This paper presents a simple MLP-like architecture, CycleMLP, which is a versatile backbone for visual recognition and dense predictions. As compared to modern MLP architectures, e.g., MLP-Mixer, ResMLP, and gMLP, whose architectures are correlated to image size and thus are infeasible in object detection and segmentation, CycleMLP has two advantages compared to modern approaches. (1) It can cope with various image sizes. (2) It achieves linear computational complexity to image size by using local windows. In contrast, previous MLPs have $O(N^2)$ computations due to fully spatial connections. We build a family of models which surpass existing MLPs and even state-of-the-art Transformer-based models, e.g., Swin Transformer, while using fewer parameters and FLOPs. We expand the MLP-like models' applicability, making them a versatile backbone for dense prediction tasks. CycleMLP achieves competitive results on object detection, instance segmentation, and semantic segmentation. In particular, CycleMLP-Tiny outperforms Swin-Tiny by 1.3% mIoU on ADE20K dataset with fewer FLOPs. Moreover, CycleMLP also shows excellent zero-shot robustness on ImageNet-C dataset. Code is available at https://github.com/ShoufaChen/CycleMLP.
By considering the spatial correspondence, dense self-supervised representation learning has achieved superior performance on various dense prediction tasks. However, the pixel-level correspondence tends to be noisy because of many similar misleading pixels, e.g., backgrounds. To address this issue, in this paper, we propose to explore \textbf{set} \textbf{sim}ilarity (SetSim) for dense self-supervised representation learning. We generalize pixel-wise similarity learning to set-wise one to improve the robustness because sets contain more semantic and structure information. Specifically, by resorting to attentional features of views, we establish corresponding sets, thus filtering out noisy backgrounds that may cause incorrect correspondences. Meanwhile, these attentional features can keep the coherence of the same image across different views to alleviate semantic inconsistency. We further search the cross-view nearest neighbours of sets and employ the structured neighbourhood information to enhance the robustness. Empirical evaluations demonstrate that SetSim is superior to state-of-the-art methods on object detection, keypoint detection, instance segmentation, and semantic segmentation.
We propose an approach to instance segmentation from 3D point clouds based on dynamic convolution. This enables it to adapt, at inference, to varying feature and object scales. Doing so avoids some pitfalls of bottom up approaches, including a dependence on hyper-parameter tuning and heuristic post-processing pipelines to compensate for the inevitable variability in object sizes, even within a single scene. The representation capability of the network is greatly improved by gathering homogeneous points that have identical semantic categories and close votes for the geometric centroids. Instances are then decoded via several simple convolution layers, where the parameters are generated conditioned on the input. The proposed approach is proposal-free, and instead exploits a convolution process that adapts to the spatial and semantic characteristics of each instance. A light-weight transformer, built on the bottleneck layer, allows the model to capture long-range dependencies, with limited computational overhead. The result is a simple, efficient, and robust approach that yields strong performance on various datasets: ScanNetV2, S3DIS, and PartNet. The consistent improvements on both voxel- and point-based architectures imply the effectiveness of the proposed method. Code is available at: https://git.io/DyCo3D
Amodal segmentation is a new direction of instance segmentation while considering the segmentation of the visible and occluded parts of the instance. The existing state-of-the-art method uses multi-task branches to predict the amodal part and the visible part separately and subtract the visible part from the amodal part to obtain the occluded part. However, the amodal part contains visible information. Therefore, the separated prediction method will generate duplicate information. Different from this method, we propose a method of amodal segmentation based on the idea of the jigsaw. The method uses multi-task branches to predict the two naturally decoupled parts of visible and occluded, which is like getting two matching jigsaw pieces. Then put the two jigsaw pieces together to get the amodal part. This makes each branch focus on the modeling of the object. And we believe that there are certain rules in the occlusion relationship in the real world. This is a kind of occlusion context information. This jigsaw method can better model the occlusion relationship and use the occlusion context information, which is important for amodal segmentation. Experiments on two widely used amodally annotated datasets prove that our method exceeds existing state-of-the-art methods. The source code of this work will be made public soon.
