In this article, we consider a special class of liveness properties for systems consisting of concurrent objects. These properties ensure the termination of methods calls under certain fairness assumptions and thus the progress of the execution. Liveness properties are defined for concurrent objects and they typically include lock-freedom , wait-freedom , deadlock-freedom , starvation-freedom, and obstruction-freedom . It is known that these five liveness properties are decidable for sequential consistency (SC) memory model of finite-state programs with a bounded number of processes. However, the problem of decidability of liveness for finite state concurrent programs running on relaxed memory models remains open. In this article, we address the decidability problem of liveness properties of concurrent objects for the total store order (TSO) memory model which is used in the x86 architecture. In particular, we prove that for a bounded number of processes, lock-freedom, wait-freedom, deadlock-freedom and starvation-freedom are undecidable, and that obstruction-freedom is decidable on TSO for a bounded number of processes. Further on, we investigate the verification problem of k -bounded wait-freedom , a bounded version of wait-freedom, and show that for each bound k , the problem of checking k -bounded wait-freedom is decidable on TSO for a bounded number of processes. We show that the complexity for checking obstruction-freedom and checking k -bounded wait-freedom are both non-primitive recursive. We also discover an interesting difference between liveness on TSO and that on SC. Our finding is that wait-freedom implies k -bounded wait-freedom for some k on SC memory model, but this implication does not hold on the TSO model. We prove this by generating a concrete object on TSO that is wait-free but not k -bounded wait-free for any k .

Multi-object navigation (MON) tasks involve sequentially locating multiple targets in an unknown environment, requiring global long-term planning under incomplete information. This necessitates that the agent dynamically balance immediate actions and long-term rewards while considering both local adaptability and global foresight. However, current methods overly focus on local path optimization, which leads to slower convergence in sparse reward settings and increases the risk of deadlocks or trap states. The core challenge of MON lies in the deformation of the shared decision space, where independent optimization leads to redundant and overlapping paths. Thus, path planning requires dynamic, cross-task optimization rather than simple subtask aggregation. To minimize overall effort, the optimization process should adaptively balance task contributions through weight adjustment. Thus, we propose the Goal-oriented Dynamic Weight Optimization (GDWO) algorithm. GDWO integrates target-specific value loss functions into a unified optimization framework and dynamically adjusts weights through gradient-based updates. To prevent over-optimization, weights are normalized during training according to navigation success rates, prioritizing more challenging targets. This adaptive mechanism effectively addresses the challenge of sparse rewards and improves convergence efficiency. By leveraging this mechanism, GDWO unifies multiple objectives within a unified decision space, achieving efficient optimization and balancing short-term gains with long-term goals. Additionally, we introduce two auxiliary modules: prior knowledge-based navigation and frontier-aware exploration to further enhance GDWO's performance. Experimental results on the Gibson and Matterport3D datasets demonstrate that GDWO achieves improvements in key metrics for MON tasks. It optimizes path planning, reduces exploration costs, and enhances navigation efficiency, enabling the agent to perform tasks more effectively in complex environments.

The goal of object navigation task is to reach the expected objects using visual information in unseen environments. Previous works typically implement deep models as agents that are trained to predict actions based on visual observations. Despite extensive training, agents often fail to make wise decisions when navigating in unseen environments toward invisible targets. In contrast, humans demonstrate a remarkable talent to navigate toward targets even in unseen environments. This superior capability is attributed to the cognitive map in the hippocampus, which enables humans to recall past experiences in similar situations and anticipate future occurrences during navigation. It is also dynamically updated with new observations from unseen environments. The cognitive map equips humans with a wealth of prior knowledge, significantly enhancing their navigation capabilities. Inspired by human navigation mechanisms, we propose the Hierarchical Object-to-Zone (HOZ++) graph, which encapsulates the regularities among objects, zones, and scenes. The HOZ++ graph helps the agent to identify the current zone and the target zone, and computes an optimal path between them, then selects the next zone along the path as the guidance for the agent. Moreover, the HOZ++ graph continuously updates based on real-time observations in new environments, thereby enhancing its adaptability to new environments. Our HOZ++ graph is versatile and can be integrated into existing methods, including end-to-end RL and modular methods. Our method is evaluated across four simulators, including AI2-THOR, RoboTHOR, Gibson, and Matterport 3D. Additionally, we build a realistic environment to evaluate our method in the real world. Experimental results demonstrate the effectiveness and efficiency of our proposed method.

This paper presents a novel category agnostic model for visual rearrangement task, which can help an embodied agent to physically recover the shuffled scene configuration without any category concepts to the goal configuration. Previous methods usually follow a similar architecture, completing the rearrangement task by aligning the scene changes of the goal and shuffled configuration, according to the semantic scene graphs. However, constructing scene graphs requires the inference of category labels, which not only causes the accuracy drop of the entire task but also limits the application in real world scenario. In this paper, we delve deep into the essence of visual re-arrangement task and focus on the two most essential issues, scene change detection and scene change matching. We utilize the movement and the protrusion of point cloud to accurately identify the scene changes and match these changes depending on the similarity of category agnostic appearance feature. Moreover, to assist the agent to explore the environment more efficiently and comprehensively, we propose a closer-aligned-retrace exploration policy, aiming to observe more details of the scene at a closer distance. We conduct extensive experiments onAI2THOR Rearrangement Challenge based on RoomR dataset and a new multi-room multi-instance dataset MrMiR collected by us. The experimental results demonstrate the effectiveness of our proposed method.

