The idea behind object-centric representation learning is that natural scenes can better be modeled as compositions of objects and their relations as opposed to distributed representations. This inductive bias can be injected into neural networks to potentially improve systematic generalization and performance of downstream tasks in scenes with multiple objects. In this paper, we train state-of-the-art unsupervised models on five common multi-object datasets and evaluate segmentation metrics and downstream object property prediction. In addition, we study generalization and robustness by investigating the settings where either a single object is out of distribution -- e.g., having an unseen color, texture, or shape -- or global properties of the scene are altered -- e.g., by occlusions, cropping, or increasing the number of objects. From our experimental study, we find object-centric representations to be useful for downstream tasks and generally robust to most distribution shifts affecting objects. However, when the distribution shift affects the input in a less structured manner, robustness in terms of segmentation and downstream task performance may vary significantly across models and distribution shifts.
Modern top-performing object detectors depend heavily on backbone networks, whose advances bring consistent performance gains through exploring more effective network structures. In this paper, we propose a novel and flexible backbone framework, namely CBNetV2, to construct high-performance detectors using existing open-sourced pre-trained backbones under the pre-training fine-tuning paradigm. In particular, CBNetV2 architecture groups multiple identical backbones, which are connected through composite connections. Specifically, it integrates the high- and low-level features of multiple backbone networks and gradually expands the receptive field to more efficiently perform object detection. We also propose a better training strategy with assistant supervision for CBNet-based detectors. Without additional pre-training of the composite backbone, CBNetV2 can be adapted to various backbones (CNN-based vs. Transformer-based) and head designs of most mainstream detectors (one-stage vs. two-stage, anchor-based vs. anchor-free-based). Experiments provide strong evidence that, compared with simply increasing the depth and width of the network, CBNetV2 introduces a more efficient, effective, and resource-friendly way to build high-performance backbone networks. Particularly, our Dual-Swin-L achieves 59.4% box AP and 51.6% mask AP on COCO test-dev under the single-model and single-scale testing protocol, which is significantly better than the state-of-the-art result (57.7% box AP and 50.2% mask AP) achieved by Swin-L, while the training schedule is reduced by 6$\times$. With multi-scale testing, we push the current best single model result to a new record of 60.1% box AP and 52.3% mask AP without using extra training data. Code is available at https://github.com/VDIGPKU/CBNetV2.
The speed-accuracy Pareto curve of object detection systems have advanced through a combination of better model architectures, training and inference methods. In this paper, we methodically evaluate a variety of these techniques to understand where most of the improvements in modern detection systems come from. We benchmark these improvements on the vanilla ResNet-FPN backbone with RetinaNet and RCNN detectors. The vanilla detectors are improved by 7.7% in accuracy while being 30% faster in speed. We further provide simple scaling strategies to generate family of models that form two Pareto curves, named RetinaNet-RS and Cascade RCNN-RS. These simple rescaled detectors explore the speed-accuracy trade-off between the one-stage RetinaNet detectors and two-stage RCNN detectors. Our largest Cascade RCNN-RS models achieve 52.9% AP with a ResNet152-FPN backbone and 53.6% with a SpineNet143L backbone. Finally, we show the ResNet architecture, with three minor architectural changes, outperforms EfficientNet as the backbone for object detection and instance segmentation systems.
Monocular 3D object detection is an important task in autonomous driving. It can be easily intractable where there exists ego-car pose change w.r.t. ground plane. This is common due to the slight fluctuation of road smoothness and slope. Due to the lack of insight in industrial application, existing methods on open datasets neglect the camera pose information, which inevitably results in the detector being susceptible to camera extrinsic parameters. The perturbation of objects is very popular in most autonomous driving cases for industrial products. To this end, we propose a novel method to capture camera pose to formulate the detector free from extrinsic perturbation. Specifically, the proposed framework predicts camera extrinsic parameters by detecting vanishing point and horizon change. A converter is designed to rectify perturbative features in the latent space. By doing so, our 3D detector works independent of the extrinsic parameter variations and produces accurate results in realistic cases, e.g., potholed and uneven roads, where almost all existing monocular detectors fail to handle. Experiments demonstrate our method yields the best performance compared with the other state-of-the-arts by a large margin on both KITTI 3D and nuScenes datasets.
Many man-made objects are characterised by a shape that is symmetric along one or more planar directions. Estimating the location and orientation of such symmetry planes can aid many tasks such as estimating the overall orientation of an object of interest or performing shape completion, where a partial scan of an object is reflected across the estimated symmetry plane in order to obtain a more detailed shape. Many methods processing 3D data rely on expensive 3D convolutions. In this paper we present an alternative novel encoding that instead slices the data along the height dimension and passes it sequentially to a 2D convolutional recurrent regression scheme. The method also comprises a differentiable least squares step, allowing for end-to-end accurate and fast processing of both full and partial scans of symmetric objects. We use this approach to efficiently handle 3D inputs to design a method to estimate planar reflective symmetries. We show that our approach has an accuracy comparable to state-of-the-art techniques on the task of planar reflective symmetry estimation on full synthetic objects. Additionally, we show that it can be deployed on partial scans of objects in a real-world pipeline to improve the outputs of a 3D object detector.
