Handwritten documents are often characterized by dense and uneven layout. Despite advances, standard deep network based approaches for semantic layout segmentation are not robust to complex deformations seen across semantic regions. This phenomenon is especially pronounced for the low-resource Indic palm-leaf manuscript domain. To address the issue, we first introduce Indiscapes2, a new large-scale diverse dataset of Indic manuscripts with semantic layout annotations. Indiscapes2 contains documents from four different historical collections and is 150% larger than its predecessor, Indiscapes. We also propose a novel deep network Palmira for robust, deformation-aware instance segmentation of regions in handwritten manuscripts. We also report Hausdorff distance and its variants as a boundary-aware performance measure. Our experiments demonstrate that Palmira provides robust layouts, outperforms strong baseline approaches and ablative variants. We also include qualitative results on Arabic, South-East Asian and Hebrew historical manuscripts to showcase the generalization capability of Palmira.
We present a novel method, called CenterPoly, for real-time instance segmentation using bounding polygons. We apply it to detect road users in dense urban environments, making it suitable for applications in intelligent transportation systems like automated vehicles. CenterPoly detects objects by their center keypoint while predicting a fixed number of polygon vertices for each object, thus performing detection and segmentation in parallel. Most of the network parameters are shared by the network heads, making it fast and lightweight enough to run at real-time speed. To properly convert mask ground-truth to polygon ground-truth, we designed a vertex selection strategy to facilitate the learning of the polygons. Additionally, to better segment overlapping objects in dense urban scenes, we also train a relative depth branch to determine which instances are closer and which are further, using available weak annotations. We propose several models with different backbones to show the possible speed / accuracy trade-offs. The models were trained and evaluated on Cityscapes, KITTI and IDD and the results are reported on their public benchmark, which are state-of-the-art at real-time speeds. Code is available at https://github.com/hu64/CenterPoly
The conventional detectors tend to make imbalanced classification and suffer performance drop, when the distribution of the training data is severely skewed. In this paper, we propose to use the mean classification score to indicate the classification accuracy for each category during training. Based on this indicator, we balance the classification via an Equilibrium Loss (EBL) and a Memory-augmented Feature Sampling (MFS) method. Specifically, EBL increases the intensity of the adjustment of the decision boundary for the weak classes by a designed score-guided loss margin between any two classes. On the other hand, MFS improves the frequency and accuracy of the adjustment of the decision boundary for the weak classes through over-sampling the instance features of those classes. Therefore, EBL and MFS work collaboratively for finding the classification equilibrium in long-tailed detection, and dramatically improve the performance of tail classes while maintaining or even improving the performance of head classes. We conduct experiments on LVIS using Mask R-CNN with various backbones including ResNet-50-FPN and ResNet-101-FPN to show the superiority of the proposed method. It improves the detection performance of tail classes by 15.6 AP, and outperforms the most recent long-tailed object detectors by more than 1 AP. Code is available at https://github.com/fcjian/LOCE.
Instance segmentation in 3D scenes is fundamental in many applications of scene understanding. It is yet challenging due to the compound factors of data irregularity and uncertainty in the numbers of instances. State-of-the-art methods largely rely on a general pipeline that first learns point-wise features discriminative at semantic and instance levels, followed by a separate step of point grouping for proposing object instances. While promising, they have the shortcomings that (1) the second step is not supervised by the main objective of instance segmentation, and (2) their point-wise feature learning and grouping are less effective to deal with data irregularities, possibly resulting in fragmented segmentations. To address these issues, we propose in this work an end-to-end solution of Semantic Superpoint Tree Network (SSTNet) for proposing object instances from scene points. Key in SSTNet is an intermediate, semantic superpoint tree (SST), which is constructed based on the learned semantic features of superpoints, and which will be traversed and split at intermediate tree nodes for proposals of object instances. We also design in SSTNet a refinement module, termed CliqueNet, to prune superpoints that may be wrongly grouped into instance proposals. Experiments on the benchmarks of ScanNet and S3DIS show the efficacy of our proposed method. At the time of submission, SSTNet ranks top on the ScanNet (V2) leaderboard, with 2% higher of mAP than the second best method. The source code in PyTorch is available at https://github.com/Gorilla-Lab-SCUT/SSTNet.