Visual navigation is to let the agent reach the target according to the continuous visual input. In most previous works, visual navigation is usually assumed to be done in a static and ideal environment: the target is always reachable with no need to alter the environment. However, the “messy” environments are more general and practical in our daily lives, where the agent may get blocked by obstacles. Thus Interactive Navigation (InterNav) is introduced to navigate to the objects in more realistic “messy” environments according to the object interaction. Prior work on InterNav learns shortterm interaction through extensive trials with reinforcement learning. However, interaction does not guarantee efficient navigation, that is, plan-ning obstacle interactions that make shorter paths and con-sume less effort is also crucial. In this paper, we introduce an effect-oriented affordance map to enable longterm interactive navigation, extending the existing map-based nav-igation framework to the domain of dynamic environment. We train a set of affordance functions predicting available interactions and the time cost of removing obstacles, which informatively support an interactive modular system to ad-dress interaction and longterm planning. Experiments on the ProcTHOR simulator demonstrate the capability of our affordance-driven system in longterm navigation in complex dynamic environments.

The Object Goal navigation (ObjectNav) task requires the agent to navigate to a specified target in an unseen environment. Since the environment layout is unknown, the agent needs to infer the unknown contextual objects from partially observations, thereby deducing the likely location of the target. Previous end-to-end RL methods capture contextual relationships through implicit representations while they lack notion of geometry. Alternatively, modular methods construct local maps for recording the observed geometric structure of unseen environment, however, lacking the reasoning of contextual relation limits the exploration efficiency. In this work, we propose the self-supervised generative map (SGM), a modular method that learns the explicit context relation via self-supervised learning. The SGM is trained to leverage both episodic observations and general knowledge to reconstruct the masked pixels of a cropped global map. During navigation, the agent maintains an incomplete local semantic map, meanwhile, the unknown regions of the local map are generated by the pretrained SGM. Based on the generated map, the agent sets the predicted location of the target as the goal and moves towards it. Experiments on Gibson, MP3D and HM3D show the effectiveness of our method. The code is available at https://github.com/sx-zhang/SGM.

Object goal navigation requires an agent to navigate to a specified object in an unseen environment based on visual observations and user-specified goals. Human decision-making in navigation is sequential, planning a most likely sequence of actions toward the goal. However, existing ObjectNav methods, both end-to-end learning methods and modular methods, rely on single-step planning. They output the next action based on the current model input, which easily overlooks temporal consistency and leads to myopic planning. To this end, we aim to learn sequence planning for ObjectNav. Specifically, we propose trajectory diffusion to learn the distribution of trajectory sequences conditioned on the current observation and the goal. We utilize DDPM and automatically collected optimal trajectory segments to train the trajectory diffusion. Once the trajectory diffusion model is trained, it can generate a temporally coherent sequence of future trajectory for agent based on its current observations. Experimental results on the Gibson and MP3D datasets demonstrate that the generated trajectories effectively guide the agent, resulting in more accurate and efficient navigation.

The goal of few-shot learning is to learn the generalization from seen to unseen data with only a few samples. Most previous few-shot learning methods focus on learning the generalization within particular domains. However, the more practical scenarios may also require the generalization ability across domains. In this paper, we study the problem of few-shot domain generalization (FSDG), which is a more challenging variant of few-shot classification. FSDG requires additional generalization with larger gap from seen domains to unseen domains. We address FSDG problem by meta-learning two levels of meta-knowledge, where the lower-level meta-knowledge is domain-specific embedding spaces as subspaces of a base space for intra-domain generalization, and the upper-level meta-knowledge is the base space and a prior subspace over domain-specific spaces for inter-domain generalization. We formulate the two levels of meta-knowledge learning problem with bi-level optimization, and further develop an optimization algorithm without higher-order derivative information to solve it. We demonstrate our method is significantly superior to the previous works by evaluating it on the widely used benchmark Meta-Dataset.

Previous works for ObjectNav task attempt to learn the association (e.g. relation graph) between the visual inputs and the goal during training. Such association contains the prior knowledge of navigating in training environments, which is denoted as the experience. The experience performs a positive effect on helping the agent infer the likely location of the goal when the layout gap between the unseen environments of the test and the prior knowledge obtained in training is minor. However, when the layout gap is significant, the experience exerts a negative effect on navigation. Motivated by keeping the positive effect and removing the negative effect of the experience, we propose the layout-based soft Total Direct Effect (L-sTDE) framework based on the causal inference to adjust the prediction of the navigation policy. In particular, we propose to calculate the layout gap which is defined as the KL divergence between the posterior and the prior distribution of the object layout. Then the sTDE is proposed to appropriately control the effect of the experience based on the layout gap. Experimental results on AI2THOR, RoboTHOR, and Habitat demonstrate the effectiveness of our method. The code is available at https://github.com/sx-zhang/Layout-based-sTDE.git.

