Moving object detection has been a central topic of discussion in computer vision for its wide range of applications like in self-driving cars, video surveillance, security, and enforcement. Neuromorphic Vision Sensors (NVS) are bio-inspired sensors that mimic the working of the human eye. Unlike conventional frame-based cameras, these sensors capture a stream of asynchronous 'events' that pose multiple advantages over the former, like high dynamic range, low latency, low power consumption, and reduced motion blur. However, these advantages come at a high cost, as the event camera data typically contains more noise and has low resolution. Moreover, as event-based cameras can only capture the relative changes in brightness of a scene, event data do not contain usual visual information (like texture and color) as available in video data from normal cameras. So, moving object detection in event-based cameras becomes an extremely challenging task. In this paper, we present an unsupervised Graph Spectral Clustering technique for Moving Object Detection in Event-based data (GSCEventMOD). We additionally show how the optimum number of moving objects can be automatically determined. Experimental comparisons on publicly available datasets show that the proposed GSCEventMOD algorithm outperforms a number of state-of-the-art techniques by a maximum margin of 30%.
Localizing objects in image collections without supervision can help to avoid expensive annotation campaigns. We propose a simple approach to this problem, that leverages the activation features of a vision transformer pre-trained in a self-supervised manner. Our method, LOST, does not require any external object proposal nor any exploration of the image collection; it operates on a single image. Yet, we outperform state-of-the-art object discovery methods by up to 8 CorLoc points on PASCAL VOC 2012. We also show that training a class-agnostic detector on the discovered objects boosts results by another 7 points. Moreover, we show promising results on the unsupervised object discovery task. The code to reproduce our results can be found at https://github.com/valeoai/LOST.
Recent advances in object segmentation have demonstrated that deep neural networks excel at object segmentation for specific classes in color and depth images. However, their performance is dictated by the number of classes and objects used for training, thereby hindering generalization to never seen objects or zero-shot samples. To exacerbate the problem further, object segmentation using image frames rely on recognition and pattern matching cues. Instead, we utilize the 'active' nature of a robot and their ability to 'interact' with the environment to induce additional geometric constraints for segmenting zero-shot samples. In this paper, we present the first framework to segment unknown objects in a cluttered scene by repeatedly 'nudging' at the objects and moving them to obtain additional motion cues at every step using only a monochrome monocular camera. We call our framework NudgeSeg. These motion cues are used to refine the segmentation masks. We successfully test our approach to segment novel objects in various cluttered scenes and provide an extensive study with image and motion segmentation methods. We show an impressive average detection rate of over 86% on zero-shot objects.
Rotation augmentations generally improve a model's invariance/equivariance to rotation - except in object detection. In object detection the shape is not known, therefore rotation creates a label ambiguity. We show that the de-facto method for bounding box label rotation, the Largest Box Method, creates very large labels, leading to poor performance and in many cases worse performance than using no rotation at all. We propose a new method of rotation augmentation that can be implemented in a few lines of code. First, we create a differentiable approximation of label accuracy and show that axis-aligning the bounding box around an ellipse is optimal. We then introduce Rotation Uncertainty (RU) Loss, allowing the model to adapt to the uncertainty of the labels. On five different datasets (including COCO, PascalVOC, and Transparent Object Bin Picking), this approach improves the rotational invariance of both one-stage and two-stage architectures when measured with AP, AP50, and AP75. The code is available at https://github.com/akasha-imaging/ICCV2021.
Event cameras are novel sensors that perceive the per-pixel intensity changes and output asynchronous event streams, showing lots of advantages over traditional cameras, such as high dynamic range (HDR) and no motion blur. It has been shown that events alone can be used for object tracking by motion compensation or prediction. However, existing methods assume that the target always moves and is the stand-alone object. Moreover, they fail to track the stopped non-independent moving objects on fixed scenes. In this paper, we propose a novel event-based object tracking framework, called SiamEvent, using Siamese networks via edge-aware similarity learning. Importantly, to find the part having the most similar edge structure of target, we propose to correlate the embedded events at two timestamps to compute the target edge similarity. The Siamese network enables tracking arbitrary target edge by finding the part with the highest similarity score. This extends the possibility of event-based object tracking applied not only for the independent stand-alone moving objects, but also for various settings of the camera and scenes. In addition, target edge initialization and edge detector are also proposed to prevent SiamEvent from the drifting problem. Lastly, we built an open dataset including various synthetic and real scenes to train and evaluate SiamEvent. Extensive experiments demonstrate that SiamEvent achieves up to 15% tracking performance enhancement than the baselines on the real-world scenes and more robust tracking performance in the challenging HDR and motion blur conditions.