Most existing video tasks related to "human" focus on the segmentation of salient humans, ignoring the unspecified others in the video. Few studies have focused on segmenting and tracking all humans in a complex video, including pedestrians and humans of other states (e.g., seated, riding, or occluded). In this paper, we propose a novel framework, abbreviated as HVISNet, that segments and tracks all presented people in given videos based on a one-stage detector. To better evaluate complex scenes, we offer a new benchmark called HVIS (Human Video Instance Segmentation), which comprises 1447 human instance masks in 805 high-resolution videos in diverse scenes. Extensive experiments show that our proposed HVISNet outperforms the state-of-the-art methods in terms of accuracy at a real-time inference speed (30 FPS), especially on complex video scenes. We also notice that using the center of the bounding box to distinguish different individuals severely deteriorates the segmentation accuracy, especially in heavily occluded conditions. This common phenomenon is referred to as the ambiguous positive samples problem. To alleviate this problem, we propose a mechanism named Inner Center Sampling to improve the accuracy of instance segmentation. Such a plug-and-play inner center sampling mechanism can be incorporated in any instance segmentation models based on a one-stage detector to improve the performance. In particular, it gains 4.1 mAP improvement on the state-of-the-art method in the case of occluded humans. Code and data are available at https://github.com/IIGROUP/HVISNet.
Our work aims to obtain 3D reconstruction of hands and manipulated objects from monocular videos. Reconstructing hand-object manipulations holds a great potential for robotics and learning from human demonstrations. The supervised learning approach to this problem, however, requires 3D supervision and remains limited to constrained laboratory settings and simulators for which 3D ground truth is available. In this paper we first propose a learning-free fitting approach for hand-object reconstruction which can seamlessly handle two-hand object interactions. Our method relies on cues obtained with common methods for object detection, hand pose estimation and instance segmentation. We quantitatively evaluate our approach and show that it can be applied to datasets with varying levels of difficulty for which training data is unavailable.
Most recent transformer-based models show impressive performance on vision tasks, even better than Convolution Neural Networks (CNN). In this work, we present a novel, flexible, and effective transformer-based model for high-quality instance segmentation. The proposed method, Segmenting Objects with TRansformers (SOTR), simplifies the segmentation pipeline, building on an alternative CNN backbone appended with two parallel subtasks: (1) predicting per-instance category via transformer and (2) dynamically generating segmentation mask with the multi-level upsampling module. SOTR can effectively extract lower-level feature representations and capture long-range context dependencies by Feature Pyramid Network (FPN) and twin transformer, respectively. Meanwhile, compared with the original transformer, the proposed twin transformer is time- and resource-efficient since only a row and a column attention are involved to encode pixels. Moreover, SOTR is easy to be incorporated with various CNN backbones and transformer model variants to make considerable improvements for the segmentation accuracy and training convergence. Extensive experiments show that our SOTR performs well on the MS COCO dataset and surpasses state-of-the-art instance segmentation approaches. We hope our simple but strong framework could serve as a preferment baseline for instance-level recognition. Our code is available at https://github.com/easton-cau/SOTR.
Scene understanding is crucial for autonomous systems which intend to operate in the real world. Single task vision networks extract information only based on some aspects of the scene. In multi-task learning (MTL), on the other hand, these single tasks are jointly learned, thereby providing an opportunity for tasks to share information and obtain a more comprehensive understanding. To this end, we develop UniNet, a unified scene understanding network that accurately and efficiently infers vital vision tasks including object detection, semantic segmentation, instance segmentation, monocular depth estimation, and monocular instance depth prediction. As these tasks look at different semantic and geometric information, they can either complement or conflict with each other. Therefore, understanding inter-task relationships can provide useful cues to enable complementary information sharing. We evaluate the task relationships in UniNet through the lens of adversarial attacks based on the notion that they can exploit learned biases and task interactions in the neural network. Extensive experiments on the Cityscapes dataset, using untargeted and targeted attacks reveal that semantic tasks strongly interact amongst themselves, and the same holds for geometric tasks. Additionally, we show that the relationship between semantic and geometric tasks is asymmetric and their interaction becomes weaker as we move towards higher-level representations.
Breast cancer is one of the factors that cause the increase of mortality of women. The most widely used method for diagnosing this geological disease i.e. breast cancer is the ultrasound scan. Several key features such as the smoothness and the texture of the tumor captured through ultrasound scans encode the abnormality of the breast tumors (malignant from benign). However, ultrasound scans are often noisy and include irrelevant parts of the breast that may bias the segmentation of eventual tumors. In this paper, we are going to extract the region of interest ( i.e, bounding boxes of the tumors) and feed-forward them to one semantic segmentation encoder-decoder structure based on its classification (i.e, malignant or benign). the whole process aims to build an instance-based segmenter from a semantic segmenter and an object detector.
