scene completion on the Se-manticKITTI dataset and 3D occupancy prediction on the nuScenes dataset.

In this paper, we propose a post-training quantization framework of large vision-language models (LVLMs) for efficient multi-modal inference. Conventional quantization methods sequentially search the layer-wise rounding functions by minimizing activation discretization errors, which fails to acquire optimal quantization strategy without considering cross-layer dependency. On the contrary, we mine the cross-layer dependency that significantly influences discretization errors of the entire vision-language model, and embed this dependency into optimal quantization strategy searching with low search cost. Specifically, we observe the strong correlation between the activation entropy and the cross-layer dependency concerning output discretization errors. Therefore, we employ the entropy as the proxy to partition blocks optimally, which aims to achieve satisfying trade-offs between discretization errors and the search cost. Moreover, we optimize the visual encoder to disentangle the cross-layer dependency for fine-grained decomposition of search space, so that the search cost is further reduced without harming the quantization accuracy. Experimental results demonstrate that our method compresses the memory by 2.78x and increase generate speed by 1.44x about 13B LLaVA model without performance degradation on diverse multi-modal reasoning tasks.

In this paper, we propose a weakly-supervised approach for 3D object detection, which makes it possible to train a strong 3D detector with position-level annotations (i.e. annotations of object centers and categories). In order to remedy the information loss from box annotations to centers, our method makes use of synthetic 3D shapes to convert the position-level annotations into virtual scenes with box-level annotations, and in turn utilizes the fully-annotated virtual scenes to complement the real labels. Specifically, we first present a shape-guided label-enhancement method, which assembles 3D shapes into physically reasonable virtual scenes according to the coarse scene layout extracted from position-level annotations. Then we transfer the information contained in the virtual scenes back to real ones by applying a virtual-to-real domain adaptation method, which refines the annotated object centers and additionally supervises the training of detector with the virtual scenes. Since the shape-guided label enhancement method generates virtual scenes by human-heuristic physical constraints, the layout of the fixed virtual scenes may be unreasonable with varied object combinations. To address this, we further present differentiable label enhancement to optimize the virtual scenes including object scales, orientations and locations in a data-driven manner. Moreover, we further propose a label-assisted self-training strategy to fully exploit the capability of detector. By reusing the position-level annotations and virtual scenes, we fuse the information from both domains and generate box-level pseudo labels on the real scenes, which enables us to directly train a detector in fully-supervised manner. Extensive experiments on the widely used ScanNet and Matterport3D datasets show that our approach surpasses current weakly-supervised and semi-supervised methods by a large margin, and achieves comparable detection performance with some popular fully-supervised methods with less than 5% of the labeling labor.

In this paper, we propose an anycost network quantization method for efficient image super-resolution with variable resource budgets. Conventional quantization approaches acquire discrete network parameters for deployment with fixed complexity constraints, while image super-resolution networks are usually applied on mobile devices with frequently modified resource budgets due to the change of battery levels or computing chips. Hence, exhaustively optimizing quantized networks with each complexity constraint results in unacceptable training costs. On the contrary, we construct a hyper-network whose parameters can efficiently adapt to different resource budgets with negligible finetuning cost, so that the image super-resolution networks can be feasibly deployed in diversified devices with variable resource budgets. Specifically, we dynamically search the optimal bitwidth for each patch in convolution according to feature maps and complexity constraints, which aims to achieve the best efficiency-accuracy trade-off in image super-resolution given the resource budget. To acquire the hyper-network that can be efficiently adapted to different bitwidth settings, we actively sample the patch-wise bitwidth during training and adaptively ensemble gradients from hyper-network in different precision for faster convergence and higher generalization ability. Compared with existing quantization methods, experimental results demonstrate that our method significantly reduces the cost of adapting models in new resource budgets with comparable efficiency-accuracy trade-offs.