We present a fast, fully parameterizable GPU implementation of Convolutional Neural Network variants. Our feature extractors are neither carefully designed nor pre-wired, but rather learned in a supervised way. Our deep hierarchical architectures achieve the best published results on benchmarks for object classification (NORB, CIFAR10) and handwritten digit recognition (MNIST), with error rates of 2.53%, 19.51%, 0.35%, respectively. Deep nets trained by simple back-propagation perform better than more shallow ones. Learning is surprisingly rapid. NORB is completely trained within five epochs. Test error rates on MNIST drop to 2.42%, 0.97% and 0.48% after 1, 3 and 17 epochs, respectively.
Effectively structuring deep knowledge plays a pivotal role in transfer from teacher to student, especially in semantic vision tasks. In this paper, we present a simple knowledge structure to exploit and encode information inside the detection system to facilitate detector knowledge distillation. Specifically, aiming at solving the feature imbalance problem while further excavating the missing relation inside semantic instances, we design a graph whose nodes correspond to instance proposal-level features and edges represent the relation between nodes. To further refine this graph, we design an adaptive background loss weight to reduce node noise and background samples mining to prune trivial edges. We transfer the entire graph as encoded knowledge representation from teacher to student, capturing local and global information simultaneously. We achieve new state-of-the-art results on the challenging COCO object detection task with diverse student-teacher pairs on both one- and two-stage detectors. We also experiment with instance segmentation to demonstrate robustness of our method. It is notable that distilled Faster R-CNN with ResNet18-FPN and ResNet50-FPN yields 38.68 and 41.82 Box AP respectively on the COCO benchmark, Faster R-CNN with ResNet101-FPN significantly achieves 43.38 AP, which outperforms ResNet152-FPN teacher about 0.7 AP. Code: https://github.com/dvlab-research/Dsig.
Detecting and tracking vehicles in urban scenes is a crucial step in many traffic-related applications as it helps to improve road user safety among other benefits. Various challenges remain unresolved in multi-object tracking (MOT) including target information description, long-term occlusions and fast motion. We propose a multi-vehicle detection and tracking system following the tracking-by-detection paradigm that tackles the previously mentioned challenges. Our MOT method extends an Intersection-over-Union (IOU)-based tracker with vehicle re-identification features. This allows us to utilize appearance information to better match objects after long occlusion phases and/or when object location is significantly shifted due to fast motion. We outperform our baseline MOT method on the UA-DETRAC benchmark while maintaining a total processing speed suitable for online use cases.
Recent approaches have shown that training deep neural networks directly on large-scale image-text pair collections enables zero-shot transfer on various recognition tasks. One central issue is how this can be generalized to object detection, which involves the non-semantic task of localization as well as semantic task of classification. To solve this problem, we introduce a vision-language embedding alignment method that transfers the generalization capabilities of a pretrained model such as CLIP to an object detector like YOLOv5. We formulate a loss function that allows us to align the image and text embeddings from the pretrained model CLIP with the modified semantic prediction head from the detector. With this method, we are able to train an object detector that achieves state-of-the-art performance on the COCO, ILSVRC, and Visual Genome zero-shot detection benchmarks. During inference, our model can be adapted to detect any number of object classes without additional training. We also find that standard object detection scaling can transfer well to our method and find consistent improvements across various scales of YOLOv5 models and the YOLOv3 model. Lastly, we develop a self-labeling method that provides a significant score improvement without needing extra images nor labels.
In this project we trained a neural network to perform specific interactions between a robot and objects in the environment, through imitation learning. In particular, we tackle the task of moving the robot to a fixed pose with respect to a certain object and later extend our method to handle any arbitrary pose around this object. We show that a simple network, with relatively little training data, is able to reach very good performance on the fixed-pose task, while more work is needed to perform the arbitrary-pose task satisfactorily. We also explore the effect of ambiguities in the sensor readings, in particular caused by symmetries in the target object, on the behaviour of the learned controller.
We classify the discontinuity of loss in both five-param and eight-param rotated object detection methods as rotation sensitivity error (RSE) which will result in performance degeneration. We introduce a novel modulated rotation loss to alleviate the problem and propose a rotation sensitivity detection network (RSDet) which is consists of an eight-param single-stage rotated object detector and the modulated rotation loss. Our proposed RSDet has several advantages: 1) it reformulates the rotated object detection problem as predicting the corners of objects while most previous methods employ a five-para-based regression method with different measurement units. 2) modulated rotation loss achieves consistent improvement on both five-param and eight-param rotated object detection methods by solving the discontinuity of loss. To further improve the accuracy of our method on objects smaller than 10 pixels, we introduce a novel RSDet++ which is consists of a point-based anchor-free rotated object detector and a modulated rotation loss. Extensive experiments demonstrate the effectiveness of both RSDet and RSDet++, which achieve competitive results on rotated object detection in the challenging benchmarks DOTA1.0, DOTA1.5, and DOTA2.0. We hope the proposed method can provide a new perspective for designing algorithms to solve rotated object detection and pay more attention to tiny objects. The codes and models are available at: https://github.com/yangxue0827/RotationDetection.