The goal of visual object navigation for an agent is to find the target objects accurately. Recent works mainly focus on the feature of embedding, attempting to learn better features with different variants, such as object distribution and graph representations. However, some typical navigation problems in complex environments, such as partially known and obstacle problems, may not be effectively addressed by previous feature embedding methods. In this paper, we propose a framework with a long-short objective policy, where the hidden states are classified according to the navigation objectives at that moment and separately rewarded. Specifically, we consider two objectives: the long-term objective is to go closer to the target, and the short-term objective is for obstacle avoidance and exploration. To alleviate the effect of long-term and short-term alternation, we build a state memory and propose an adjustment gate to update the state memory. Finally, all past hidden states are reweighted and combined for action prediction with an action-boosting gate. Experimental results on RoboTHOR show that the proposed method can significantly outperform the state-of-the-art.

Getting trust is crucial for embodied agents (such as robots and autonomous vehicles) to collaborate with human beings, especially non-experts. The most direct way for mutual understanding is through natural language explanation. Existing researches consider generating visual explanations for object recognition, while the exploration of explaining embodied decisions remains vacant. In this paper, we study generating action decisions and explanations based on visual observation. Distinct to explanations for recognition, justifying an action needs to show why it's better than other actions. Besides, the understanding of scene structure is required since the agent needs to interact with the environment (e.g. navigation, moving objects). We introduce a new dataset THOR-EAE (Embodied Action Explanation) collected based on AI2-THOR simulator. The dataset consists of over 840,000 egocentric images of indoor embodied observation which are annotated with the optimal action labels and explanation sentences. An explainable decision-making criterion is developed considering scene layout and action attributes for efficient annotation. We propose a graph action justification model, exploiting graph neural networks for obstacle-surroundings relations representation and justifying the actions under the guidance of decision results. Experimental results on THOR-EAE dataset showcase its challenge and the effectiveness of the proposed method.

Visual navigation has been widely studied under the assumption that there may be several clear routes to reach the goal. However, in more practical scenarios such as a house with several messy rooms, there may not. Interactive Navigation (InterNav) considers agents navigating to their goals more effectively with object interactions, posing new challenges of learning interaction dynamics and extra action space. Previous works learn single vision-to-action policy with the guidance of designed representations. However, the causality between actions and outcomes is prone to be confounded when the attributes of obstacles are diverse and hard to measure. Learning policy for long-term action planning in complex scenes also leads to extensive inefficient exploration. In this paper, we introduce a causal diagram of InterNav clarifying the confounding bias caused by obstacles. To address the problem, we propose a multi-policy model that enables the exploration of counterfactual interactions as well as reduces unnecessary exploration. We develop a large-scale dataset containing 600k task episodes in 12k multi-room scenes based on the ProcTHOR simulator and showcase the effectiveness of our method with the evaluations on our dataset.

Embodied AI, where the agent learns to accomplish tasks through interaction with its surrounding environment, is drawing increasing attention in the community. As a challenging Embodied AI task, visual room rearrangement aims to restore the initially misplaced objects in a room to the target state. Existing approaches usually use a single policy to learn a mapping from visual observation to action. Those methods may be capable of accomplishing tasks with simple goals such as visual navigation. However, the agent in the rearrangement task has to explore various types of interaction for a long time. Only considering a single policy may easily get stuck in local optimum. In this paper, we propose a Phase Discriminated Multi-Policy (PDMP) model, decomposing the task into specific phases and tackling them with customized policies. In particular, we first introduce the graph representation of object relationships providing scene layout knowledge, which is discriminated to task phases. Then based on the knowledge a hierarchical actor-critic module is proposed to dynamically call the policies capable of navigation or object interaction. Each policy is trained with narrowed action space and dense rewards so that they can better converge and cooperate to reach long-term goals. Comprehensive experiments based on the AI2-THOR platform, show that the proposed model achieves better performance than baselines.

The goal of few-shot image recognition (FSIR) is to identify novel categories with a small number of annotated samples by exploiting transferable knowledge from training data (base categories). Most current studies assume that the transferable knowledge can be well used to identify novel categories. However, such transferable capability may be impacted by the dataset bias, and this problem has rarely been investigated before. Besides, most of few-shot learning methods are biased to different datasets, which is also an important issue that needs to be investigated deeply. In this paper, we first investigate the impact of transferable capabilities learned from base categories. Specifically, we use the relevance to measure relationships between base categories and novel categories. Distributions of base categories are depicted via the instance density and category diversity. The FSIR model learns better transferable knowledge from relevant training data. In the relevant data, dense instances or diverse categories can further enrich the learned knowledge. Experimental results on different sub-datasets of Imagenet demonstrate category relevance, instance density and category diversity can depict transferable bias from distributions of base categories. Second, we investigate performance differences on different datasets from the aspects of dataset structures and different few-shot learning methods. Specifically, we introduce image complexity, intra-concept visual consistency, and inter-concept visual similarity to quantify characteristics of dataset structures. We use these quantitative characteristics and eight few-shot learning methods to analyze performance differences on multiple datasets. Based on the experimental analysis, some insightful observations are obtained from the perspective of both dataset structures and few-shot learning methods. We hope these observations are useful to guide future few-shot learning research on new datasets or tasks. Our data is available at http://123.57.42.89/dataset-bias/dataset-bias.html.

