Deep neural networks for medical image reconstruction are traditionally trained using high-quality ground-truth images as training targets. Recent work on Noise2Noise (N2N) has shown the potential of using multiple noisy measurements of the same object as an alternative to having a ground-truth. However, existing N2N-based methods are not suitable for learning from the measurements of an object undergoing nonrigid deformation. This paper addresses this issue by proposing the deformation-compensated learning (DeCoLearn) method for training deep reconstruction networks by compensating for object deformations. A key component of DeCoLearn is a deep registration module, which is jointly trained with the deep reconstruction network without any ground-truth supervision. We validate DeCoLearn on both simulated and experimentally collected magnetic resonance imaging (MRI) data and show that it significantly improves imaging quality.
Object detection in autonomous cars is commonly based on camera images and Lidar inputs, which are often used to train prediction models such as deep artificial neural networks for decision making for object recognition, adjusting speed, etc. A mistake in such decision making can be damaging; thus, it is vital to measure the reliability of decisions made by such prediction models via uncertainty measurement. Uncertainty, in deep learning models, is often measured for classification problems. However, deep learning models in autonomous driving are often multi-output regression models. Hence, we propose a novel method called PURE (Prediction sURface uncErtainty) for measuring prediction uncertainty of such regression models. We formulate the object recognition problem as a regression model with more than one outputs for finding object locations in a 2-dimensional camera view. For evaluation, we modified three widely-applied object recognition models (i.e., YoLo, SSD300 and SSD512) and used the KITTI, Stanford Cars, Berkeley DeepDrive, and NEXET datasets. Results showed the statistically significant negative correlation between prediction surface uncertainty and prediction accuracy suggesting that uncertainty significantly impacts the decisions made by autonomous driving.
The vast majority of uncertainty quantification methods for deep object detectors such as variational inference are based on the network output. Here, we study gradient-based epistemic uncertainty metrics for deep object detectors to obtain reliable confidence estimates. We show that they contain predictive information and that they capture information orthogonal to that of common, output-based uncertainty estimation methods like Monte-Carlo dropout and deep ensembles. To this end, we use meta classification and meta regression to produce confidence estimates using gradient metrics and other baselines for uncertainty quantification which are in principle applicable to any object detection architecture. Specifically, we employ false positive detection and prediction of localization quality to investigate uncertainty content of our metrics and compute the calibration errors of meta classifiers. Moreover, we use them as a post-processing filter mechanism to the object detection pipeline and compare object detection performance. Our results show that gradient-based uncertainty is itself on par with output-based methods across different detectors and datasets. More significantly, combined meta classifiers based on gradient and output-based metrics outperform the standalone models. Based on this result, we conclude that gradient uncertainty adds orthogonal information to output-based methods. This suggests that variational inference may be supplemented by gradient-based uncertainty to obtain improved confidence measures, contributing to down-stream applications of deep object detectors and improving their probabilistic reliability.
Multi-object tracking is a critical component in autonomous navigation, as it provides valuable information for decision-making. Many researchers tackled the 3D multi-object tracking task by filtering out the frame-by-frame 3D detections; however, their focus was mainly on finding useful features or proper matching metrics. Our work focuses on a neglected part of the tracking system: score refinement and tracklet termination. We show that manipulating the scores depending on time consistency while terminating the tracklets depending on the tracklet score improves tracking results. We do this by increasing the matched tracklets' score with score update functions and decreasing the unmatched tracklets' score. Compared to count-based methods, our method consistently produces better AMOTA and MOTA scores when utilizing various detectors and filtering algorithms on different datasets. The improvements in AMOTA score went up to 1.83 and 2.96 in MOTA. We also used our method as a late-fusion ensembling method, and it performed better than voting-based ensemble methods by a solid margin. It achieved an AMOTA score of 67.6 on nuScenes test evaluation, which is comparable to other state-of-the-art trackers. Code is publicly available at: \url{https://github.com/cogsys-tuebingen/CBMOT}.
