We introduce the task of 3D object localization in RGB-D scans using natural language descriptions. As input, we assume a point cloud of a scanned 3D scene along with a free-form description of a specified target object. To address this task, we propose ScanRefer, learning a fused descriptor from 3D object proposals and encoded sentence embeddings. This fused descriptor correlates language expressions with geometric features, enabling regression of the 3D bounding box of a target object. We also introduce the ScanRefer dataset, containing 51,583 descriptions of 11,046 objects from 800 ScanNet scenes. ScanRefer is the first large-scale effort to perform object localization via natural language expression directly in 3D.
Weakly supervised object localization is a challenging task in which the object of interest should be localized while learning its appearance. State-of-the-art methods recycle the architecture of a standard CNN by using the activation maps of the last layer for localizing the object. While this approach is simple and works relatively well, object localization relies on different features than classification, thus, a specialized localization mechanism is required during training to improve performance. In this paper, we propose a convolutional, multi-scale spatial localization network that provides accurate localization for the object of interest. Experimental results on CUB-200-2011 and ImageNet datasets show that our proposed approach provides competitive performance for weakly supervised localization.
Object detection in aerial images is a challenging task due to the lack of visible features and variant orientation of objects. Significant progress has been made recently for predicting targets from aerial images with horizontal bounding boxes (HBBs) and oriented bounding boxes (OBBs) using two-stage detectors with region based convolutional neural networks (R-CNN), involving object localization in one stage and object classification in the other. However, the computational complexity in two-stage detectors is often high, especially for orientational object detection, due to anchor matching and using regions of interest (RoI) pooling for feature extraction. In this paper, we propose a one-stage anchor free detector for orientational object detection, namely, an interactive embranchment network (IENet), which is built upon a detector with prediction in per-pixel fashion. First, a novel geometric transformation is employed to better represent the oriented object in angle prediction, then a branch interactive module with a self-attention mechanism is developed to fuse features from classification and box regression branches. Finally, we introduce an enhanced intersection over union (IoU) loss for OBB detection, which is computationally more efficient than regular polygon IoU. Experiments conducted demonstrate the effectiveness and the superiority of our proposed method, as compared with state-of-the-art detectors.
Interpretability is an emerging area of research in trustworthy machine learning. Safe deployment of machine learning system mandates that the prediction and its explanation be reliable and robust. Recently, it has been shown that the explanations could be manipulated easily by adding visually imperceptible perturbations to the input while keeping the model's prediction intact. In this work, we study the problem of attributional robustness (i.e. models having robust explanations) by showing an upper bound for attributional vulnerability in terms of spatial correlation between the input image and its explanation map. We propose a training methodology that learns robust features by minimizing this upper bound using soft-margin triplet loss. Our methodology of robust attribution training (\textit{ART}) achieves the new state-of-the-art attributional robustness measure by a margin of $\approx$ 6-18 $\%$ on several standard datasets, ie. SVHN, CIFAR-10 and GTSRB. We further show the utility of the proposed robust training technique (\textit{ART}) in the downstream task of weakly supervised object localization by achieving the new state-of-the-art performance on CUB-200 dataset.
Mistakes/uncertainties in object detection could lead to catastrophes when deploying robots in the real world. In this paper, we measure the uncertainties of object localization to minimize this kind of risk. Uncertainties emerge upon challenging cases like occlusion. The bounding box borders of an occluded object can have multiple plausible configurations. We propose a deep multivariate mixture of Gaussians model for probabilistic object detection. The covariances help to learn the relationship between the borders, and the mixture components potentially learn different configurations of an occluded part. Quantitatively, our model improves the AP of the baselines by 3.9% and 1.4% on CrowdHuman and MS-COCO respectively with almost no computational or memory overhead. Qualitatively, our model enjoys explainability since the resulting covariance matrices and the mixture components help measure uncertainties.
We propose a novel method that tracks fast moving objects, mainly non-uniform spherical, in full 6 degrees of freedom, estimating simultaneously their 3D motion trajectory, 3D pose and object appearance changes with a time step that is a fraction of the video frame exposure time. The sub-frame object localization and appearance estimation allows realistic temporal super-resolution and precise shape estimation. The method, called TbD-3D (Tracking by Deblatting in 3D) relies on a novel reconstruction algorithm which solves a piece-wise deblurring and matting problem. The 3D rotation is estimated by minimizing the reprojection error. As a second contribution, we present a new challenging dataset with fast moving objects that change their appearance and distance to the camera. High speed camera recordings with zero lag between frame exposures were used to generate videos with different frame rates annotated with ground-truth trajectory and pose.