The goal of few-shot image recognition is to classify different categories with only one or a few training samples. Previous works of few-shot learning mainly focus on simple images, such as object or character images. Those works usually use a convolutional neural network (CNN) to learn the global image representations from training tasks, which are then adapted to novel tasks. However, there are many more abstract and complex images in real world, such as scene images, consisting of many object entities with flexible spatial relations among them. In such cases, global features can hardly obtain satisfactory generalization ability due to the large diversity of object relations in the scenes, which may hinder the adaptability to novel scenes. This paper proposes a composite object relation modeling method for few-shot scene recognition, capturing the spatial structural characteristic of scene images to enhance adaptability on novel scenes, considering that objects commonly co- occurred in different scenes. In different few-shot scene recognition tasks, the objects in the same images usually play different roles. Thus we propose a task-aware region selection module (TRSM) to further select the detected regions in different few-shot tasks. In addition to detecting object regions, we mainly focus on exploiting the relations between objects, which are more consistent to the scenes and can be used to cleave apart different scenes. Objects and relations are used to construct a graph in each image, which is then modeled with graph convolutional neural network. The graph modeling is jointly optimized with few-shot recognition, where the loss of few-shot learning is also capable of adjusting graph based representations. Typically, the proposed graph based representations can be plugged in different types of few-shot architectures, such as metric-based and meta-learning methods. Experimental results of few-shot scene recognition show the effectiveness of the proposed method.

Visual object navigation is an essential task of embodied AI, which is letting the agent navigate to the goal object under the user's demand. Previous methods often focus on single-object navigation. However, in real life, human demands are generally continuous and multiple, requiring the agent to implement multiple tasks in sequence. These demands can be addressed by repeatedly performing previous single task methods. However, by dividing multiple tasks into several independent tasks to perform, without the global optimization between different tasks, the agents' trajectories may overlap, reducing the efficiency of navigation. In this paper, we propose an efficient reinforcement learning framework with a hybrid policy for multi-object navigation, aiming to maximally eliminate noneffective actions. First, the visual observations are embedded to detect the semantic entities (such as objects). And the detected objects are memorized and projected into semantic maps, which can also be regarded as a long-term memory of the observed environment. Then a hybrid policy consisting of exploration and long-term planning strategies is proposed to predict the potential target position. In particular, when the target is directly oriented, the policy function makes long-term planning for the target based on the semantic map, which is implemented by a sequence of motion actions. In the alternative, when the target is not oriented, the policy function estimates an object's potential position toward exploring the most possible objects (positions) that have close relations to the target. The relation between different objects is obtained with prior knowledge, which is used to predict the potential target position by integrating with the memorized semantic map. And then a path to the potential target is planned by the policy function. We evaluate our proposed method on two large-scale 3D realistic environment datasets, Gibson and Matterport3D, and the experimental results demonstrate the effectiveness and generalization of the proposed method.

Scene images are usually composed of foreground and background regional contents. Some existing methods propose to extract regional contents with dense grids or objectness region proposals. However, dense grids may split the object into several discrete parts, learning semantic ambiguity for the patches. The objectness methods may focus on particular objects but only pay attention to the foreground contents and do not exploit the background that is key to scene recognition. In contrast, we propose a novel scene recognition framework with amorphous region detection and context modeling. In the proposed framework, discriminative regions are first detected with amorphous contours that can tightly surround the targets through semantic segmentation techniques. In addition, both foreground and background regions are jointly embedded to obtain the scene representations with the graph model. Based on the graph modeling module, we explore the contextual relations between the regions in geometric and morphology aspects, and generate the discriminative representations for scene recognition. Experimental results on MIT67 and SUN397 demonstrate the effectiveness and generality of the proposed method.

The goal of object navigation is to reach the expected objects according to visual information in the unseen environments. Previous works usually implement deep models to train an agent to predict actions in real-time. However, in the unseen environment, when the target object is not in egocentric view, the agent may not be able to make wise decisions due to the lack of guidance. In this paper, we propose a hierarchical object-to-zone (HOZ) graph to guide the agent in a coarse-to-fine manner, and an online-learning mechanism is also proposed to update HOZ according to the real-time observation in new environments. In particular, the HOZ graph is composed of scene nodes, zone nodes and object nodes. With the pre-learned HOZ graph, the real-time observation and the target goal, the agent can constantly plan an optimal path from zone to zone. In the estimated path, the next potential zone is regarded as sub-goal, which is also fed into the deep reinforcement learning model for action prediction. Our methods are evaluated on the AI2-Thor simulator. In addition to widely used evaluation metrics SR and SPL, we also propose a new evaluation metric of SAE that focuses on the effective action rate. Experimental results demonstrate the effectiveness and efficiency of our proposed method. The code is available at https://github.com/sx-zhang/HOZ.git.