Point clouds and RGB images are naturally complementary modalities for 3D visual understanding - the former provides sparse but accurate locations of points on objects, while the latter contains dense color and texture information. Despite this potential for close sensor fusion, many methods train two models in isolation and use simple feature concatenation to represent 3D sensor data. This separated training scheme results in potentially sub-optimal performance and prevents 3D tasks from being used to benefit 2D tasks that are often useful on their own. To provide a more integrated approach, we propose a novel Multi-Modality Task Cascade network (MTC-RCNN) that leverages 3D box proposals to improve 2D segmentation predictions, which are then used to further refine the 3D boxes. We show that including a 2D network between two stages of 3D modules significantly improves both 2D and 3D task performance. Moreover, to prevent the 3D module from over-relying on the overfitted 2D predictions, we propose a dual-head 2D segmentation training and inference scheme, allowing the 2nd 3D module to learn to interpret imperfect 2D segmentation predictions. Evaluating our model on the challenging SUN RGB-D dataset, we improve upon state-of-the-art results of both single modality and fusion networks by a large margin ($\textbf{+3.8}$ [email omitted]). Code will be released $\href{https://github.com/Divadi/MTC_RCNN}{\text{here.}}$
Deep learning-based dense object detectors have achieved great success in the past few years and have been applied to numerous multimedia applications such as video understanding. However, the current training pipeline for dense detectors is compromised to lots of conjunctions that may not hold. In this paper, we investigate three such important conjunctions: 1) only samples assigned as positive in classification head are used to train the regression head; 2) classification and regression share the same input feature and computational fields defined by the parallel head architecture; and 3) samples distributed in different feature pyramid layers are treated equally when computing the loss. We first carry out a series of pilot experiments to show disentangling such conjunctions can lead to persistent performance improvement. Then, based on these findings, we propose Disentangled Dense Object Detector (DDOD), in which simple and effective disentanglement mechanisms are designed and integrated into the current state-of-the-art dense object detectors. Extensive experiments on MS COCO benchmark show that our approach can lead to 2.0 mAP, 2.4 mAP and 2.2 mAP absolute improvements on RetinaNet, FCOS, and ATSS baselines with negligible extra overhead. Notably, our best model reaches 55.0 mAP on the COCO test-dev set and 93.5 AP on the hard subset of WIDER FACE, achieving new state-of-the-art performance on these two competitive benchmarks. Code is available at https://github.com/zehuichen123/DDOD.
Image composition aims to generate realistic composite image by inserting an object from one image into another background image, where the placement (e.g., location, size, occlusion) of inserted object may be unreasonable, which would significantly degrade the quality of the composite image. Although some works attempted to learn object placement to create realistic composite images, they did not focus on assessing the plausibility of object placement. In this paper, we focus on object placement assessment task, which verifies whether a composite image is plausible in terms of the object placement. To accomplish this task, we construct the first Object Placement Assessment (OPA) dataset consisting of composite images and their rationality labels. We also propose a simple yet effective baseline for this task. Dataset is available at https://github.com/bcmi/Object-Placement-Assessment-Dataset-OPA.
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.
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 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.
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.
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.
Automotive traffic scenes are complex due to the variety of possible scenarios, objects, and weather conditions that need to be handled. In contrast to more constrained environments, such as automated underground trains, automotive perception systems cannot be tailored to a narrow field of specific tasks but must handle an ever-changing environment with unforeseen events. As currently no single sensor is able to reliably perceive all relevant activity in the surroundings, sensor data fusion is applied to perceive as much information as possible. Data fusion of different sensors and sensor modalities on a low abstraction level enables the compensation of sensor weaknesses and misdetections among the sensors before the information-rich sensor data are compressed and thereby information is lost after a sensor-individual object detection. This paper develops a low-level sensor fusion network for 3D object detection, which fuses lidar, camera, and radar data. The fusion network is trained and evaluated on the nuScenes data set. On the test set, fusion of radar data increases the resulting AP (Average Precision) detection score by about 5.1% in comparison to the baseline lidar network. The radar sensor fusion proves especially beneficial in inclement conditions such as rain and night scenes. Fusing additional camera data contributes positively only in conjunction with the radar fusion, which shows that interdependencies of the sensors are important for the detection result. Additionally, the paper proposes a novel loss to handle the discontinuity of a simple yaw representation for object detection. Our updated loss increases the detection and orientation estimation performance for all sensor input configurations. The code for this research has been made available on GitHub.