Identifying potential threats concealed within the baggage is of prime concern for the security staff. Many researchers have developed frameworks that can detect baggage threats from X-ray scans. However, to the best of our knowledge, all of these frameworks require extensive training on large-scale and well-annotated datasets, which are hard to procure in the real world. This paper presents a novel unsupervised anomaly instance segmentation framework that recognizes baggage threats, in X-ray scans, as anomalies without requiring any ground truth labels. Furthermore, thanks to its stylization capacity, the framework is trained only once, and at the inference stage, it detects and extracts contraband items regardless of their scanner specifications. Our one-staged approach initially learns to reconstruct normal baggage content via an encoder-decoder network utilizing a proposed stylization loss function. The model subsequently identifies the abnormal regions by analyzing the disparities within the original and the reconstructed scans. The anomalous regions are then clustered and post-processed to fit a bounding box for their localization. In addition, an optional classifier can also be appended with the proposed framework to recognize the categories of these extracted anomalies. A thorough evaluation of the proposed system on four public baggage X-ray datasets, without any re-training, demonstrates that it achieves competitive performance as compared to the conventional fully supervised methods (i.e., the mean average precision score of 0.7941 on SIXray, 0.8591 on GDXray, 0.7483 on OPIXray, and 0.5439 on COMPASS-XP dataset) while outperforming state-of-the-art semi-supervised and unsupervised baggage threat detection frameworks by 67.37%, 32.32%, 47.19%, and 45.81% in terms of F1 score across SIXray, GDXray, OPIXray, and COMPASS-XP datasets, respectively.
We propose a method for multi-object tracking and segmentation (MOTS) that does not require fine-tuning or per benchmark hyperparameter selection. The proposed method addresses particularly the data association problem. Indeed, the recently introduced HOTA metric, that has a better alignment with the human visual assessment by evenly balancing detections and associations quality, has shown that improvements are still needed for data association. After creating tracklets using instance segmentation and optical flow, the proposed method relies on a space-time memory network (STM) developed for one-shot video object segmentation to improve the association of tracklets with temporal gaps. To the best of our knowledge, our method, named MeNToS, is the first to use the STM network to track object masks for MOTS. We took the 4th place in the RobMOTS challenge. The project page is https://mehdimiah.com/mentos.html.
In the process of intelligently segmenting foods in images using deep neural networks for diet management, data collection and labeling for network training are very important but labor-intensive tasks. In order to solve the difficulties of data collection and annotations, this paper proposes a food segmentation method applicable to real-world through synthetic data. To perform food segmentation on healthcare robot systems, such as meal assistance robot arm, we generate synthetic data using the open-source 3D graphics software Blender placing multiple objects on meal plate and train Mask R-CNN for instance segmentation. Also, we build a data collection system and verify our segmentation model on real-world food data. As a result, on our real-world dataset, the model trained only synthetic data is available to segment food instances that are not trained with 52.2% mask AP@all, and improve performance by +6.4%p after fine-tuning comparing to the model trained from scratch. In addition, we also confirm the possibility and performance improvement on the public dataset for fair analysis. Our code and pre-trained weights are avaliable online at: https://github.com/gist-ailab/Food-Instance-Segmentation
Segmenting 3D cell nuclei from microscopy image volumes is critical for biological and clinical analysis, enabling the study of cellular expression patterns and cell lineages. However, current datasets for neuronal nuclei usually contain volumes smaller than $10^{\text{-}3}\ mm^3$ with fewer than 500 instances per volume, unable to reveal the complexity in large brain regions and restrict the investigation of neuronal structures. In this paper, we have pushed the task forward to the sub-cubic millimeter scale and curated the NucMM dataset with two fully annotated volumes: one $0.1\ mm^3$ electron microscopy (EM) volume containing nearly the entire zebrafish brain with around 170,000 nuclei; and one $0.25\ mm^3$ micro-CT (uCT) volume containing part of a mouse visual cortex with about 7,000 nuclei. With two imaging modalities and significantly increased volume size and instance numbers, we discover a great diversity of neuronal nuclei in appearance and density, introducing new challenges to the field. We also perform a statistical analysis to illustrate those challenges quantitatively. To tackle the challenges, we propose a novel hybrid-representation learning model that combines the merits of foreground mask, contour map, and signed distance transform to produce high-quality 3D masks. The benchmark comparisons on the NucMM dataset show that our proposed method significantly outperforms state-of-the-art nuclei segmentation approaches. Code and data are available at https://connectomics-bazaar.github.io/proj/nucMM/index.html.