Visual object navigation is a fundamental task in Embodied AI. Previous works focus on the category-wise navigation, in which navigating to any possible instance of target object category is considered a success. Those methods may be effective to find the general objects. However, it may be more practical to navigate to the specific instance in our real life, since our particular requirements are usually satisfied with specific instances rather than all instances of one category. How to navigate to the specific instance has been rarely researched before and is typically challenging to current works. In this paper, we introduce a new task of Instance Object Navigation (ION), where instance-level descriptions of targets are provided and instance-level navigation is required. In particular, multiple types of attributes such as colors, materials and object references are involved in the instance-level descriptions of the targets. In order to allow the agent to maintain the ability of instance navigation, we propose a cascade framework with Instance-Relation Graph (IRG) based navigator and instance grounding module. To specify the different instances of the same object categories, we construct instance-level graph instead of category-level one, where instances are regarded as nodes, encoded with the representation of colors, materials and locations (bounding boxes). During navigation, the detected instances can activate corresponding nodes in IRG, which are updated with graph convolutional neural network (GCNN). The final instance prediction is obtained with the grounding module by selecting the candidates (instances) with maximum probability (a joint probability of category, color and material, obtained by corresponding regressors with softmax). For the task evaluation, we build a benchmark for instance-level object navigation on AI2-Thor simulator, where over 27,735 object instance descriptions and navigation groundtruth are automatically obtained through the interaction with the simulator. The proposed model outperforms the baseline in instance-level metrics, showing that our proposed graph model can guide instance object navigation, as well as leaving promising room for further improvement. The project is available at https://github.com/LWJ312/ION.

Scene classification is a valuable classification subtask and has its own characteristics which still needs more in-depth studies. Basically, scene characteristics are distributed over the whole image, which cause the need of “seeing” comprehensive and informative regions. Previous works mainly focus on region discovery and aggregation, while rarely involves the inherent properties of CNN along with its potential ability to satisfy the requirements of scene classification. In this paper, we propose to understand scene images and the scene classification CNN models in terms of the focus area. From this new perspective, we find that large focus area is preferred in scene classification CNN models as a consequence of learning scene characteristics. Meanwhile, the analysis about existing training schemes helps us to understand the effects of focus area, and also raises the question about optimal training method for scene classification. Pursuing the better usage of scene characteristics, we propose a new learning scheme with a tailored loss in the goal of activating larger focus area on scene images. Since the supervision of the target regions to be enlarged is usually lacked, our alternative learning scheme is to erase already activated area, and allow the CNN models to activate more area during training. The proposed scheme is implemented by keeping the pairwise consistency between the output of the erased image and its original one. In particular, a tailored loss is proposed to keep such pairwise consistency by leveraging category-relevance information. Experiments on Places365 show the significant improvements of our method with various CNNs. Our method shows an inferior result on the object-centric dataset, ImageNet, which experimentally indicates that it captures the unique characteristics of scenes.

Recognizing visual categories from semantic descriptions is a promising way to extend the capability of a visual classifier beyond the concepts represented in the training data (i.e. seen categories). This problem is addressed by (generalized) zero-shot learning methods (GZSL), which leverage semantic descriptions that connect them to seen categories (e.g. label embedding, attributes). Conventional GZSL are designed mostly for object recognition. In this paper we focus on zero-shot scene recognition, a more challenging setting with hundreds of categories where their differences can be subtle and often localized in certain objects or regions. Conventional GZSL representations are not rich enough to capture these local discriminative differences. Addressing these limitations, we propose a feature generation framework with two novel components: 1) multiple sources of semantic information (i.e. attributes, word embeddings and descriptions), 2) region descriptions that can enhance scene discrimination. To generate synthetic visual features we propose a two-step generative approach, where local descriptions are sampled and used as conditions to generate visual features. The generated features are then aggregated and used together with real features to train a joint classifier. In order to evaluate the proposed method, we introduce a new dataset for zero-shot scene recognition with multi-semantic annotations. Experimental results on the proposed dataset and SUN Attribute dataset illustrate the effectiveness of the proposed method.

Exploiting the spatial structure in scene images is a key research direction for scene recognition. Due to the large intra-class structural diversity, building and modeling flexible structural layout to adapt various image characteristics is a challenge. Existing structural modeling methods in scene recognition either focus on predefined grids or rely on learned prototypes, which all have limited representative ability. In this paper, we propose Prototype-agnostic Scene Layout (PaSL) construction method to build the spatial structure for each image without conforming to any prototype. Our PaSL can flexibly capture the diverse spatial characteristic of scene images and have considerable generalization capability. Given a PaSL, we build Layout Graph Network (LGN) where regions in PaSL are defined as nodes and two kinds of independent relations between regions are encoded as edges. The LGN aims to incorporate two topological structures (formed in spatial and semantic similarity dimensions) into image representations through graph convolution. Extensive experiments show that our approach achieves state-of-the-art results on widely recognized MIT67 and SUN397 datasets without multi-model or multi-scale fusion. Moreover, we also conduct the experiments on one of the largest scale datasets, Places365. The results demonstrate the proposed method can be well generalized and obtains competitive performance.

Visual attention has been successfully applied in image captioning to selectively incorporate the most relevant areas to the language generation procedure. However, the attention in current image captioning methods is only guided by the hidden state of language model, e.g. LSTM (Long-Short Term Memory), indirectly and implicitly, and thus the attended areas are weakly relevant at different time steps. Besides the spatial relationship of attention areas, the temporal relationship in attention is crucial for image captioning according to the attention transmission mechanism of human vision. In this paper, we propose a new spatio-temporal memory attention (STMA) model to learn the spatio-temporal relationship in attention for image captioning. The STMA introduces the memory mechanism to the attention model through a tailored LSTM, where the new cell is used to memorize and propagate the attention information, and the output gate is used to generate attention weights. The attention in STMA transmits with memory adaptively and dependently, which builds strong temporal connections of attentions and learns the spatio-temporal relationship of attended areas simultaneously. Besides, the proposed STMA is flexible to combine with attention-based image captioning frameworks. Experiments on MS COCO dataset demonstrate the superiority of the proposed STMA model in exploring the spatio-temporal relationship in attention and improving the current attention-based image captioning.