In this paper, we propose binary sparse convolutional networks called BSC-Net for efficient point cloud analysis. We empirically observe that sparse convolution operation causes larger quantization errors than standard convolution. However, conventional network quantization methods directly binarize the weights and activations in sparse convolution, resulting in performance drop due to the significant quantization loss. On the contrary, we search the optimal subset of convolution operation that activates the sparse convolution at various locations for quantization error alleviation, and the performance gap between real-valued and binary sparse convolutional networks is closed without complexity overhead. Specifically, we first present the shifted sparse convolution that fuses the information in the receptive field for the active sites that match the pre-defined positions. Then we employ the differentiable search strategies to discover the optimal opsitions for active site matching in the shifted sparse convolution, and the quantization errors are significantly alleviated for efficient point cloud analysis. For fair evaluation of the proposed method, we empirically select the recently advances that are beneficial for sparse convolution network binarization to construct a strong baseline. The experimental results on ScanNet and NYU Depth v2 show that our BSC-Net achieves significant improvement upon our srtong baseline and outperforms the state-of-the-art network binarization methods by a remarkable margin without additional computation overhead for binarizing sparse convolutional networks.

Due to the high price and heavy energy consumption of GPUs, deploying deep models on IoT devices such as microcontrollers makes significant contributions for ecological AI. Conventional methods successfully enable convolutional neural network inference of high resolution images on microcontrollers, while the framework for vision transformers that achieve the state-of-the-art performance in many vision applications still remains unexplored. In this paper, we propose a hardware-algorithm co-optimizations method called MCUFormer to deploy vision transformers on microcontrollers with extremely limited memory, where we jointly design transformer architecture and construct the inference operator library to fit the memory resource constraint. More specifically, we generalize the one-shot network architecture search (NAS) to discover the optimal architecture with highest task performance given the memory budget from the microcontrollers, where we enlarge the existing search space of vision transformers by considering the low-rank decomposition dimensions and patch resolution for memory reduction. For the construction of the inference operator library of vision transformers, we schedule the memory buffer during inference through operator integration, patch embedding decomposition, and token overwriting, allowing the memory buffer to be fully utilized to adapt to the forward pass of the vision transformer. Experimental results demonstrate that our MCUFormer achieves 73.62\% top-1 accuracy on ImageNet for image classification with 320KB memory on STM32F746 microcontroller. Code is available at https://github.com/liangyn22/MCUFormer.

In this article, we propose extremely low-precision vision transformers called Quantformer for efficient inference. Conventional network quantization methods directly quantize weights and activations of fully-connected layers without considering properties of transformer architectures. Quantization sizably deviates the self-attention compared with full-precision counterparts, and the shared quantization strategy for diversely distributed patch features causes severe quantization errors. To address these issues, we enforce the self-attention rank in quantized transformers to mimic that in full-precision counterparts with capacity-aware distribution for information retention, and quantize patch features with group-wise discretization strategy for quantization error minimization. Specifically, we efficiently preserve the self-attention rank consistency by minimizing the distance between the self-attention in quantized and real-valued transformers with adaptive concentration degree, where the optimal concentration degree is selected according to the self-attention entropy for model capacity adaptation. Moreover, we partition patch features in different dimensions with differentiable group assignment, so that features in different groups leverage various discretization strategies with minimal rounding and clipping errors. Experimental results show that our Quantformer outperforms the state-of-the-art network quantization methods by a sizable margin across various vision transformer architectures in image classification and object detection. We also integrate our Quantformer with mixed-precision quantization to further enhance the performance of the vanilla models.

In this paper, we propose a GraphBit method to learn unsupervised deep binary descriptors for efficient image representation. Conventional binary representation learning methods directly quantize each element according to the threshold without considering the quantization ambiguousness. The elements near the boundary dubbed as "ambiguous bits" fail to collect effective information for reliable binarization and are sensitive to noise that causes reversed bits. We argue that there are implicit inner relationships among bits in binary descriptors called bitwise interaction, where the related bits can provide extra instruction as prior knowledge for ambiguousness reduction. Specifically, we design a deep reinforcement learning model to learn the structure of the graph for bitwise interaction mining, and the uncertainty of binary codes is reduced by maximizing the mutual information with input and related bits. Consequently, the ambiguous bits receive additional instruction from the graph for reliable binarization. Moreover, we further present a differentiable search method (GraphBit+) that mines the bitwise interaction in continuous space, so that the heavy search cost caused by the training difficulties in reinforcement learning is significantly reduced. Since the GraphBit and GraphBit+ methods learn fixed bitwise interaction which is suboptimal for various input, the inaccurate instruction from the fixed bitwise interaction cannot effectively decrease the ambiguousness of binary descriptors. To address this, we further propose the unsupervised binary descriptor learning method via dynamic bitwise interaction mining (D-GraphBit), where a graph convolutional network called GraphMiner reasons the optimal bitwise interaction for each input sample. Extensive experimental results on the CIFAR-10, NUS-WIDE, ImageNet-100, Brown and HPatches datasets demonstrate the efficiency and effectiveness of the proposed GraphBit, GraphBit+ and D-GraphBit.