Weakly Supervised Object Localization (WSOL) methods generate both classification and localization results by learning from only image category labels. Previous methods usually utilize class activation map (CAM) to obtain target object regions. However, most of them only focus on improving foreground object parts in CAM, but ignore the important effect of its background contents. In this paper, we propose a confidence segmentation (ConfSeg) module that builds confidence score for each pixel in CAM without introducing additional hyper-parameters. The generated sample-specific confidence mask is able to indicate the extent of determination for each pixel in CAM, and further supervises additional CAM extended from internal feature maps. Besides, we introduce Co-supervised Augmentation (CoAug) module to capture feature-level representation for foreground and background parts in CAM separately. Then a metric loss is applied at batch sample level to augment distinguish ability of our model, which helps a lot to localize more related object parts. Our final model, CSoA, combines the two modules and achieves superior performance, e.g. $37.69\%$ and $48.81\%$ Top-1 localization error on CUB-200 and ILSVRC datasets, respectively, which outperforms all previous methods and becomes the new state-of-the-art.
Humans are able to localize objects in the environment using both visual and auditory cues, integrating information from multiple modalities into a common reference frame. We introduce a system that can leverage unlabeled audio-visual data to learn to localize objects (moving vehicles) in a visual reference frame, purely using stereo sound at inference time. Since it is labor-intensive to manually annotate the correspondences between audio and object bounding boxes, we achieve this goal by using the co-occurrence of visual and audio streams in unlabeled videos as a form of self-supervision, without resorting to the collection of ground-truth annotations. In particular, we propose a framework that consists of a vision "teacher" network and a stereo-sound "student" network. During training, knowledge embodied in a well-established visual vehicle detection model is transferred to the audio domain using unlabeled videos as a bridge. At test time, the stereo-sound student network can work independently to perform object localization us-ing just stereo audio and camera meta-data, without any visual input. Experimental results on a newly collected Au-ditory Vehicle Tracking dataset verify that our proposed approach outperforms several baseline approaches. We also demonstrate that our cross-modal auditory localization approach can assist in the visual localization of moving vehicles under poor lighting conditions.
Within the last decade, neural network based predictors have demonstrated impressive - and at times super-human - capabilities. This performance is often paid for with an intransparent prediction process and thus has sparked numerous contributions in the novel field of explainable artificial intelligence (XAI). In this paper, we focus on a popular and widely used method of XAI, the Layer-wise Relevance Propagation (LRP). Since its initial proposition LRP has evolved as a method, and a best practice for applying the method has tacitly emerged, based however on humanly observed evidence alone. In this paper we investigate - and for the first time quantify - the effect of this current best practice on feedforward neural networks in a visual object detection setting. The results verify that the layer-dependent approach to LRP applied in recent literature better represents the model's reasoning, and at the same time increases the object localization and class discriminativity of LRP.
Human-object interaction detection is an important and relatively new class of visual relationship detection tasks, essential for deeper scene understanding. Most existing approaches decompose the problem into object localization and interaction recognition. Despite showing progress, these approaches only rely on the appearances of humans and objects and overlook the available context information, crucial for capturing subtle interactions between them. We propose a contextual attention framework for human-object interaction detection. Our approach leverages context by learning contextually-aware appearance features for human and object instances. The proposed attention module then adaptively selects relevant instance-centric context information to highlight image regions likely to contain human-object interactions. Experiments are performed on three benchmarks: V-COCO, HICO-DET and HCVRD. Our approach outperforms the state-of-the-art on all datasets. On the V-COCO dataset, our method achieves a relative gain of 4.4% in terms of role mean average precision ($mAP_{role}$), compared to the existing best approach.
Weakly supervised object localization has recently attracted attention since it aims to identify both class labels and locations of objects by using image-level labels. Most previous methods utilize the activation map corresponding to the highest activation source. Exploiting only one activation map of the highest probability class is often biased into limited regions or sometimes even highlights background regions. To resolve these limitations, we propose to use activation maps, named combinational class activation maps (CCAM), which are linear combinations of activation maps from the highest to the lowest probability class. By using CCAM for localization, we suppress background regions to help highlighting foreground objects more accurately. In addition, we design the network architecture to consider spatial relationships for localizing relevant object regions. Specifically, we integrate non-local modules into an existing base network at both low- and high-level layers. Our final model, named non-local combinational class activation maps (NL-CCAM), obtains superior performance compared to previous methods on representative object localization benchmarks including ILSVRC 2016 and CUB-200-2011. Furthermore, we show that the proposed method has a great capability of generalization by visualizing other datasets.