Scene recognition is challenging due to the intra-class diversity and inter-class similarity. Previous works recognize scenes either with global representations or with the intermediate representations of objects. In contrast, we investigate more discriminative image representations of object-to-object relations for scene recognition, which are based on the triplets of obtained with detection techniques. Particularly, two types of representations, including co-occurring frequency of object-to-object relation (denoted as COOR) and sequential representation of object-to-object relation (denoted as SOOR), are proposed to describe objects and their relative relations in different forms. COOR is represented as the intermediate representation of co-occurring frequency of objects and their relations, with a three order tensor that can be fed to scene classifier without further embedding. SOOR is represented in a more explicit and freer form that sequentially describe image contents with local captions. And a sequence encoding model (e.g., recurrent neural network (RNN)) is implemented to encode SOOR to the features for feeding the classifiers. In order to better capture the spatial information, the proposed COOR and SOOR are adapted to RGB-D data, where a RGB-D proposal fusion method is proposed for RGB-D object detection. With the proposed approaches COOR and SOOR, we obtain the state-of-the-art results of RGB-D scene recognition on SUN RGB-D and NYUD2 datasets.

Scene recognition has been a challenging task in the field of computer vision and multimedia for a long time. The current scene recognition works often extract object features and scene features through CNN, and combine these two types of features to obtain complementary and discriminative scene representations. However, when the scene categories are visually similar, the object features might lack of discriminations. Therefore, it may be debatable to consider only object features. In contrast to the existing works, in this paper, we discuss the discrimination of scene attributes in local regions and utilize scene attributes as the complementary features of object and scene features. We extract these visual features from two individual CNN branches, one extracting the global features of the image while the other extracting the features of local regions. Through contextual modeling framework, we aggregate these features and generate more discriminative scene representations, which achieve better performance than the feature aggregation of object and scene. Moreover, we achieve the new state-of-the-art performance on both standard scene recognition benchmarks by aggregating more complementary visual features: MIT67 (88.06%) and SUN397 (74.12%).

Depth data captured by the cameras such as Microsoft Kinect can bring depth information than traditional RGB data, which is also more robust to different environments, such as dim or dark lighting conditions. In this technical demonstration, we build a scene recognition system based on real-time processing of RGB-D video streams. Our system recognizes the scenes with video clips, where three types of threads are implemented to ensure the realtime. This system first buffers the frames of both RGB and depth videos with the capturing threads. When the buffered videos reach the certain length, the frames will be packed into clips and forwarded in a pre-trained C3D model to predict scene labels with the scene recognition thread. Finally, the predicted scene labels and captured videos are illustrated in our user interface with illustration thread.

Due to the abstraction of scenes, comprehensive scene understanding requires semantic modeling in both global and local aspects. Scene recognition is usually researched from a global point of view, while dense captioning is typically studied for local regions. Previous works separately research on the modeling of scene recognition and dense captioning. In contrast, we propose a joint learning framework that benefits from the mutual coupling of scene recognition and dense captioning models. Generally, these two tasks are coupled through two steps, 1) fusing the supervision by considering the contexts between scene labels and local captions, and 2) jointly optimizing semantically symmetric LSTM models. Particularly, in order to balance bias between dense captioning and scene recognition, a scene adaptive non-maximum suppression (NMS) method is proposed to emphasize the scene related regions in region proposal procedure, and a region-wise and category-wise weighted pooling method is proposed to avoid over attention on particular regions in local to global pooling procedure. For the model training and evaluation, scene labels are manually annotated for Visual Genome database. The experimental results on Visual Genome show the effectiveness of the proposed method. Moreover, the proposed method also can improve previous CNN based works on public scene databases, such as MIT67 and SUN397.

With the developments of deep learning, previous approaches have made successes in scene recognition with massive RGB data obtained from the ideal environments. However, scene recognition in real world may face various types of aberrant conditions caused by different unavoidable factors, such as the lighting variance of the environments and the limitations of cameras, which may damage the performance of previous models. In addition to ideal conditions, our motivation is to investigate researches on robust scene recognition models for unconstrained environments. In this paper, we propose an aberrance-aware framework for RGB-D scene recognition, where several types of attentions, such as temporal, spatial and modal attentions are integrated to spatio-temporal RGB-D CNN models to avoid the interference of RGB frame blurring, depth missing, and light variance. All the attentions are homogeneously obtained by projecting the gradient-sensitive maps of visual data into corresponding spaces. Particularly, the gradient maps are captured with the convolutional operations with the typically designed kernels, which can be seamlessly integrated into end-to-end CNN training. The experiments under different challenging conditions demonstrate the effectiveness of the proposed method.