Leaf segmentation is the most direct and effective way for high-throughput plant phenotype data analysis and quantitative researches of complex traits. Currently, the primary goal of plant phenotyping is to raise the accuracy of the autonomous phenotypic measurement. In this work, we present the LeafMask neural network, a new end-to-end model to delineate each leaf region and count the number of leaves, with two main components: 1) the mask assembly module merging position-sensitive bases of each predicted box after non-maximum suppression (NMS) and corresponding coefficients to generate original masks; 2) the mask refining module elaborating leaf boundaries from the mask assembly module by the point selection strategy and predictor. In addition, we also design a novel and flexible multi-scale attention module for the dual attention-guided mask (DAG-Mask) branch to effectively enhance information expression and produce more accurate bases. Our main contribution is to generate the final improved masks by combining the mask assembly module with the mask refining module under the anchor-free instance segmentation paradigm. We validate our LeafMask through extensive experiments on Leaf Segmentation Challenge (LSC) dataset. Our proposed model achieves the 90.09% BestDice score outperforming other state-of-the-art approaches.
3D Instance segmentation is a fundamental task in computer vision. Effective segmentation plays an important role in robotic tasks, augmented reality, autonomous driving, etc. With the ascendancy of convolutional neural networks in 2D image processing, the use of deep learning methods to segment 3D point clouds receives much attention. A great convergence of training loss often requires a large amount of human-annotated data, while making such a 3D dataset is time-consuming. This paper proposes a method for training convolutional neural networks to predict instance segmentation results using synthetic data. The proposed method is based on the SGPN framework. We replaced the original feature extractor with “dynamic graph convolutional neural networks” that learned how to extract local geometric features and proposed a simple and effective loss function, making the network more focused on hard examples. We experimentally proved that the proposed method significantly outperforms the state-of-the-art method in both Stanford 3D Indoor Semantics Dataset and our datasets.
Unsupervised learning is just at a tipping point where it could really take off. Among these approaches, contrastive learning has seen tremendous progress and led to state-of-the-art performance. In this paper, we construct a novel probabilistic graphical model that effectively incorporates the low rank promoting prior into the framework of contrastive learning, referred to as LORAC. In contrast to the existing conventional self-supervised approaches that only considers independent learning, our hypothesis explicitly requires that all the samples belonging to the same instance class lie on the same subspace with small dimension. This heuristic poses particular joint learning constraints to reduce the degree of freedom of the problem during the search of the optimal network parameterization. Most importantly, we argue that the low rank prior employed here is not unique, and many different priors can be invoked in a similar probabilistic way, corresponding to different hypotheses about underlying truth behind the contrastive features. Empirical evidences show that the proposed algorithm clearly surpasses the state-of-the-art approaches on multiple benchmarks, including image classification, object detection, instance segmentation and keypoint detection.
Instance segmentation on point clouds is a fundamental task in 3D scene perception. In this work, we propose a concise clustering-based framework named HAIS, which makes full use of spatial relation of points and point sets. Considering clustering-based methods may result in over-segmentation or under-segmentation, we introduce the hierarchical aggregation to progressively generate instance proposals, i.e., point aggregation for preliminarily clustering points to sets and set aggregation for generating complete instances from sets. Once the complete 3D instances are obtained, a sub-network of intra-instance prediction is adopted for noisy points filtering and mask quality scoring. HAIS is fast (only 410ms per frame) and does not require non-maximum suppression. It ranks 1st on the ScanNet v2 benchmark, achieving the highest 69.9% AP50 and surpassing previous state-of-the-art (SOTA) methods by a large margin. Besides, the SOTA results on the S3DIS dataset validate the good generalization ability. Code will be available at https://github.com/hustvl/HAIS.