This work investigates the problem of sketch-guided object localization (SGOL), where human sketches are used as queries to conduct the object localization in natural images. In this cross-modal setting, we first contribute with a tough-to-beat baseline that without any specific SGOL training is able to outperform the previous works on a fixed set of classes. The baseline is useful to analyze the performance of SGOL approaches based on available simple yet powerful methods. We advance prior arts by proposing a sketch-conditioned DETR (DEtection TRansformer) architecture which avoids a hard classification and alleviates the domain gap between sketches and images to localize object instances. Although the main goal of SGOL is focused on object detection, we explored its natural extension to sketch-guided instance segmentation. This novel task allows to move towards identifying the objects at pixel level, which is of key importance in several applications. We experimentally demonstrate that our model and its variants significantly advance over previous state-of-the-art results. All training and testing code of our model will be released to facilitate future research{{https://github.com/priba/sgol_wild}}.
Humans learn to recognize and manipulate new objects in lifelong settings without forgetting the previously gained knowledge under non-stationary and sequential conditions. In autonomous systems, the agents also need to mitigate similar behavior to continually learn the new object categories and adapt to new environments. In most conventional deep neural networks, this is not possible due to the problem of catastrophic forgetting, where the newly gained knowledge overwrites existing representations. Furthermore, most state-of-the-art models excel either in recognizing the objects or in grasp prediction, while both tasks use visual input. The combined architecture to tackle both tasks is very limited. In this paper, we proposed a hybrid model architecture consists of a dynamically growing dual-memory recurrent neural network (GDM) and an autoencoder to tackle object recognition and grasping simultaneously. The autoencoder network is responsible to extract a compact representation for a given object, which serves as input for the GDM learning, and is responsible to predict pixel-wise antipodal grasp configurations. The GDM part is designed to recognize the object in both instances and categories levels. We address the problem of catastrophic forgetting using the intrinsic memory replay, where the episodic memory periodically replays the neural activation trajectories in the absence of external sensory information. To extensively evaluate the proposed model in a lifelong setting, we generate a synthetic dataset due to lack of sequential 3D objects dataset. Experiment results demonstrated that the proposed model can learn both object representation and grasping simultaneously in continual learning scenarios.
Intelligent navigation among social crowds is an essential aspect of mobile robotics for applications such as delivery, health care, or assistance. Deep Reinforcement Learning emerged as an alternative planning method to conservative approaches and promises more efficient and flexible navigation. However, in highly dynamic environments employing different kinds of obstacle classes, safe navigation still presents a grand challenge. In this paper, we propose a semantic Deep-reinforcement-learning-based navigation approach that teaches object-specific safety rules by considering high-level obstacle information. In particular, the agent learns object-specific behavior by contemplating the specific danger zones to enhance safety around vulnerable object classes. We tested the approach against a benchmark obstacle avoidance approach and found an increase in safety. Furthermore, we demonstrate that the agent could learn to navigate more safely by keeping an individual safety distance dependent on the semantic information.
In this paper, we propose the first self-distillation framework for general object detection, termed LGD (Label-Guided self-Distillation). Previous studies rely on a strong pretrained teacher to provide instructive knowledge that could be unavailable in real-world scenarios. Instead, we generate an instructive knowledge based only on student representations and regular labels. Our framework includes sparse label-appearance encoder, inter-object relation adapter and intra-object knowledge mapper that jointly form an implicit teacher at training phase, dynamically dependent on labels and evolving student representations. They are trained end-to-end with detector and discarded in inference. Experimentally, LGD obtains decent results on various detectors, datasets, and extensive tasks like instance segmentation. For example in MS-COCO dataset, LGD improves RetinaNet with ResNet-50 under 2x single-scale training from 36.2% to 39.0% mAP (+ 2.8%). It boosts much stronger detectors like FCOS with ResNeXt-101 DCN v2 under 2x multi-scale training from 46.1% to 47.9% (+ 1.8%). Compared with a classical teacher-based method FGFI, LGD not only performs better without requiring pretrained teacher but also reduces 51% training cost beyond inherent student learning. Codes are available at https://github.com/megvii-research/LGD.
