In recent years, the brain and cognitive sciences have made great strides developing a mechanistic understanding of object recognition in mature brains. Despite this progress, fundamental questions remain about the origins and computational foundations of object recognition. What learning algorithms underlie object recognition in newborn brains? Since newborn animals learn largely through unsupervised learning, we explored whether unsupervised learning algorithms can be used to predict the view-invariant object recognition behavior of newborn chicks. Specifically, we used feature representations derived from unsupervised deep neural networks (DNNs) as inputs to cognitive models of categorization. We show that features derived from unsupervised DNNs make competitive predictions about chick behavior compared to supervised features. More generally, we argue that linking controlled-rearing studies to image-computable DNN models opens new experimental avenues for studying the origins and computational basis of object recognition in newborn animals.
Capturing contextual dependencies has proven useful to improve the representational power of deep neural networks. Recent approaches that focus on modeling global context, such as self-attention and non-local operation, achieve this goal by enabling unconstrained pairwise interactions between elements. In this work, we consider learning representations for deformable objects which can benefit from context exploitation by modeling the structural dependencies that the data intrinsically possesses. To this end, we provide a novel structure-regularized attention mechanism, which formalizes feature interaction as structural factorization through the use of a pair of light-weight operations. The instantiated building blocks can be directly incorporated into modern convolutional neural networks, to boost the representational power in an efficient manner. Comprehensive studies on multiple tasks and empirical comparisons with modern attention mechanisms demonstrate the gains brought by our method in terms of both performance and model complexity. We further investigate its effect on feature representations, showing that our trained models can capture diversified representations characterizing object parts without resorting to extra supervision.
Entities Object Localization (EOL) aims to evaluate how grounded or faithful a description is, which consists of caption generation and object grounding. Previous works tackle this problem by jointly training the two modules in a framework, which limits the complexity of each module. Therefore, in this work, we propose to divide these two modules into two stages and improve them respectively to boost the whole system performance. For the caption generation, we propose a Unified Multi-modal Pre-training Model (UMPM) to generate event descriptions with rich objects for better localization. For the object grounding, we fine-tune the state-of-the-art detection model MDETR and design a post processing method to make the grounding results more faithful. Our overall system achieves the state-of-the-art performances on both sub-tasks in Entities Object Localization challenge at Activitynet 2021, with 72.57 localization accuracy on the testing set of sub-task I and 0.2477 F1_all_per_sent on the hidden testing set of sub-task II.
Most object detection methods use bounding boxes to encode and represent the object shape and location. In this work, we explore a fuzzy representation of object regions using Gaussian distributions, which provides an implicit binary representation as (potentially rotated) ellipses. We also present a similarity measure for the Gaussian distributions based on the Hellinger Distance, which can be viewed as a Probabilistic Intersection-over-Union (ProbIoU). Our experimental results show that the proposed Gaussian representations are closer to annotated segmentation masks in publicly available datasets, and that loss functions based on ProbIoU can be successfully used to regress the parameters of the Gaussian representation. Furthermore, we present a simple mapping scheme from traditional (or rotated) bounding boxes to Gaussian representations, allowing the proposed ProbIoU-based losses to be seamlessly integrated into any object detector.
We present a local anomaly detection method in videos. As opposed to most existing methods that are computationally expensive and are not very generalizable across different video scenes, we propose an adversarial framework that learns the temporal local appearance variations by predicting the appearance of a normally behaving object in the next frame of a scene by only relying on its current and past appearances. In the presence of an abnormally behaving object, the reconstruction error between the real and the predicted next appearance of that object indicates the likelihood of an anomaly. Our method is competitive with the existing state-of-the-art while being significantly faster for both training and inference and being better at generalizing to unseen video scenes.