Robot manipulation of unknown objects in unstructured environments is a challenging problem due to the variety of shapes, materials, arrangements and lighting conditions. Even with large-scale real-world data collection, robust perception and manipulation of transparent and reflective objects across various lighting conditions remain challenging. To address these challenges we propose an approach to performing sim-to-real transfer of robotic perception. The underlying model, SimNet, is trained as a single multi-headed neural network using simulated stereo data as input and simulated object segmentation masks, 3D oriented bounding boxes (OBBs), object keypoints, and disparity as output. A key component of SimNet is the incorporation of a learned stereo sub-network that predicts disparity. SimNet is evaluated on 2D car detection, unknown object detection, and deformable object keypoint detection and significantly outperforms a baseline that uses a structured light RGB-D sensor. By inferring grasp positions using the OBB and keypoint predictions, SimNet can be used to perform end-to-end manipulation of unknown objects in both easy and hard scenarios using our fleet of Toyota HSR robots in four home environments. In unknown object grasping experiments, the predictions from the baseline RGB-D network and SimNet enable successful grasps of most of the easy objects. However, the RGB-D baseline only grasps 35% of the hard (e.g., transparent) objects, while SimNet grasps 95%, suggesting that SimNet can enable robust manipulation of unknown objects, including transparent objects, in unknown environments.
In applications such as object tracking, time-series data inevitably carry missing observations. Following the success of deep learning-based models for various sequence learning tasks, these models increasingly replace classic approaches in object tracking applications for inferring the objects' motion states. While traditional tracking approaches can deal with missing observations, most of their deep counterparts are, by default, not suited for this. Towards this end, this paper introduces a transformer-based approach for handling missing observations in variable input length trajectory data. The model is formed indirectly by successively increasing the complexity of the demanded inference tasks. Starting from reproducing noise-free trajectories, the model then learns to infer trajectories from noisy inputs. By providing missing tokens, binary-encoded missing events, the model learns to in-attend to missing data and infers a complete trajectory conditioned on the remaining inputs. In the case of a sequence of successive missing events, the model then acts as a pure prediction model. The abilities of the approach are demonstrated on synthetic data and real-world data reflecting prototypical object tracking scenarios.
To reduce annotation labor associated with object detection, an increasing number of studies focus on transferring the learned knowledge from a labeled source domain to another unlabeled target domain. However, existing methods assume that the labeled data are sampled from a single source domain, which ignores a more generalized scenario, where labeled data are from multiple source domains. For the more challenging task, we propose a unified Faster R-CNN based framework, termed Divide-and-Merge Spindle Network (DMSN), which can simultaneously enhance domain invariance and preserve discriminative power. Specifically, the framework contains multiple source subnets and a pseudo target subnet. First, we propose a hierarchical feature alignment strategy to conduct strong and weak alignments for low- and high-level features, respectively, considering their different effects for object detection. Second, we develop a novel pseudo subnet learning algorithm to approximate optimal parameters of pseudo target subset by weighted combination of parameters in different source subnets. Finally, a consistency regularization for region proposal network is proposed to facilitate each subnet to learn more abstract invariances. Extensive experiments on different adaptation scenarios demonstrate the effectiveness of the proposed model.
Point clouds have recently gained interest, especially for real-time applications and for 3D-scanned material, such as is used in autonomous driving, architecture, and engineering, to model real estate for renovation or display. Point clouds are associated with geometry information and attributes such as color. Be the color unique or direction-dependent (in the case of plenoptic point clouds), it reflects the colors observed by cameras displaced around the object. Hence, not only are the viewing references assumed, but the illumination spectrum and illumination geometry is also implicit. We propose a model-centric description of the 3D object, that is independent of the illumination and of the position of the cameras. We want to be able to describe the objects themselves such that, at a later stage, the rendering of the model may decide where to place illumination, from which it may calculate the image viewed by a given camera. We want to be able to describe transparent or translucid objects, mirrors, fishbowls, fog and smoke. Volumetric clouds may allow us to describe the air, however ``empty'', and introduce air particles, in a manner independent of the viewer position. For that, we rely on some eletromagnetic properties to arrive at seven attributes per voxel that would describe the material and its color or transparency. Three attributes are for the transmissivity of each color, three are for the attenuation of each color, and another attribute is for diffuseness. These attributes give information about the object to the renderer, with whom lies the decision on how to render and depict each object.