Transformers have made great progress in dealing with computer vision tasks. However, existing vision transformers do not yet possess the ability of building the interactions among features of different scales, which is perceptually important to visual inputs. The reasons are two-fold: (1) Input embeddings of each layer are equal-scale, so no cross-scale feature can be extracted; (2) to lower the computational cost, some vision transformers merge adjacent embeddings inside the self-attention module, thus sacrificing small-scale (fine-grained) features of the embeddings and also disabling the cross-scale interactions. To this end, we propose Cross-scale Embedding Layer (CEL) and Long Short Distance Attention (LSDA). On the one hand, CEL blends each embedding with multiple patches of different scales, providing the self-attention module itself with cross-scale features. On the other hand, LSDA splits the self-attention module into a short-distance one and a long-distance counterpart, which not only reduces the computational burden but also keeps both small-scale and large-scale features in the embeddings. Through the above two designs, we achieve cross-scale attention. Besides, we put forward a dynamic position bias for vision transformers to make the popular relative position bias apply to variable-sized images. Hinging on the cross-scale attention module, we construct a versatile vision architecture, dubbed CrossFormer, which accommodates variable-sized inputs. Extensive experiments show that CrossFormer outperforms the other vision transformers on image classification, object detection, instance segmentation, and semantic segmentation tasks. The code has been released: https://github.com/cheerss/CrossFormer.
Video Instance Segmentation (VIS) is a new and inherently multi-task problem, which aims to detect, segment, and track each instance in a video sequence. Existing approaches are mainly based on single-frame features or single-scale features of multiple frames, where either temporal information or multi-scale information is ignored. To incorporate both temporal and scale information, we propose a Temporal Pyramid Routing (TPR) strategy to conditionally align and conduct pixel-level aggregation from a feature pyramid pair of two adjacent frames. Specifically, TPR contains two novel components, including Dynamic Aligned Cell Routing (DACR) and Cross Pyramid Routing (CPR), where DACR is designed for aligning and gating pyramid features across temporal dimension, while CPR transfers temporally aggregated features across scale dimension. Moreover, our approach is a light-weight and plug-and-play module and can be easily applied to existing instance segmentation methods. Extensive experiments on three datasets including YouTube-VIS (2019, 2021) and Cityscapes-VPS demonstrate the effectiveness and efficiency of the proposed approach on several state-of-the-art video instance and panoptic segmentation methods. Codes will be publicly available at \url{https://github.com/lxtGH/TemporalPyramidRouting}.
Objective assessment of Magnetic Resonance Imaging (MRI) scans of osteoarthritis (OA) can address the limitation of the current OA assessment. Segmentation of bone, cartilage, and joint fluid is necessary for the OA objective assessment. Most of the proposed segmentation methods are not performing instance segmentation and suffer from class imbalance problems. This study deployed Mask R-CNN instance segmentation and improved it (improved-Mask R-CNN (iMaskRCNN)) to obtain a more accurate generalized segmentation for OA-associated tissues. Training and validation of the method were performed using 500 MRI knees from the Osteoarthritis Initiative (OAI) dataset and 97 MRI scans of patients with symptomatic hip OA. Three modifications to Mask R-CNN yielded the iMaskRCNN: adding a 2nd ROIAligned block, adding an extra decoder layer to the mask-header, and connecting them by a skip connection. The results were assessed using Hausdorff distance, dice score, and coefficients of variation (CoV). The iMaskRCNN led to improved bone and cartilage segmentation compared to Mask RCNN as indicated with the increase in dice score from 95% to 98% for the femur, 95% to 97% for tibia, 71% to 80% for femoral cartilage, and 81% to 82% for tibial cartilage. For the effusion detection, dice improved with iMaskRCNN 72% versus MaskRCNN 71%. The CoV values for effusion detection between Reader1 and Mask R-CNN (0.33), Reader1 and iMaskRCNN (0.34), Reader2 and Mask R-CNN (0.22), Reader2 and iMaskRCNN (0.29) are close to CoV between two readers (0.21), indicating a high agreement between the human readers and both Mask R-CNN and iMaskRCNN. Mask R-CNN and iMaskRCNN can reliably and simultaneously extract different scale articular tissues involved in OA, forming the foundation for automated assessment of OA. The iMaskRCNN results show that the modification improved the network performance around the edges.