This paper presents an approach of two-stage convolutional neural network (CNN) for detection of fingertips so that an interaction of the fingertips with a 3D object in the virtual environment (VR) can be established. The first-stage CNN is assigned to detect and locate the hand. Subsequently, the detected hand is cropped, resized, and fed to the second stage CNN for predicting the coordinates of fingertips. Next, a tracker is employed to track the hand continuously so that the system becomes reliable in real-time performance. The VR environments are designed to demonstrate the performance of the fingertip-based interaction system. The proposed method focuses on the geometric transformation of a virtual 3D object by using the gesture of the thumb and index finger. In particular, the distance of the thumb and index fingertips is employed to scale a 3D object in virtual environment. To realize the system, a dataset of 1000 images, named, Thumb Index 1000 (TI1K) dataset, is developed including those variations that are commonly-seen in real-life thumb and index fingers. The system is evaluated with the aid of a number of participants and virtual objects that are distinctive in nature. The proposed approach attains the desired goal and performs in real-time seamlessly to facilitate the human-computer interaction in the VR environment.
The paper proposes a method to convert a deep learning object detector into an equivalent spiking neural network. The aim is to provide a conversion framework that is not constrained to shallow network structures and classification problems as in state-of-the-art conversion libraries. The results show that models of higher complexity, such as the RetinaNet object detector, can be converted with limited loss in performance.
Obtaining a useful estimate of an object from highly incomplete imaging measurements remains a holy grail of imaging science. Deep learning methods have shown promise in learning object priors or constraints to improve the conditioning of an ill-posed imaging inverse problem. In this study, a framework for estimating an object of interest that is semantically related to a known prior image, is proposed. An optimization problem is formulated in the disentangled latent space of a style-based generative model, and semantically meaningful constraints are imposed using the disentangled latent representation of the prior image. Stable recovery from incomplete measurements with the help of a prior image is theoretically analyzed. Numerical experiments demonstrating the superior performance of our approach as compared to related methods are presented.
Unsupervised domain adaptation for object detection is a challenging problem with many real-world applications. Unfortunately, it has received much less attention than supervised object detection. Models that try to address this task tend to suffer from a shortage of annotated training samples. Moreover, existing methods of feature alignments are not sufficient to learn domain-invariant representations. To address these limitations, we propose a novel augmented feature alignment network (AFAN) which integrates intermediate domain image generation and domain-adversarial training into a unified framework. An intermediate domain image generator is proposed to enhance feature alignments by domain-adversarial training with automatically generated soft domain labels. The synthetic intermediate domain images progressively bridge the domain divergence and augment the annotated source domain training data. A feature pyramid alignment is designed and the corresponding feature discriminator is used to align multi-scale convolutional features of different semantic levels. Last but not least, we introduce a region feature alignment and an instance discriminator to learn domain-invariant features for object proposals. Our approach significantly outperforms the state-of-the-art methods on standard benchmarks for both similar and dissimilar domain adaptations. Further extensive experiments verify the effectiveness of each component and demonstrate that the proposed network can learn domain-invariant representations.
We propose a method for object-aware 3D egocentric pose estimation that tightly integrates kinematics modeling, dynamics modeling, and scene object information. Unlike prior kinematics or dynamics-based approaches where the two components are used disjointly, we synergize the two approaches via dynamics-regulated training. At each timestep, a kinematic model is used to provide a target pose using video evidence and simulation state. Then, a prelearned dynamics model attempts to mimic the kinematic pose in a physics simulator. By comparing the pose instructed by the kinematic model against the pose generated by the dynamics model, we can use their misalignment to further improve the kinematic model. By factoring in the 6DoF pose of objects (e.g., chairs, boxes) in the scene, we demonstrate for the first time, the ability to estimate physically-plausible 3D human-object interactions using a single wearable camera. We evaluate our egocentric pose estimation method in both controlled laboratory settings and real-world scenarios.
We present Temporal and Object Quantification Networks (TOQ-Nets), a new class of neuro-symbolic networks with a structural bias that enables them to learn to recognize complex relational-temporal events. This is done by including reasoning layers that implement finite-domain quantification over objects and time. The structure allows them to generalize directly to input instances with varying numbers of objects in temporal sequences of varying lengths. We evaluate TOQ-Nets on input domains that require recognizing event-types in terms of complex temporal relational patterns. We demonstrate that TOQ-Nets can generalize from small amounts of data to scenarios containing more objects than were present during training and to temporal warpings of input sequences.