In this paper, we propose a Shapley value based method to evaluate operation contribution (Shapley-NAS) for neural architecture search. Differentiable architecture search (DARTS) acquires the optimal architectures by optimizing the architecture parameters with gradient descent, which significantly reduces the search cost. However, the magnitude of architecture parameters updated by gradient descent fails to reveal the actual operation importance to the task performance and therefore harms the effectiveness of obtained architectures. By contrast, we propose to evaluate the direct influence of operations on validation accuracy. To deal with the complex relationships between supernet components, we leverage Shapley value to quantify their marginal contributions by considering all possible combinations. Specifically, we iteratively optimize the supernet weights and update the architecture parameters by evaluating operation contributions via Shapley value, so that the optimal architectures are derived by selecting the operations that contribute significantly to the tasks. Since the exact computation of Shapley value is NP-hard, the Monte-Carlo sampling based algorithm with early truncation is employed for efficient approximation, and the momentum update mechanism is adopted to alleviate fluctuation of the sampling process. Extensive experiments on various datasets and various search spaces show that our Shapley-NAS outperforms the state-of-the-art methods by a considerable margin with light search cost. The code is available at https://github.com/Euphoria16/Shapley-NAS.git.

Grasping in dense clutter is a fundamental skill for autonomous robots. However, the crowdedness and oc-clusions in the cluttered scenario cause significant difficul-ties to generate valid grasp poses without collisions, which results in low efficiency and high failure rates. To address these, we present a generic framework called GE-Grasp for robotic motion planning in dense clutter, where we leverage diverse action primitives for occluded object removal and present the generator-evaluator architecture to avoid spatial collisions. Therefore, our GE-Grasp is capable of grasping objects in dense clutter efficiently with promising success rates. Specifically, we define three action primitives: target-oriented grasping for target capturing, pushing, and nontarget-oriented grasping to reduce the crowdedness and occlusions. The gen-erators effectively provide various action candidates referring to the spatial information. Meanwhile, the evaluators assess the selected action primitive candidates, where the optimal action is implemented by the robot. Extensive experiments in simulated and real-world environments show that our approach outperforms the state-of-the-art methods of grasping in clutter with respect to motion efficiency and success rates. Moreover, we achieve comparable performance in the real world as that in the simulation environment, which indicates the strong gen-eralization ability of our GE-Grasp. Supplementary material is available at: https://github.com/CaptainWuDaoKou/GE-Grasp.

Recognizing objects in dense clutter accurately plays an important role to a wide variety of robotic manipulation tasks including grasping, packing, rearranging and many others. However, conventional visual recognition models usually miss objects because of the significant occlusion among instances and causes incorrect prediction due to the visual ambiguity with the high object crowdedness. In this paper, we propose an interactive exploration framework called Smart Explorer for recognizing all objects in dense clutters. Our Smart Explorer physically interacts with the clutter to maximize the recognition performance while minimize the number of motions, where the false positives and negatives can be alleviated effectively with the optimal accuracy-efficiency trade-offs. Specifically, we first collect the multi-view RGB-D images of the clutter and reconstruct the corresponding point cloud. By aggregating the instance segmentation of RGB images across views, we acquire the instance-wise point cloud partition of the clutter through which the existed classes and the number of objects for each class are predicted. The pushing actions for effective physical interaction are generated to sizably reduce the recognition uncertainty that consists of the instance segmentation entropy and multi-view object disagreement. Therefore, the optimal accuracy-efficiency trade-off of object recognition in dense clutter is achieved via iterative instance prediction and physical interaction. Extensive experiments demonstrate that our Smart Explorer acquires promising recognition accuracy with only a few actions, which also outperforms the random pushing by a large margin.