Human vision is able to capture the part-whole hierarchical information from the entire scene. This paper presents the Visual Parser (ViP) that explicitly constructs such a hierarchy with transformers. ViP divides visual representations into two levels, the part level and the whole level. Information of each part represents a combination of several independent vectors within the whole. To model the representations of the two levels, we first encode the information from the whole into part vectors through an attention mechanism, then decode the global information within the part vectors back into the whole representation. By iteratively parsing the two levels with the proposed encoder-decoder interaction, the model can gradually refine the features on both levels. Experimental results demonstrate that ViP can achieve very competitive performance on three major tasks e.g. classification, detection and instance segmentation. In particular, it can surpass the previous state-of-the-art CNN backbones by a large margin on object detection. The tiny model of the ViP family with $7.2\times$ fewer parameters and $10.9\times$ fewer FLOPS can perform comparably with the largest model ResNeXt-101-64$\times$4d of ResNe(X)t family. Visualization results also demonstrate that the learnt parts are highly informative of the predicting class, making ViP more explainable than previous fundamental architectures. Code is available at https://github.com/kevin-ssy/ViP.
Self-Attention has become prevalent in computer vision models. Inspired by fully connected Conditional Random Fields (CRFs), we decompose self-attention into local and context terms. They correspond to the unary and binary terms in CRF and are implemented by attention mechanisms with projection matrices. We observe that the unary terms only make small contributions to the outputs, and meanwhile standard CNNs that rely solely on the unary terms achieve great performances on a variety of tasks. Therefore, we propose Locally Enhanced Self-Attention (LESA), which enhances the unary term by incorporating it with convolutions, and utilizes a fusion module to dynamically couple the unary and binary operations. In our experiments, we replace the self-attention modules with LESA. The results on ImageNet and COCO show the superiority of LESA over convolution and self-attention baselines for the tasks of image recognition, object detection, and instance segmentation. The code is made publicly available.
Electron microscopy (EM) enables the reconstruction of neural circuits at the level of individual synapses, which has been transformative for scientific discoveries. However, due to the complex morphology, an accurate reconstruction of cortical axons has become a major challenge. Worse still, there is no publicly available large-scale EM dataset from the cortex that provides dense ground truth segmentation for axons, making it difficult to develop and evaluate large-scale axon reconstruction methods. To address this, we introduce the AxonEM dataset, which consists of two 30x30x30 um^3 EM image volumes from the human and mouse cortex, respectively. We thoroughly proofread over 18,000 axon instances to provide dense 3D axon instance segmentation, enabling large-scale evaluation of axon reconstruction methods. In addition, we densely annotate nine ground truth subvolumes for training, per each data volume. With this, we reproduce two published state-of-the-art methods and provide their evaluation results as a baseline. We publicly release our code and data at https://connectomics-bazaar.github.io/proj/AxonEM/index.html to foster the development of advanced methods.
Recently, there has been a panoptic segmentation task combining semantic and instance segmentation, in which the goal is to classify each pixel with the corresponding instance ID. In this work, we propose a solution to tackle the panoptic segmentation task. The overall structure combines the bottom-up method and the top-down method. Therefore, not only can there be better performance, but also the execution speed can be maintained. The network mainly pays attention to the quality of the mask. In the previous work, we can see that the uneven contour of the object is more likely to appear, resulting in low-quality prediction. Accordingly, we propose enhancement features and corresponding loss functions for the silhouette of objects and backgrounds to improve the mask. Meanwhile, we use the new proposed confidence score to solve the occlusion problem and make the network tend to use higher quality masks as prediction results. To verify our research, we used the COCO dataset and CityScapes dataset to do experiments and obtained competitive results with fast inference time.