Deep convolutional networks (CNN) can achieve impressive results on RGB scene recognition thanks to large datasets such as Places. In contrast, RGB-D scene recognition is still underdeveloped in comparison, due to two limitations of RGB-D data we address in this paper. The first limitation is the lack of depth data for training deep learning models. Rather than fine tuning or transferring RGB-specific features, we address this limitation by proposing an architecture and a twostep training approach that directly learns effective depth-specific features using weak supervision via patches. The resulting RGBD model also benefits from more complementary multimodal features. Another limitation is the short range of depth sensors (typically 0.5m to 5.5m), resulting in depth images not capturing distant objects in the scenes that RGB images can. We show that this limitation can be addressed by using RGB-D videos, where more comprehensive depth information is accumulated as the camera travels across the scenes. Focusing on this scenario, we introduce the ISIA RGB-D video dataset to evaluate RGB-D scene recognition with videos. Our video recognition architecture combines convolutional and recurrent neural networks (RNNs) that are trained in three steps with increasingly complex data to learn effective features (i.e. patches, frames and sequences). Our approach obtains state-of-the-art performances on RGB-D image (NYUD2 and SUN RGB-D) and video (ISIA RGB-D) scene recognition.

Scene classification is a challenging problem. Compared with object images, scene images are more abstract, as they are composed of objects. Object and scene images have different characteristics with different scales and composition structures. How to effectively integrate the local mid-level semantic representations including both object and scene concepts needs to be investigated, which is an important aspect for scene classification. In this article, the idea of a sharing codebook is introduced by organically integrating deep learning, concept feature, and local feature encoding techniques. More specifically, the shared local feature codebook is generated from the combined ImageNet1K and Places365 concepts (Mixed1365) using convolutional neural networks. As the Mixed1365 features cover all the semantic information including both object and scene concepts, we can extract a shared codebook from the Mixed1365 features, which only contain a subset of the whole 1,365 concepts with the same codebook size. The shared codebook can not only provide complementary representations without additional codebook training but also be adaptively extracted toward different scene classification tasks. A method of fusing the encoded features with both the original codebook and the shared codebook is proposed for scene classification. In this way, more comprehensive and representative image features can be generated for classification. Extensive experimentations conducted on two public datasets validate the effectiveness of the proposed method. Besides, some useful observations are also revealed to show the advantage of shared codebook.

Scene recognition with RGB images has been extensively studied and has reached very remarkable recognition levels, thanks to convolutional neural networks (CNN) and large scene datasets. In contrast, current RGB-D scene data is much more limited, so often leverages RGB large datasets, by transferring pretrained RGB CNN models and fine-tuning with the target RGB-D dataset. However, we show that this approach has the limitation of hardly reaching bottom layers, which is key to learn modality-specific features. In contrast, we focus on the bottom layers, and propose an alternative strategy to learn depth features combining local weakly supervised training from patches followed by global fine tuning with images. This strategy is capable of learning very discriminative depth-specific features with limited depth images, without resorting to Places-CNN. In addition we propose a modified CNN architecture to further match the complexity of the model and the amount of data available. For RGB-D scene recognition, depth and RGB features are combined by projecting them in a common space and further leaning a multilayer classifier, which is jointly optimized in an end-to-end network. Our framework achieves state-of-the-art accuracy on NYU2 and SUN RGB-D in both depth only and combined RGB-D data.

Image captioning has recently received much attention. Existing approaches, however, are limited to describing images with simple contextual information, which typically generate one sentence to describe each image with only a single contextual emphasis. In this paper, we address this limitation from a user perspective with a novel approach. Given some keywords as additional inputs, the proposed method would generate various descriptions according to the provided guidance. Hence, descriptions with different focuses can be generated for the same image. Our method is based on a new Context-dependent Bilateral Long Short-Term Memory (CDB-LSTM) model to predict a keyword-driven sentence by considering the word dependence. The word dependence is explored externally with a bilateral pipeline, and internally with a unified and joint training process. Experiments on the MS COCO dataset demonstrate that the proposed approach not only significantly outperforms the baseline method but also shows good adaptation and consistency with various keywords.

Depth can complement RGB with useful cues about object volumes and scene layout. However, RGB-D image datasets are still too small for directly training deep convolutional neural networks (CNNs), in contrast to the massive monomodal RGB datasets. Previous works in RGB-D recognition typically combine two separate networks for RGB and depth data, pretrained with a large RGB dataset and then fine tuned to the respective target RGB and depth datasets. These approaches have several limitations: 1) only use low-level filters learned from RGB data, thus not being able to exploit properly depth-specific patterns, and 2) RGB and depth features are only combined at high-levels but rarely at lower-levels. In this paper, we propose a framework that leverages both knowledge acquired from large RGB datasets together with depth-specific cues learned from the limited depth data, obtaining more effective multi-source and multi-modal representations. We propose a multi-modal combination method that selects discriminative combinations of layers from the different source models and target modalities, capturing both high-level properties of the task and intrinsic low-level properties of both modalities.