We present Hierarchical Memory Matching Network (HMMN) for semi-supervised video object segmentation. Based on a recent memory-based method [33], we propose two advanced memory read modules that enable us to perform memory reading in multiple scales while exploiting temporal smoothness. We first propose a kernel guided memory matching module that replaces the non-local dense memory read, commonly adopted in previous memory-based methods. The module imposes the temporal smoothness constraint in the memory read, leading to accurate memory retrieval. More importantly, we introduce a hierarchical memory matching scheme and propose a top-k guided memory matching module in which memory read on a fine-scale is guided by that on a coarse-scale. With the module, we perform memory read in multiple scales efficiently and leverage both high-level semantic and low-level fine-grained memory features to predict detailed object masks. Our network achieves state-of-the-art performance on the validation sets of DAVIS 2016/2017 (90.8% and 84.7%) and YouTube-VOS 2018/2019 (82.6% and 82.5%), and test-dev set of DAVIS 2017 (78.6%). The source code and model are available online: https://github.com/Hongje/HMMN.
Real-world object detection is highly desired to be equipped with the learning expandability that can enlarge its detection classes incrementally. Moreover, such learning from only few annotated training samples further adds the flexibility for the object detector, which is highly expected in many applications such as autonomous driving, robotics, etc. However, such sequential learning scenario with few-shot training samples generally causes catastrophic forgetting and dramatic overfitting. In this paper, to address the above incremental few-shot learning issues, a novel Incremental Few-Shot Object Detection (iFSOD) method is proposed to enable the effective continual learning from few-shot samples. Specifically, a Double-Branch Framework (DBF) is proposed to decouple the feature representation of base and novel (few-shot) class, which facilitates both the old-knowledge retention and new-class adaption simultaneously. Furthermore, a progressive model updating rule is carried out to preserve the long-term memory on old classes effectively when adapt to sequential new classes. Moreover, an inter-task class separation loss is proposed to extend the decision region of new-coming classes for better feature discrimination. We conduct experiments on both Pascal VOC and MS-COCO, which demonstrate that our method can effectively solve the problem of incremental few-shot detection and significantly improve the detection accuracy on both base and novel classes.
Instance-aware segmentation of unseen objects is essential for a robotic system in an unstructured environment. Although previous works achieved encouraging results, they were limited to segmenting the only visible regions of unseen objects. For robotic manipulation in a cluttered scene, amodal perception is required to handle the occluded objects behind others. This paper addresses Unseen Object Amodal Instance Segmentation (UOAIS) to detect 1) visible masks, 2) amodal masks, and 3) occlusions on unseen object instances. For this, we propose a Hierarchical Occlusion Modeling (HOM) scheme designed to reason about the occlusion by assigning a hierarchy to a feature fusion and prediction order. We evaluated our method on three benchmarks (tabletop, indoors, and bin environments) and achieved state-of-the-art (SOTA) performance. Robot demos for picking up occluded objects, codes, and datasets are available at https://sites.google.com/view/uoais
Monocular 3D object detection encounters occlusion problems in many application scenarios, such as traffic monitoring, pedestrian monitoring, etc., which leads to serious false negative. Multi-view object detection effectively solves this problem by combining data from different perspectives. However, due to label confusion and feature confusion, the orientation estimation of multi-view 3D object detection is intractable, which is important for object tracking and intention prediction. In this paper, we propose a novel multi-view 3D object detection method named MVM3Det which simultaneously estimates the 3D position and orientation of the object according to the multi-view monocular information. The method consists of two parts: 1) Position proposal network, which integrates the features from different perspectives into consistent global features through feature orthogonal transformation to estimate the position. 2) Multi-branch orientation estimation network, which introduces feature perspective pooling to overcome the two confusion problems during the orientation estimation. In addition, we present a first dataset for multi-view 3D object detection named MVM3D. Comparing with State-Of-The-Art (SOTA) methods on our dataset and public dataset WildTrack, our method achieves very competitive results.
Methodologies for incorporating the uncertainties characteristic of data-driven object detectors into object tracking algorithms are explored. Object tracking methods rely on measurement error models, typically in the form of measurement noise, false positive rates, and missed detection rates. Each of these quantities, in general, can be dependent on object or measurement location. However, for detections generated from neural-network processed camera inputs, these measurement error statistics are not sufficient to represent the primary source of errors, namely a dissimilarity between run-time sensor input and the training data upon which the detector was trained. To this end, we investigate incorporating data uncertainty into object tracking methods such as to improve the ability to track objects, and particularly those which out-of-distribution w.r.t. training data. The proposed methodologies are validated on an object tracking benchmark as well on experiments with a real autonomous aircraft.