In order to handle the challenges of autonomous driving, deep learning has proven to be crucial in tackling increasingly complex tasks, such as 3D detection or instance segmentation. State-of-the-art approaches for image-based detection tasks tackle this complexity by operating in a cascaded fashion: they first extract a 2D bounding box based on which additional attributes, e.g. instance masks, are inferred. While these methods perform well, a key challenge remains the lack of accurate and cheap annotations for the growing variety of tasks. Synthetic data presents a promising solution but, despite the effort in domain adaptation research, the gap between synthetic and real data remains an open problem. In this work, we propose a weakly supervised domain adaptation setting which exploits the structure of cascaded detection tasks. In particular, we learn to infer the attributes solely from the source domain while leveraging 2D bounding boxes as weak labels in both domains to explain the domain shift. We further encourage domain-invariant features through class-wise feature alignment using ground-truth class information, which is not available in the unsupervised setting. As our experiments demonstrate, the approach is competitive with fully supervised settings while outperforming unsupervised adaptation approaches by a large margin.
We present UrbanScene3D, a large-scale data platform for research of urban scene perception and reconstruction. UrbanScene3D contains over 128k high-resolution images covering 16 scenes including large-scale real urban regions and synthetic cities with 136 km^2 area in total. The dataset also contains high-precision LiDAR scans and hundreds of image sets with different observation patterns, which provide a comprehensive benchmark to design and evaluate aerial path planning and 3D reconstruction algorithms. In addition, the dataset, which is built on Unreal Engine and Airsim simulator together with the manually annotated unique instance label for each building in the dataset, enables the generation of all kinds of data, e.g., 2D depth maps, 2D/3D bounding boxes, and 3D point cloud/mesh segmentations, etc. The simulator with physical engine and lighting system not only produce variety of data but also enable users to simulate cars or drones in the proposed urban environment for future research.
Independently exploring unknown spaces or finding objects in an indoor environment is a daily but challenging task for visually impaired people. However, common 2D assistive systems lack depth relationships between various objects, resulting in difficulty to obtain accurate spatial layout and relative positions of objects. To tackle these issues, we propose HIDA, a lightweight assistive system based on 3D point cloud instance segmentation with a solid-state LiDAR sensor, for holistic indoor detection and avoidance. Our entire system consists of three hardware components, two interactive functions~(obstacle avoidance and object finding) and a voice user interface. Based on voice guidance, the point cloud from the most recent state of the changing indoor environment is captured through an on-site scanning performed by the user. In addition, we design a point cloud segmentation model with dual lightweight decoders for semantic and offset predictions, which satisfies the efficiency of the whole system. After the 3D instance segmentation, we post-process the segmented point cloud by removing outliers and projecting all points onto a top-view 2D map representation. The system integrates the information above and interacts with users intuitively by acoustic feedback. The proposed 3D instance segmentation model has achieved state-of-the-art performance on ScanNet v2 dataset. Comprehensive field tests with various tasks in a user study verify the usability and effectiveness of our system for assisting visually impaired people in holistic indoor understanding, obstacle avoidance and object search.
Stixels have been successfully applied to a wide range of vision tasks in autonomous driving, recently including instance segmentation. However, due to their sparse occurrence in the image, until now Stixels seldomly served as input for Deep Learning algorithms, restricting their utility for such approaches. In this work we present StixelPointNet, a novel method to perform fast instance segmentation directly on Stixels. By regarding the Stixel representation as unstructured data similar to point clouds, architectures like PointNet are able to learn features from Stixels. We use a bounding box detector to propose candidate instances, for which the relevant Stixels are extracted from the input image. On these Stixels, a PointNet models learns binary segmentations, which we then unify throughout the whole image in a final selection step. StixelPointNet achieves state-of-the-art performance on Stixel-level, is considerably faster than pixel-based segmentation methods, and shows that with our approach the Stixel domain can be introduced to many new 3D Deep Learning tasks.