In this paper, we propose a binarized neural network learning method (BiDet) for efficient object detection. Conventional network binarization methods directly quantize the weights and activations in one-stage or two-stage detectors with constrained representational capacity, so that the information redundancy in the networks causes numerous false positives and degrades the performance significantly. On the contrary, our BiDet fully utilizes the representational capacity of the binary neural networks by redundancy removal, through which the detection precision is enhanced with alleviated false positives. Specifically, we generalize the information bottleneck (IB) principle to object detection, where the amount of information in the high-level feature maps is constrained and the mutual information between the feature maps and object detection is maximized. Meanwhile, we learn sparse object priors so that the posteriors are concentrated on informative detection prediction with false positive elimination. Since BiDet employs a fixed IB trade-off to balance the total and relative information contained in the high-level feature maps, the information compression leads to ineffective utilization of the network capacity or insufficient redundancy removal for input in different complexity. To address this, we further present binary neural networks with automatic information compression (AutoBiDet) to automatically adjust the IB trade-off for each input according to the complexity. Moreover, we further propose the class-aware sparse object priors by assigning different sparsity to objects in various classes, so that the false positives are alleviated more effectively without recall decrease. Extensive experiments on the PASCAL VOC and COCO datasets show that our BiDet and AutoBiDet outperform the state-of-the-art binarized object detectors by a sizable margin.

In this paper, we propose an instance similarity learning (ISL) method for unsupervised feature representation. Conventional methods assign close instance pairs in the feature space with high similarity, which usually leads to wrong pairwise relationship for large neighborhoods because the Euclidean distance fails to depict the true semantic similarity on the feature manifold. On the contrary, our method mines the feature manifold in an unsupervised manner, through which the semantic similarity among instances is learned in order to obtain discriminative representations. Specifically, we employ the Generative Adversarial Networks (GAN) to mine the underlying feature manifold, where the generated features are applied as the proxies to progressively explore the feature manifold so that the semantic similarity among instances is acquired as reliable pseudo supervision. Extensive experiments on image classification demonstrate the superiority of our method compared with the state-of-the-art methods. The code is available at https://github.com/ZiweiWangTHU/ISL.git.

In this paper, we propose a generalizable mixed-precision quantization (GMPQ) method for efficient inference. Conventional methods require the consistency of datasets for bitwidth search and model deployment to guarantee the policy optimality, leading to heavy search cost on challenging largescale datasets in realistic applications. On the contrary, our GMPQ searches the mixed-quantization policy that can be generalized to largescale datasets with only a small amount of data, so that the search cost is significantly reduced without performance degradation. Specifically, we observe that locating network attribution correctly is general ability for accurate visual analysis across different data distribution. Therefore, despite of pursuing higher model accuracy and complexity, we preserve attribution rank consistency between the quantized models and their full-precision counterparts via efficient capacity-aware attribution imitation for generalizable mixed-precision quantization strategy search. Extensive experiments show that our method obtains competitive accuracy-complexity trade-off compared with the state-of-the-art mixed-precision networks in significantly reduced search cost. The code is available at https://github.com/ZiweiWangTHU/GMPQ.git.

In this paper, we propose a channel-wise interaction based binary convolutional neural networks (CI-BCNN) approach for efficient inference. Conventional binary convolutional neural networks usually apply the xnor and bitcount operations in the binary convolution with notable quantization errors, which obtain opposite signs of pixels in binary feature maps compared to their full-precision counterparts and lead to significant information loss. In our proposed CI-BCNN method, we exploit the channel-wise interactions with the prior knowledge which aims to alleviate inconsistency of signs in binary feature maps and preserves the information of input samples during inference. Specifically, we mine the channel-wise interactions by using a reinforcement learning model, and impose channel-wise priors on the intermediate feature maps to correct inconsistent signs through the interacted bitcount. Since CI-BCNN mines the channel-wise interactions in a large search space where each channel may correlate with others, the search deficiency caused by sparse interactions obstacles the agent to obtain the optimal policy. To address this, we further present a hierarchical channel-wise interaction based binary convolutional neural networks (HCI-BCNN) method to shrink the search space via hierarchical reinforcement learning. Moreover, we propose a denoising interacted bitcount operation in binary convolution by smoothing the channel-wise interactions, so that noise in channel-wise priors can be alleviated. Extensive experimental results on the CIFAR-10 and ImageNet datasets demonstrate the effectiveness of the proposed CI-BCNN and HCI-BCNN.