TANet is one of state-of-the-art 3D object detection method on KITTI and JRDB benchmark, the network contains a Triple Attention module and Coarse-to-Fine Regression module to improve the robustness and accuracy of 3D Detection. However, since the original input data (point clouds) contains a lot of noise during collecting the data, which will further affect the training of the model. For example, the object is far from the robot, the sensor is difficult to obtain enough pointcloud. If the objects only contains few point clouds, and the samples are fed into model with the normal samples together during training, the detector will be difficult to distinguish the individual with few pointcloud belong to object or background. In this paper, we propose TANet++ to improve the performance on 3D Detection, which adopt a novel training strategy on training the TANet. In order to reduce the negative impact by the weak samples, the training strategy previously filtered the training data, and then the TANet++ is trained by the rest of data. The experimental results shows that AP score of TANet++ is 8.98 higher than TANet on JRDB benchmark.
TextVQA requires models to read and reason about text in images to answer questions about them. Specifically, models need to incorporate a new modality of text present in the images and reason over it to answer TextVQA questions. In this challenge, we use generative model T5 for TextVQA task. Based on pre-trained checkpoint T5-3B from HuggingFace repository, two other pre-training tasks including masked language modeling(MLM) and relative position prediction(RPP) are designed to better align object feature and scene text. In the stage of pre-training, encoder is dedicate to handle the fusion among multiple modalities: question text, object text labels, scene text labels, object visual features, scene visual features. After that decoder generates the text sequence step-by-step, cross entropy loss is required by default. We use a large-scale scene text dataset in pre-training and then fine-tune the T5-3B with the TextVQA dataset only.
Object recognition from live video streams comes with numerous challenges such as the variation in illumination conditions and poses. Convolutional neural networks (CNNs) have been widely used to perform intelligent visual object recognition. Yet, CNNs still suffer from severe accuracy degradation, particularly on illumination-variant datasets. To address this problem, we propose a new CNN method based on orientation fusion for visual object recognition. The proposed cloud-based video analytics system pioneers the use of bi-dimensional empirical mode decomposition to split a video frame into intrinsic mode functions (IMFs). We further propose these IMFs to endure Reisz transform to produce monogenic object components, which are in turn used for the training of CNNs. Past works have demonstrated how the object orientation component may be used to pursue accuracy levels as high as 93\%. Herein we demonstrate how a feature-fusion strategy of the orientation components leads to further improving visual recognition accuracy to 97\%. We also assess the scalability of our method, looking at both the number and the size of the video streams under scrutiny. We carry out extensive experimentation on the publicly available Yale dataset, including also a self generated video datasets, finding significant improvements (both in accuracy and scale), in comparison to AlexNet, LeNet and SE-ResNeXt, which are the three most commonly used deep learning models for visual object recognition and classification.
Contrary to the vast literature in modeling, perceiving, and understanding agent-object (e.g., human-object, hand-object, robot-object) interaction in computer vision and robotics, very few past works have studied the task of object-object interaction, which also plays an important role in robotic manipulation and planning tasks. There is a rich space of object-object interaction scenarios in our daily life, such as placing an object on a messy tabletop, fitting an object inside a drawer, pushing an object using a tool, etc. In this paper, we propose a unified affordance learning framework to learn object-object interaction for various tasks. By constructing four object-object interaction task environments using physical simulation (SAPIEN) and thousands of ShapeNet models with rich geometric diversity, we are able to conduct large-scale object-object affordance learning without the need for human annotations or demonstrations. At the core of technical contribution, we propose an object-kernel point convolution network to reason about detailed interaction between two objects. Experiments on large-scale synthetic data and real-world data prove the effectiveness of the proposed approach. Please refer to the project webpage for code, data, video, and more materials: https://cs.stanford.edu/~kaichun/o2oafford
In this paper, we introduce a new technique that combines two popular methods to estimate uncertainty in object detection. Quantifying uncertainty is critical in real-world robotic applications. Traditional detection models can be ambiguous even when they provide a high-probability output. Robot actions based on high-confidence, yet unreliable predictions, may result in serious repercussions. Our framework employs deep ensembles and Monte Carlo dropout for approximating predictive uncertainty, and it improves upon the uncertainty estimation quality of the baseline method. The proposed approach is evaluated on publicly available synthetic image datasets captured from sequences of video.
We present an object detection based approach to localize handwritten regions from documents, which initially aims to enhance the anonymization during the data transmission. The concatenated fusion of original and preprocessed images containing both printed texts and handwritten notes or signatures are fed into the convolutional neural network, where the bounding boxes are learned to detect the handwriting. Afterwards, the handwritten regions can be processed (e.g. replaced with redacted signatures) to conceal the personally identifiable information (PII). This processing pipeline based on the deep learning network Cascade R-CNN works at 10 fps on a GPU during the inference, which ensures the enhanced anonymization with minimal computational overheads. Furthermore, the impressive generalizability has been empirically showcased: the trained model based on the English-dominant dataset works well on the fictitious unseen invoices, even in Chinese. The proposed approach is also expected to facilitate other tasks such as handwriting recognition and signature verification.