We tackle the problem of object-centric learning on point clouds, which is crucial for high-level relational reasoning and scalable machine intelligence. In particular, we introduce a framework, SPAIR3D, to factorize a 3D point cloud into a spatial mixture model where each component corresponds to one object. To model the spatial mixture model on point clouds, we derive the Chamfer Mixture Loss, which fits naturally into our variational training pipeline. Moreover, we adopt an object-specification scheme that describes each object's location relative to its local voxel grid cell. Such a scheme allows SPAIR3D to model scenes with an arbitrary number of objects. We evaluate our method on the task of unsupervised scene decomposition. Experimental results demonstrate that SPAIR3D has strong scalability and is capable of detecting and segmenting an unknown number of objects from a point cloud in an unsupervised manner.
Phase can be reliably estimated from a single diffracted intensity image, if a faithful prior information about the object is available. Examples include amplitude bounds, object support, sparsity in the spatial or a transform domain, deep image prior and the prior learnt from the labelled datasets by a deep neural network. Deep learning facilitates state of art reconstruction quality but requires a large labelled dataset (ground truth-measurement pair acquired in the same experimental conditions) for training. To alleviate this data requirement problem, this letter proposes a zero-shot learning method. The letter demonstrates that the object-prior learnt by a deep neural network while being trained for a denoising task can also be utilized for the phase retrieval, if the diffraction physics is effectively enforced on the network output. The letter additionally demonstrates that the incorporation of total variation in the proposed zero-shot framework facilitates the reconstruction of similar quality in lesser time (e.g. ~8.5 fold, for a test reported in this letter).
Estimating 3D hand and object pose from a single image is an extremely challenging problem: hands and objects are often self-occluded during interactions, and the 3D annotations are scarce as even humans cannot directly label the ground-truths from a single image perfectly. To tackle these challenges, we propose a unified framework for estimating the 3D hand and object poses with semi-supervised learning. We build a joint learning framework where we perform explicit contextual reasoning between hand and object representations by a Transformer. Going beyond limited 3D annotations in a single image, we leverage the spatial-temporal consistency in large-scale hand-object videos as a constraint for generating pseudo labels in semi-supervised learning. Our method not only improves hand pose estimation in challenging real-world dataset, but also substantially improve the object pose which has fewer ground-truths per instance. By training with large-scale diverse videos, our model also generalizes better across multiple out-of-domain datasets. Project page and code: https://stevenlsw.github.io/Semi-Hand-Object
Instance segmentation can detect where the objects are in an image, but hard to understand the relationship between them. We pay attention to a typical relationship, relative saliency. A closely related task, salient object detection, predicts a binary map highlighting a visually salient region while hard to distinguish multiple objects. Directly combining two tasks by post-processing also leads to poor performance. There is a lack of research on relative saliency at present, limiting the practical applications such as content-aware image cropping, video summary, and image labeling. In this paper, we study the Salient Object Ranking (SOR) task, which manages to assign a ranking order of each detected object according to its visual saliency. We propose the first end-to-end framework of the SOR task and solve it in a multi-task learning fashion. The framework handles instance segmentation and salient object ranking simultaneously. In this framework, the SOR branch is independent and flexible to cooperate with different detection methods, so that easy to use as a plugin. We also introduce a Position-Preserved Attention (PPA) module tailored for the SOR branch. It consists of the position embedding stage and feature interaction stage. Considering the importance of position in saliency comparison, we preserve absolute coordinates of objects in ROI pooling operation and then fuse positional information with semantic features in the first stage. In the feature interaction stage, we apply the attention mechanism to obtain proposals' contextualized representations to predict their relative ranking orders. Extensive experiments have been conducted on the ASR dataset. Without bells and whistles, our proposed method outperforms the former state-of-the-art method significantly. The code will be released publicly available.