In this paper, we propose a binarized neural network learning method called BiDet for efficient object detection. Conventional network binarization methods directly quantize the weights and activations in one-stage or two-stage detectors with constrained representational capacity, so that the information redundancy in the networks causes numerous false positives and degrades the performance significantly. On the contrary, our BiDet fully utilizes the representational capacity of the binary neural networks for object detection by redundancy removal, through which the detection precision is enhanced with alleviated false positives. Specifically, we generalize the information bottleneck (IB) principle to object detection, where the amount of information in the high-level feature maps is constrained and the mutual information between the feature maps and object detection is maximized. Meanwhile, we learn sparse object priors so that the posteriors are concentrated on informative detection prediction with false positive elimination. Extensive experiments on the PASCAL VOC and COCO datasets show that our method outperforms the state-of-the-art binary neural networks by a sizable margin.

In this paper, we propose a channel-wise interaction based binary convolutional neural network learning method (CI-BCNN) for efficient inference. Conventional methods apply xnor and bitcount operations in binary convolution with notable quantization error, which usually obtains inconsistent signs in binary feature maps compared with their full-precision counterpart and leads to significant information loss. In contrast, our CI-BCNN mines the channel-wise interactions, through which prior knowledge is provided to alleviate inconsistency of signs in binary feature maps and preserves the information of input samples during inference. Specifically, we mine the channel-wise interactions by a reinforcement learning model, and impose channel-wise priors on the intermediate feature maps through the interacted bitcount function. Extensive experiments on the CIFAR-10 and ImageNet datasets show that our method outperforms the state-of-the-art binary convolutional neural networks with less computational and storage cost.

In this paper, we propose an unsupervised feature learning method called deep binary descriptor with multi-quantization (DBD-MQ) for visual analysis. Existing learning-based binary descriptors such as compact binary face descriptor (CBFD) and DeepBit utilize the rigid sign function for binarization despite of data distributions, which usually suffer from severe quantization loss. In order to address the limitation, we propose a deep multi-quantization network to learn a data-dependent binarization in an unsupervised manner. More specifically, we design a K-Autoencoders (KAEs) network to jointly learn the parameters of feature extractor and the binarization functions under a deep learning framework, so that discriminative binary descriptors can be obtained with a fine-grained multi-quantization. As DBD-MQ simply allocates the same number of quantizers to each real-valued feature dimension ignoring the elementwise diversity of informativeness, we further propose a deep competitive binary descriptor with multi-quantization (DCBD-MQ) method to learn optimal allocation of bits with the fixed binary length in a competitive manner, where informative dimensions gain more bits for complete representation. Moreover, we present a similarity-aware binary encoding strategy based on the earth mover's distance of Autoencoders, so that elements that are quantized into similar Autoencoders will have smaller Hamming distances. Extensive experimental results on six widely-used datasets show that our DBD-MQ and DCBD-MQ outperform most state-of-the-art unsupervised binary descriptors.

In this paper, we propose a GraphBit method to learn deep binary descriptors in a directed acyclic graph unsupervisedly, representing bitwise interactions as edges between the nodes of bits. Conventional binary representation learning methods enforce each element to be binarized into zero or one. However, there are elements lying in the boundary which suffer from doubtful binarization as "ambiguous bits". Ambiguous bits fail to collect effective information for confident binarization, which are unreliable and sensitive to noise. We argue that there are implicit inner relationships between bits in binary descriptors, where the related bits can provide extra instruction as prior knowledge for ambiguity elimination. Specifically, we design a deep reinforcement learning model to learn the structure of the graph for bitwise interaction mining, reducing the uncertainty of binary codes by maximizing the mutual information with inputs and related bits, so that the ambiguous bits receive additional instruction from the graph for confident binarization. Due to the reliability of the proposed binary codes with bitwise interaction, we obtain an average improvement of 9.64%, 8.84% and 3.22% on the CIFAR-10, Brown and HPatches datasets respectively compared with the state-of-the-art unsupervised binary descriptors.

In this paper, we propose an unsupervised feature learning method called deep binary descriptor with multi-quantization (DBD-MQ) for visual matching. Existing learning-based binary descriptors such as compact binary face descriptor (CBFD) and DeepBit utilize the rigid sign function for binarization despite of data distributions, thereby suffering from severe quantization loss. In order to address the limitation, our DBD-MQ considers the binarization as a multi-quantization task. Specifically, we apply a K-AutoEncoders (KAEs) network to jointly learn the parameters and the binarization functions under a deep learning framework, so that discriminative binary descriptors can be obtained with a fine-grained multi-quantization. Extensive experimental results on different visual analysis including patch retrieval, image matching and image retrieval show that our DBD-MQ outperforms most existing binary feature descriptors.