Object detection plays an important role in self-driving cars for security development. However, mobile systems on self-driving cars with limited computation resources lead to difficulties for object detection. To facilitate this, we propose a compiler-aware neural pruning search framework to achieve high-speed inference on autonomous vehicles for 2D and 3D object detection. The framework automatically searches the pruning scheme and rate for each layer to find a best-suited pruning for optimizing detection accuracy and speed performance under compiler optimization. Our experiments demonstrate that for the first time, the proposed method achieves (close-to) real-time, 55ms and 99ms inference times for YOLOv4 based 2D object detection and PointPillars based 3D detection, respectively, on an off-the-shelf mobile phone with minor (or no) accuracy loss.
When producing a model to object detection in a specific context, the first obstacle is to have a dataset labeling the desired classes. In RoboCup, some leagues already have more than one dataset to train and evaluate a model. However, in the Small Size League (SSL), there is not such dataset available yet. This paper presents an open-source dataset to be used as a benchmark for real-time object detection in SSL. This work also presented a pipeline to train, deploy, and evaluate Convolutional Neural Networks (CNNs) models in a low-power embedded system. This pipeline was used to evaluate the proposed dataset with state-of-art optimized models. In this dataset, the MobileNet SSD v1 achieves 44.88% AP (68.81% AP50) at 94 Frames Per Second (FPS) while running on an SSL robot.
As a common image editing operation, image composition (object insertion) aims to combine the foreground from one image and another background image, resulting in a composite image. However, there are many issues that could make the composite images unrealistic. These issues can be summarized as the inconsistency between foreground and background, which includes appearance inconsistency (e.g., incompatible illumination), geometry inconsistency (e.g., unreasonable size), and semantic inconsistency (e.g., mismatched semantic context). Image composition task could be decomposed into multiple sub-tasks, in which each sub-task targets at one or more issues. Specifically, object placement aims to find reasonable scale, location, and shape for the foreground. Image blending aims to address the unnatural boundary between foreground and background. Image harmonization aims to adjust the illumination statistics of foreground. Shadow (resp., reflection) generation aims to generate plausible shadow (resp., reflection) for the foreground. These sub-tasks can be executed sequentially or parallelly to acquire realistic composite images. To the best of our knowledge, there is no previous survey on image composition (object insertion). In this paper, we conduct comprehensive survey over the sub-tasks and combinatorial task of image composition (object insertion). For each one, we summarize the existing methods, available datasets, and common evaluation metrics. We have also contributed the first image composition toolbox libcom, which assembles 10+ image composition related functions.
Image-based 3D object modeling refers to the process of converting raw optical images to 3D digital representations of the objects. Very often, such models are desired to be dimensionally true, semantically labeled with photorealistic appearance (reality-based modeling). Laser scanning was deemed as the standard (and direct) way to obtaining highly accurate 3D measurements of objects, while one would have to abide the high acquisition cost and its unavailability on some of the platforms. Nowadays the image-based methods backboned by the recently developed advanced dense image matching algorithms and geo-referencing paradigms, are becoming the dominant approaches, due to its high flexibility, availability and low cost. The largely automated geometric processing of images in a 3D object reconstruction workflow, from ordered/unordered raw imagery to textured meshes, is becoming a critical part of the reality-based 3D modeling. This article summarizes the overall geometric processing workflow, with focuses on introducing the state-of-the-art methods of three major components of geometric processing: 1) geo-referencing; 2) Image dense matching 3) texture mapping. Finally, we will draw conclusions and share our outlooks of the topics discussed in this article.
In this study, we propose a novel approach to predict the distances of the detected objects in an observed scene. The proposed approach modifies the recently proposed Convolutional Support Estimator Networks (CSENs). CSENs are designed to compute a direct mapping for the Support Estimation (SE) task in a representation-based classification problem. We further propose and demonstrate that representation-based methods (sparse or collaborative representation) can be used in well-designed regression problems. To the best of our knowledge, this is the first representation-based method proposed for performing a regression task by utilizing the modified CSENs; and hence, we name this novel approach as Representation-based Regression (RbR). The initial version of CSENs has a proxy mapping stage (i.e., a coarse estimation for the support set) that is required for the input. In this study, we improve the CSEN model by proposing Compressive Learning CSEN (CL-CSEN) that has the ability to jointly optimize the so-called proxy mapping stage along with convolutional layers. The experimental evaluations using the KITTI 3D Object Detection distance estimation dataset show that the proposed method can achieve a significantly improved distance estimation performance over all competing methods. Finally, the software implementations of the methods are publicly shared at https://github.com/meteahishali/CSENDistance.