As a core problem in computer vision, the performance of object detection has improved drastically in the past few years. Despite their impressive performance, object detectors suffer from a lack of interpretability. Visualization techniques have been developed and widely applied to introspect the decisions made by other kinds of deep learning models; however, visualizing object detectors has been underexplored. In this paper, we propose using inversion as a primary tool to understand modern object detectors and develop an optimization-based approach to layout inversion, allowing us to generate synthetic images recognized by trained detectors as containing a desired configuration of objects. We reveal intriguing properties of detectors by applying our layout inversion technique to a variety of modern object detectors, and further investigate them via validation experiments: they rely on qualitatively different features for classification and regression; they learn canonical motifs of commonly co-occurring objects; they use diff erent visual cues to recognize objects of varying sizes. We hope our insights can help practitioners improve object detectors.
Highly constrained manipulation tasks continue to be challenging for autonomous robots as they require high levels of precision, typically less than 1mm, which is often incompatible with what can be achieved by traditional perception systems. This paper demonstrates that the combination of state-of-the-art object tracking with passively adaptive mechanical hardware can be leveraged to complete precision manipulation tasks with tight, industrially-relevant tolerances (0.25mm). The proposed control method closes the loop through vision by tracking the relative 6D pose of objects in the relevant workspace. It adjusts the control reference of both the compliant manipulator and the hand to complete object insertion tasks via within-hand manipulation. Contrary to previous efforts for insertion, our method does not require expensive force sensors, precision manipulators, or time-consuming, online learning, which is data hungry. Instead, this effort leverages mechanical compliance and utilizes an object agnostic manipulation model of the hand learned offline, off-the-shelf motion planning, and an RGBD-based object tracker trained solely with synthetic data. These features allow the proposed system to easily generalize and transfer to new tasks and environments. This paper describes in detail the system components and showcases its efficacy with extensive experiments involving tight tolerance peg-in-hole insertion tasks of various geometries as well as open-world constrained placement tasks.
Semantic image synthesis is a process for generating photorealistic images from a single semantic mask. To enrich the diversity of multimodal image synthesis, previous methods have controlled the global appearance of an output image by learning a single latent space. However, a single latent code is often insufficient for capturing various object styles because object appearance depends on multiple factors. To handle individual factors that determine object styles, we propose a class- and layer-wise extension to the variational autoencoder (VAE) framework that allows flexible control over each object class at the local to global levels by learning multiple latent spaces. Furthermore, we demonstrate that our method generates images that are both plausible and more diverse compared to state-of-the-art methods via extensive experiments with real and synthetic datasets inthree different domains. We also show that our method enables a wide range of applications in image synthesis and editing tasks.
Multiple object tracking and segmentation requires detecting, tracking, and segmenting objects belonging to a set of given classes. Most approaches only exploit the temporal dimension to address the association problem, while relying on single frame predictions for the segmentation mask itself. We propose Prototypical Cross-Attention Network (PCAN), capable of leveraging rich spatio-temporal information for online multiple object tracking and segmentation. PCAN first distills a space-time memory into a set of prototypes and then employs cross-attention to retrieve rich information from the past frames. To segment each object, PCAN adopts a prototypical appearance module to learn a set of contrastive foreground and background prototypes, which are then propagated over time. Extensive experiments demonstrate that PCAN outperforms current video instance tracking and segmentation competition winners on both Youtube-VIS and BDD100K datasets, and shows efficacy to both one-stage and two-stage segmentation frameworks. Code and video resources are available at http://vis.xyz/pub/pcan.
Contrastive self-supervised learning has largely narrowed the gap to supervised pre-training on ImageNet. However, its success highly relies on the object-centric priors of ImageNet, i.e., different augmented views of the same image correspond to the same object. Such a heavily curated constraint becomes immediately infeasible when pre-trained on more complex scene images with many objects. To overcome this limitation, we introduce Object-level Representation Learning (ORL), a new self-supervised learning framework towards scene images. Our key insight is to leverage image-level self-supervised pre-training as the prior to discover object-level semantic correspondence, thus realizing object-level representation learning from scene images. Extensive experiments on COCO show that ORL significantly improves the performance of self-supervised learning on scene images, even surpassing supervised ImageNet pre-training on several downstream tasks. Furthermore, ORL improves the downstream performance when more unlabeled scene images are available, demonstrating its great potential of harnessing unlabeled data in the wild. We hope our approach can motivate future research on more general-purpose unsupervised representation learning from scene data.