Spatial self-attention layers, in the form of Non-Local blocks, introduce long-range dependencies in Convolutional Neural Networks by computing pairwise similarities among all possible positions. Such pairwise functions underpin the effectiveness of non-local layers, but also determine a complexity that scales quadratically with respect to the input size both in space and time. This is a severely limiting factor that practically hinders the applicability of non-local blocks to even moderately sized inputs. Previous works focused on reducing the complexity by modifying the underlying matrix operations, however in this work we aim to retain full expressiveness of non-local layers while keeping complexity linear. We overcome the efficiency limitation of non-local blocks by framing them as special cases of 3rd order polynomial functions. This fact enables us to formulate novel fast Non-Local blocks, capable of reducing the complexity from quadratic to linear with no loss in performance, by replacing any direct computation of pairwise similarities with element-wise multiplications. The proposed method, which we dub as "Poly-NL", is competitive with state-of-the-art performance across image recognition, instance segmentation, and face detection tasks, while having considerably less computational overhead.
Instance segmentation is an important computer vision problem which remains challenging despite impressive recent advances due to deep learning-based methods. Given sufficient training data, fully supervised methods can yield excellent performance, but annotation of ground-truth data remains a major bottleneck, especially for biomedical applications where it has to be performed by domain experts. The amount of labels required can be drastically reduced by using rules derived from prior knowledge to guide the segmentation. However, these rules are in general not differentiable and thus cannot be used with existing methods. Here, we relax this requirement by using stateless actor critic reinforcement learning, which enables non-differentiable rewards. We formulate the instance segmentation problem as graph partitioning and the actor critic predicts the edge weights driven by the rewards, which are based on the conformity of segmented instances to high-level priors on object shape, position or size. The experiments on toy and real datasets demonstrate that we can achieve excellent performance without any direct supervision based only on a rich set of priors.
Accurate object 6D pose estimation is a core task for robot bin-picking applications, especially when objects are randomly stacked with heavy occlusion. To address this problem, this paper proposes a simple but novel Point-wise Pose Regression Network (PPR-Net). For each point in the point cloud, the network regresses a 6D pose of the object instance that the point belongs to. We argue that the regressed poses of points from the same object instance should be located closely in pose space. Thus, these points can be clustered into different instances and their corresponding objects’ 6D poses can be estimated simultaneously. In our experiments, PPR-Net outperforms the state-of-the-art approach by 15% - 41% in average precision when evaluated on the benchmark Sileane ´ dataset. In addition, it also works well in real world robot bin-picking tasks.
Recent studies indicate that hierarchical Vision Transformer with a macro architecture of interleaved non-overlapped window-based self-attention \& shifted-window operation is able to achieve state-of-the-art performance in various visual recognition tasks, and challenges the ubiquitous convolutional neural networks (CNNs) using densely slid kernels. Most follow-up works attempt to replace the shifted-window operation with other kinds of cross-window communication paradigms, while treating self-attention as the de-facto standard for window-based information aggregation. In this manuscript, we question whether self-attention is the only choice for hierarchical Vision Transformer to attain strong performance, and the effects of different kinds of cross-window communication. To this end, we replace self-attention layers with embarrassingly simple linear mapping layers, and the resulting proof-of-concept architecture termed as LinMapper can achieve very strong performance in ImageNet-1k image recognition. Moreover, we find that LinMapper is able to better leverage the pre-trained representations from image recognition and demonstrates excellent transfer learning properties on downstream dense prediction tasks such as object detection and instance segmentation. We also experiment with other alternatives to self-attention for content aggregation inside each non-overlapped window under different cross-window communication approaches, which all give similar competitive results. Our study reveals that the \textbf{macro architecture} of Swin model families, other than specific aggregation layers or specific means of cross-window communication, may be more responsible for its strong performance and is the real challenger to the ubiquitous CNN's dense sliding window paradigm. Code and models will be publicly available to facilitate future research.