Tracking-by-detection is a very popular framework for single object tracking which attempts to search the target object within a local search window for each frame. Although such local search mechanism works well on simple videos, however, it makes the trackers sensitive to extremely challenging scenarios, such as heavy occlusion and fast motion. In this paper, we propose a novel and general target-aware attention mechanism (termed TANet) and integrate it with tracking-by-detection framework to conduct joint local and global search for robust tracking. Specifically, we extract the features of target object patch and continuous video frames, then we concatenate and feed them into a decoder network to generate target-aware global attention maps. More importantly, we resort to adversarial training for better attention prediction. The appearance and motion discriminator networks are designed to ensure its consistency in spatial and temporal views. In the tracking procedure, we integrate the target-aware attention with multiple trackers by exploring candidate search regions for robust tracking. Extensive experiments on both short-term and long-term tracking benchmark datasets all validated the effectiveness of our algorithm. The project page of this paper can be found at \url{https://sites.google.com/view/globalattentiontracking/home/extend}.
Knowledge distillation constitutes a simple yet effective way to improve the performance of a compact student network by exploiting the knowledge of a more powerful teacher. Nevertheless, the knowledge distillation literature remains limited to the scenario where the student and the teacher tackle the same task. Here, we investigate the problem of transferring knowledge not only across architectures but also across tasks. To this end, we study the case of object detection and, instead of following the standard detector-to-detector distillation approach, introduce a classifier-to-detector knowledge transfer framework. In particular, we propose strategies to exploit the classification teacher to improve both the detector's recognition accuracy and localization performance. Our experiments on several detectors with different backbones demonstrate the effectiveness of our approach, allowing us to outperform the state-of-the-art detector-to-detector distillation methods.
When transporting an object, we unconsciously adapt our movement to its properties, for instance by slowing down when the item is fragile. The most relevant features of an object are immediately revealed to a human observer by the way the handling occurs, without any need for verbal description. It would greatly facilitate collaboration to enable humanoid robots to perform movements that convey similar intuitive cues to the observers. In this work, we focus on how to generate robot motion adapted to the hidden properties of the manipulated objects, such as their weight and fragility. We explore the possibility of leveraging Generative Adversarial Networks to synthesize new actions coherent with the properties of the object. The use of a generative approach allows us to create new and consistent motion patterns, without the need of collecting a large number of recorded human-led demonstrations. Besides, the informative content of the actions is preserved. Our results show that Generative Adversarial Nets can be a powerful tool for the generation of novel and meaningful transportation actions, which result effectively modulated as a function of the object weight and the carefulness required in its handling.
Recent self-supervised pretraining methods for object detection largely focus on pretraining the backbone of the object detector, neglecting key parts of detection architecture. Instead, we introduce DETReg, a new self-supervised method that pretrains the entire object detection network, including the object localization and embedding components. During pretraining, DETReg predicts object localizations to match the localizations from an unsupervised region proposal generator and simultaneously aligns the corresponding feature embeddings with embeddings from a self-supervised image encoder. We implement DETReg using the DETR family of detectors and show that it improves over competitive baselines when finetuned on COCO, PASCAL VOC, and Airbus Ship benchmarks. In low-data regimes DETReg achieves improved performance, e.g., when training with only 1% of the labels and in the few-shot learning settings.
Recent advances in deep learning have brought significant progress in visual grounding tasks such as language-guided video object segmentation. However, collecting large datasets for these tasks is expensive in terms of annotation time, which represents a bottleneck. To this end, we propose a novel method, namely SynthRef, for generating synthetic referring expressions for target objects in an image (or video frame), and we also present and disseminate the first large-scale dataset with synthetic referring expressions for video object segmentation. Our experiments demonstrate that by training with our synthetic referring expressions one can improve the ability of a model to generalize across different datasets, without any additional annotation cost. Moreover, our formulation allows its application to any object detection or segmentation dataset.