Perturbation-based explanation methods often measure the contribution of an input feature to an image classifier's outputs by heuristically removing it via e.g. blurring, adding noise, or graying out, which often produce unrealistic, out-of-samples. Instead, we propose to integrate a generative inpainter into three representative attribution methods to remove an input feature. Our proposed change improved all three methods in (1) generating more plausible counterfactual samples under the true data distribution; (2) being more accurate according to three metrics: object localization, deletion, and saliency metrics; and (3) being more robust to hyperparameter changes. Our findings were consistent across both ImageNet and Places365 datasets and two different pairs of classifiers and inpainters.
In this paper, we address the problem of weakly supervised object localization (WSL), which trains a detection network on the dataset with only image-level annotations. The proposed approach is built on the observation that the proposal set from the training dataset is a collection of background, object parts, and objects. Several strategies are taken to adaptively eliminate the noisy proposals and generate pseudo object-level annotations for the weakly labeled dataset. A multiple instance learning (MIL) algorithm enhanced by mask-out strategy is adopted to collect the class-specific object proposals, which are then utilized to adapt a pre-trained classification network to a detection network. In addition, the detection results from the detection network are re-weighted by jointly considering the detection scores and the overlap ratio of proposals in a proposal subset optimization framework. The optimal proposals work as object-level labels that enable a pseudo-strongly supervised dataset for training the detection network. Consequently, we establish a fully adaptive detection network. Extensive evaluations on the PASCAL VOC 2007 and 2012 datasets demonstrate a significant improvement compared with the state-of-the-art methods.
Translucency is prevalent in everyday scenes. As such, perception of transparent objects is essential for robots to perform manipulation. Compared with texture-rich or texture-less Lambertian objects, transparency induces significant uncertainty on object appearances. Ambiguity can be due to changes in lighting, viewpoint, and backgrounds, each of which brings challenges to existing object pose estimation algorithms. In this work, we propose LIT, a two-stage method for transparent object pose estimation using light-field sensing and photorealistic rendering. LIT employs multiple filters specific to light-field imagery in deep networks to capture transparent material properties, with robust depth and pose estimators based on generative sampling. Along with the LIT algorithm, we introduce the light-field transparent object dataset ProLIT for the tasks of recognition, localization and pose estimation. With respect to this ProLIT dataset, we demonstrate that LIT can outperform both state-of-the-art end-to-end pose estimation methods and a generative pose estimator on transparent objects.
Localizing objects with weak supervision in an image is a key problem of the research in computer vision community. Many existing Weakly-Supervised Object Localization (WSOL) approaches tackle this problem by estimating the most discriminative regions with feature maps (activation maps) obtained by Deep Convolutional Neural Network, that is, only the objects or parts of them with the most discriminative response will be located. However, the activation maps often display different local maximum responses or relatively weak response when one image contains multiple objects with the same type or small objects. In this paper, we propose a simple yet effective multi-scale discriminative region discovery method to localize not only more integral objects but also as many as possible with only image-level class labels. The gradient weights flowing into different convolutional layers of CNN are taken as the input of our method, which is different from previous methods only considering that of the final convolutional layer. To mine more discriminative regions for the task of object localization, the multiple local maximum from the gradient weight maps are leveraged to generate the localization map with a parallel sliding window. Furthermore, multi-scale localization maps from different convolutional layers are fused to produce the final result. We evaluate the proposed method with the foundation of VGGnet on the ILSVRC 2016, CUB-200-2011 and PASCAL VOC 2012 datasets. On ILSVRC 2016, the proposed method yields the Top-1 localization error of 48.65\%, which outperforms previous results by 2.75\%. On PASCAL VOC 2012, our approach achieve the highest localization accuracy of 0.43. Even for CUB-200-2011 dataset, our method still achieves competitive results.