In this paper, we focus on the challenging multicategory instance segmentation problem in remote sensing images (RSIs), which aims at predicting the categories of all instances and localizing them with pixel-level masks. Although many landmark frameworks have demonstrated promising performance in instance segmentation, the complexity in the background and scale variability instances still remain challenging for instance segmentation of RSIs. To address the above problems, we propose an end-to-end multi-category instance segmentation model, namely Semantic Attention and Scale Complementary Network, which mainly consists of a Semantic Attention (SEA) module and a Scale Complementary Mask Branch (SCMB). The SEA module contains a simple fully convolutional semantic segmentation branch with extra supervision to strengthen the activation of interest instances on the feature map and reduce the background noise's interference. To handle the under-segmentation of geospatial instances with large varying scales, we design the SCMB that extends the original single mask branch to trident mask branches and introduces complementary mask supervision at different scales to sufficiently leverage the multi-scale information. We conduct comprehensive experiments to evaluate the effectiveness of our proposed method on the iSAID dataset and the NWPU Instance Segmentation dataset and achieve promising performance.
We propose Rank & Sort (RS) Loss, a ranking-based loss function to train deep object detection and instance segmentation methods (i.e. visual detectors). RS Loss supervises the classifier, a sub-network of these methods, to rank each positive above all negatives as well as to sort positives among themselves with respect to (wrt.) their localisation qualities (e.g. Intersection-over-Union - IoU). To tackle the non-differentiable nature of ranking and sorting, we reformulate the incorporation of error-driven update with backpropagation as Identity Update, which enables us to model our novel sorting error among positives. With RS Loss, we significantly simplify training: (i) Thanks to our sorting objective, the positives are prioritized by the classifier without an additional auxiliary head (e.g. for centerness, IoU, mask-IoU), (ii) due to its ranking-based nature, RS Loss is robust to class imbalance, and thus, no sampling heuristic is required, and (iii) we address the multi-task nature of visual detectors using tuning-free task-balancing coefficients. Using RS Loss, we train seven diverse visual detectors only by tuning the learning rate, and show that it consistently outperforms baselines: e.g. our RS Loss improves (i) Faster R-CNN by ~ 3 box AP and aLRP Loss (ranking-based baseline) by ~ 2 box AP on COCO dataset, (ii) Mask R-CNN with repeat factor sampling (RFS) by 3.5 mask AP (~ 7 AP for rare classes) on LVIS dataset; and also outperforms all counterparts. Code is available at: https://github.com/kemaloksuz/RankSortLoss
Transformer with self-attention has led to the revolutionizing of natural language processing field, and recently inspires the emergence of Transformer-style architecture design with competitive results in numerous computer vision tasks. Nevertheless, most of existing designs directly employ self-attention over a 2D feature map to obtain the attention matrix based on pairs of isolated queries and keys at each spatial location, but leave the rich contexts among neighbor keys under-exploited. In this work, we design a novel Transformer-style module, i.e., Contextual Transformer (CoT) block, for visual recognition. Such design fully capitalizes on the contextual information among input keys to guide the learning of dynamic attention matrix and thus strengthens the capacity of visual representation. Technically, CoT block first contextually encodes input keys via a $3\times3$ convolution, leading to a static contextual representation of inputs. We further concatenate the encoded keys with input queries to learn the dynamic multi-head attention matrix through two consecutive $1\times1$ convolutions. The learnt attention matrix is multiplied by input values to achieve the dynamic contextual representation of inputs. The fusion of the static and dynamic contextual representations are finally taken as outputs. Our CoT block is appealing in the view that it can readily replace each $3\times3$ convolution in ResNet architectures, yielding a Transformer-style backbone named as Contextual Transformer Networks (CoTNet). Through extensive experiments over a wide range of applications (e.g., image recognition, object detection and instance segmentation), we validate the superiority of CoTNet as a stronger backbone. Source code is available at \url{https://github.com/JDAI-CV/CoTNet}.
Identifying the locations and footprints of buildings is vital for many practical and scientific purposes. Such information can be particularly useful in developing regions where alternative data sources may be scarce. In this work, we describe a model training pipeline for detecting buildings across the entire continent of Africa, using 50 cm satellite imagery. Starting with the U-Net model, widely used in satellite image analysis, we study variations in architecture, loss functions, regularization, pre-training, self-training and post-processing that increase instance segmentation performance. Experiments were carried out using a dataset of 100k satellite images across Africa containing 1.75M manually labelled building instances, and further datasets for pre-training and self-training. We report novel methods for improving performance of building detection with this type of model, including the use of mixup (mAP +0.12) and self-training with soft KL loss (mAP +0.06). The resulting pipeline obtains good results even on a wide variety of challenging rural and urban contexts, and was used to create the Open Buildings dataset of 516M Africa-wide detected footprints.