A scene is usually abstract that consists of several less abstract entities such as objects or themes. It is very difficult to reason scenes from visual features due to the semantic gap between the abstract scenes and low-level visual features. Some alternative works recognize scenes with a two-step framework by representing images with intermediate representations of objects or themes. However, the object co-occurrences between scenes may lead to ambiguity for scene recognition. In this paper, we propose a framework to represent images with intermediate (object) representations with spatial layout, i.e., object-to-object relation (OOR) representation. In order to better capture the spatial information, the proposed OOR is adapted to RGB-D data. In the proposed framework, we first apply object detection technique on RGB and depth images separately. Then the detected results of both modalities are combined with a RGB-D proposal fusion process. Based on the detected results, we extract semantic feature OOR and regional convolutional neural network (CNN) features located by bounding boxes. Finally, different features are concatenated to feed to the classifier for scene recognition. The experimental results on SUN RGB-D and NYUD2 datasets illustrate the efficiency of the proposed method.

Before the big data era, scene recognition was often approached with two-step inference using localized intermediate representations (objects, topics, and so on). One of such approaches is the semantic manifold (SM), in which patches and images are modeled as points in a semantic probability simplex. Patch models are learned resorting to weak supervision via image labels, which leads to the problem of scene categories co-occurring in this semantic space. Fortunately, each category has its own co-occurrence patterns that are consistent across the images in that category. Thus, discovering and modeling these patterns are critical to improve the recognition performance in this representation. Since the emergence of large data sets, such as ImageNet and Places, these approaches have been relegated in favor of the much more powerful convolutional neural networks (CNNs), which can automatically learn multi-layered representations from the data. In this paper, we address many limitations of the original SM approach and related works. We propose discriminative patch representations using neural networks and further propose a hybrid architecture in which the semantic manifold is built on top of multiscale CNNs. Both representations can be computed significantly faster than the Gaussian mixture models of the original SM. To combine multiple scales, spatial relations, and multiple features, we formulate rich context models using Markov random fields. To solve the optimization problem, we analyze global and local approaches, where a top-down hierarchical algorithm has the best performance. Experimental results show that exploiting different types of contextual relations jointly consistently improves the recognition accuracy.

Recently, automatic generation of image captions has attracted great interest not only because of its extensive applications but also because it connects computer vision and natural language processing. By combining convolutional neural networks (CNNs), which learn visual representations from images, and recurrent neural networks (RNNs), which translate the learned features into text sequences, the content of a image can be transformed into linguistic sequences. Existing approaches typically focus on visual features extracted form an object-oriented CNN (train on ImageNet) and then decode them into natural language. In this paper, we propose a novel model using not only object-related, but also scene-related information extracted from the images. To make full use of both object and scene information, we first combine object information and scene information (extracted from a scene-oriented CNN), and then using as inputs to RNNs. Both types of information provide complementary aspects that help in generating a more complete description of the image. Qualitative and quantitative evaluation results validate the effectiveness of our method.

In the semantic multinomial framework patches and images are modeled as points in a semantic probability simplex. Patch theme models are learned resorting to weak supervision via image labels, which leads the problem of scene categories co-occurring in this semantic space. Fortunately, each category has its own co-occurrence patterns that are consistent across the images in that category. Thus, discovering and modeling these patterns is critical to improve the recognition performance in this representation. In this paper, we observe that not only global co-occurrences at the image-level are important, but also different regions have different category co-occurrence patterns. We exploit local contextual relations to address the problem of discovering consistent co-occurrence patterns and removing noisy ones. Our hypothesis is that a less noisy semantic representation, would greatly help the classifier to model consistent co-occurrences and discriminate better between scene categories. An important advantage of modeling features in a semantic space is that this space is feature independent. Thus, we can combine multiple features and spatial neighbors in the same common space, and formulate the problem as minimizing a context-dependent energy. Experimental results show that exploiting different types of contextual relations consistently improves the recognition accuracy. In particular, larger datasets benefit more from the proposed method, leading to very competitive performance.

Abstract Automatically describing the content of an image is a fundamental problem in artificial intelligence that connects computer vision and natural language processing. In contrast to the previous image description methods that focus on describing the whole image, this paper presents a method of generating rich image descriptions from image regions. First, we detect regions with R-CNN (regions with convolutional neural network features) framework. We then utilize the RNN (recurrent neural networks) to generate sentences for image regions. Finally, we propose an optimization method to select one suitable region. The proposed model generates several sentence description of regions in an image, which has sufficient representative power of the whole image and contains more detailed information. Comparing to general image level description, generating more specific and accurate sentences on the different regions can satisfy more personal requirements for different people. Experimental evaluations validate the effectiveness of the proposed method.

Food-related photos have become increasingly popular , due to social networks, food recommendations, and dietary assessment systems. Reliable annotation is essential in those systems, but unconstrained automatic food recognition is still not accurate enough. Most works focus on exploiting only the visual content while ignoring the context. To address this limitation, in this paper we explore leveraging geolocation and external information about restaurants to simplify the classification problem. We propose a framework incorporating discriminative classification in geolocalized settings and introduce the concept of geolocalized models, which, in our scenario, are trained locally at each restaurant location. In particular, we propose two strategies to implement this framework: geolocalized voting and combinations of bundled classifiers. Both models show promising performance, and the later is particularly efficient and scalable. We collected a restaurant-oriented food dataset with food images, dish tags, and restaurant-level information, such as the menu and geolocation. Experiments on this dataset show that exploiting geolocation improves around 30% the recognition performance, and geolocalized models contribute with an additional 3-8% absolute gain, while they can be trained up to five times faster.