Image segmentation is widely used in a variety of computer vision tasks, such as object localization and recognition, boundary detection, and medical imaging. This thesis proposes deep learning architectures to improve automatic object localization and boundary delineation for salient object segmentation in natural images and for 2D medical image segmentation. First, we propose and evaluate a novel dilated dense encoder-decoder architecture with a custom dilated spatial pyramid pooling block to accurately localize and delineate boundaries for salient object segmentation. The dilation offers better spatial understanding and the dense connectivity preserves features learned at shallower levels of the network for better localization. Tested on three publicly available datasets, our architecture outperforms the state-of-the-art for one and is very competitive on the other two. Second, we propose and evaluate a custom 2D dilated dense UNet architecture for accurate lesion localization and segmentation in medical images. This architecture can be utilized as a stand-alone segmentation framework or used as a rich feature extracting backbone to aid other models in medical image segmentation. Our architecture outperforms all baseline models for accurate lesion localization and segmentation on a new dataset. We furthermore explore the main considerations that should be taken into account for 3D medical image segmentation, among them preprocessing techniques and specialized loss functions.
Predicting the future trajectory of agents from visual observations is an important problem for realization of safe and effective navigation of autonomous systems in dynamic environments. This paper focuses on two important aspects of future trajectory forecast which are particularly relevant for mobile platforms: 1) modeling uncertainty of the predictions, particularly from egocentric views, where uncertainty in the interactive reactions and behaviors of other agents must consider the uncertainty in the ego-motion, and 2) modeling multi-modality nature of the problem, which are particularly prevalent at junctions in urban traffic scenes. To address these problems in a unified approach, we propose NEMO (Noisy Ego MOtion priors for future object localization) for future forecast of agents in the egocentric view. In the proposed approach, a predictive distribution of future forecast is jointly modeled with the uncertainty of predictions. For this, we divide the problem into two tasks: future ego-motion prediction and future object localization. We first model the multi-modal distribution of future ego-motion with uncertainty estimates. The resulting distribution of ego-behavior is used to sample multiple modes of future ego-motion. Then, each modality is used as a prior to understand the interactions between the ego-vehicle and target agent. We predict the multi-modal future locations of the target from individual modes of the ego-vehicle while modeling the uncertainty of the target's behavior. To this end, we extensively evaluate the proposed framework using the publicly available benchmark dataset (HEV-I) supplemented with odometry data from an Inertial Measurement Unit (IMU).
Despite deep convolutional neural networks boost the performance of image classification and segmentation in digital pathology analysis, they are usually weak in interpretability for clinical applications or require heavy annotations to achieve object localization. To overcome this problem, we propose a weakly supervised learning-based approach that can effectively learn to localize the discriminative evidence for a diagnostic label from weakly labeled training data. Experimental results show that our proposed method can reliably pinpoint the location of cancerous evidence supporting the decision of interest, while still achieving a competitive performance on glimpse-level and slide-level histopathologic cancer detection tasks.
The research on recognizing the most discriminative regions provides referential information for weakly supervised object localization with only image-level annotations. However, the most discriminative regions usually conceal the other parts of the object, thereby impeding entire object recognition and localization. To tackle this problem, the Dual-attention Focused Module (DFM) is proposed to enhance object localization performance. Specifically, we present a dual attention module for information fusion, consisting of a position branch and a channel one. In each branch, the input feature map is deduced into an enhancement map and a mask map, thereby highlighting the most discriminative parts or hiding them. For the position mask map, we introduce a focused matrix to enhance it, which utilizes the principle that the pixels of an object are continuous. Between these two branches, the enhancement map is integrated with the mask map, aiming at partially compensating the lost information and diversifies the features. With the dual-attention module and focused matrix, the entire object region could be precisely recognized with implicit information. We demonstrate outperforming results of DFM in experiments. In particular, DFM achieves state-of-the-art performance in localization accuracy in ILSVRC 2016 and CUB-200-2011.
With rapid adoption of deep learning in critical applications, the question of when and how much to trust these models often arises, which drives the need to quantify the inherent uncertainties. While identifying all sources that account for the stochasticity of models is challenging, it is common to augment predictions with confidence intervals to convey the expected variations in a model's behavior. We require prediction intervals to be well-calibrated, reflect the true uncertainties, and to be sharp. However, existing techniques for obtaining prediction intervals are known to produce unsatisfactory results in at least one of these criteria. To address this challenge, we develop a novel approach for building calibrated estimators. More specifically, we use separate models for prediction and interval estimation, and pose a bi-level optimization problem that allows the former to leverage estimates from the latter through an \textit{uncertainty matching} strategy. Using experiments in regression, time-series forecasting, and object localization, we show that our approach achieves significant improvements over existing uncertainty quantification methods, both in terms of model fidelity and calibration error.