WACV 2021 Papers

Visual camera localization using offline maps is widespread in robotics and mobile applications. Most state-of-the-art localization approaches assume static scenes, so maps are often reconstructed once and then kept constant. However, many scenes are dynamic and as changes in the scene happen, future localization attempts may struggle or fail entirely. Therefore, it is important for successful long-term localization to update and maintain maps as new observations of the scene, and changes in it, arrive. We propose a novel method for automatically discovering which points in a map remain stable over time, and which are due to transient changes. To this end, we calculate a stability store for each point based on its visibility over time, weighted by an exponential decay over time. This allows us to introduce the impact of time when scoring points, and distinguishes which points are useful for long-term localization. We evaluate our method on the CMU Extended Seasons dataset (outdoors) and a new indoor dataset of a retail shop, and show the benefit of maintaining a `live map' that integrates updates over time using our exponential decay based method over a static `base map'.

The efficacy of deep learning depends on large-scale data sets that have been carefully curated with reliable data acquisition and annotation processes. However, acquiring such large-scale data sets with precise annotations is very expensive and time-consuming, and the cheap alternatives often yield data sets that have noisy labels. The field has addressed this problem by focusing on training models under two types of label noise: 1) closed-set noise, where some training samples are incorrectly annotated to a training label other than their known true class; and 2) open-set noise, where the training set includes samples that possess a true class that is (strictly) not contained in the set of known training labels. In this work, we study a new variant of the noisy label problem that combines the open-set and closed-set noisy labels, and introduce a benchmark evaluation to assess the performance of training algorithms under this setup. We argue that such problem is more general and better reflects the noisy label scenarios in practice. Furthermore, we propose a novel algorithm, called EvidentialMix, that addresses this problem and compare its performance with the state-of-the-art methods for both closed-set and open-set noise on the proposed benchmark. Our results show that our method produces superior classification results and better feature representations than previous state-of-the-art methods. The code is available at https://github.com/ragavsachdeva/EvidentialMix.

Segmentation of anatomical regions of interest such as vessels or small lesions in medical images is still a difficult problem that is often tackled with manual input by an expert. One of the major challenges for this task is that the appearance of foreground (positive) regions can be similar to background (negative) regions. As a result, many automatic segmentation algorithms tend to exhibit asymmetric errors, typically producing more false positives than false negatives. In this paper, we aim to leverage this asymmetry and train a diverse ensemble of models with very high recall, while sacrificing their precision. Our core idea is straightforward: A diverse ensemble of low precision and high recall models are likely to make different false positive errors (classifying background as foreground in different parts of the image), but the true positives will tend to be consistent. Thus, in aggregate the false positive errors will cancel out, yielding high performance for the ensemble. Our strategy is general and can be applied with any segmentation model. In three different applications (carotid artery segmentation in a neck CT angiography, myocardium segmentation in a cardiovascular MRI and multiple sclerosis lesion segmentation in a brain MRI), we show how the proposed approach can significantly boost the performance of a baseline segmentation method.

The objective of unsupervised person re-identification (Re-ID) is to learn discriminative features without labor-intensive identity annotations. State-of-the-art unsupervised Re-ID methods assign pseudo labels to unlabeled images in the target domain and learn from these noisy pseudo labels. Recently introduced Mean Teacher Model is a promising way to mitigate the label noise. However, during the training, self-ensembled teacher-student networks quickly converge to a consensus which leads to a local minimum. We explore the possibility of using an asymmetric structure inside neural network to address this problem. First, asymmetric branches are proposed to extract features in different manners, which enhances the feature diversity in appearance signatures. Then, our proposed cross-branch supervision allows one branch to get supervision from the other branch, which transfers distinct knowledge and enhances the weight diversity between teacher and student networks. Extensive experiments show that our proposed method can significantly surpass the performance of previous work on both unsupervised domain adaptation and fully unsupervised Re-ID tasks.

Accurately estimating the position of static objects, such as traffic lights, from the moving camera of a self-driving car is a challenging problem. In this work, we present a system that improves the localization of static objects by jointly-optimizing the components of the system via learning. Our system is comprised of networks that perform: 1) 5DoF object pose estimation from a single image, 2) association of objects between pairs of frames, and 3) multi-object tracking to produce the final geo-localization of the static objects within the scene. We evaluate our approach using a publicly-available data set, focusing on traffic lights due to data availability. For each component, we compare against contemporary alternatives and show significantly-improved performance. We also show that the end-to-end system performance is further improved via joint-training of the constituent models.

Lane detection, the process of identifying lane markings as approximated curves, is widely used for lane departure warning and adaptive cruise control in autonomous vehicles. The popular pipeline that solves it in two steps---feature extraction plus post-processing, while useful, is too inefficient and flawed in learning the global context and lanes' long and thin structures. To tackle these issues, we propose an end-to-end method that directly outputs parameters of a lane shape model, using a network built with a transformer to learn richer structures and context. The lane shape model is formulated based on road structures and camera pose, providing physical interpretation for parameters of network output. The transformer models non-local interactions with a self-attention mechanism to capture slender structures and global context. The proposed method is validated on the TuSimple benchmark and shows state-of-the-art accuracy with the most lightweight model size and fastest speed. Additionally, our method shows excellent adaptability to a challenging self-collected lane detection dataset, showing its powerful deployment potential in real applications. Codes are available at https://github.com/liuruijin17/E2ELSPTRs.

Current GAN-based art generation methods produce unoriginal artwork due to their dependence on conditional input. Here, we propose Sketch-And-Paint GAN (SAPGAN), the first model which generates Chinese landscape paintings from end to end, without conditional input. SAPGAN is composed of two GANs: SketchGAN for generation of edge maps, and PaintGAN for subsequent edge-to-painting translation. Our model is trained on a new dataset of traditional Chinese landscape paintings never before used for generative research. A 242-person Visual Turing Test study reveals that SAPGAN paintings are mistaken as human artwork with 55% frequency, significantly outperforming paintings from baseline GANs. Our work lays a groundwork for truly machine-original art generation.

Dense registrations of huge image sets are still challenging due to exhaustive matchings and computationally expensive optimisations. Moreover, the resultant image mosaics often suffer from structural errors such as drift. Here, we propose a novel algorithm to generate global large-scale registrations from thousands of images extracted from multiple videos to derive high-resolution image mosaics which include full frame rate camera trajectories. Our algorithm does not require any initialisations and ensures the effective integration of all available image data by combining efficient and highly parallelised key-frame and loop-closure mechanisms with a novel geodesic interpolation-based reintegration strategy. As a consequence, global refinement can be done in a fraction of iterations compared to traditional optimisation strategies, while effectively avoiding drift and convergence towards inappropriate solutions. We compared our registration strategy with state-of-the-art algorithms and quantitative evaluations revealed millimetre spatial and high angular accuracy. Applicability is demonstrated by registering more than 110,000 frames from multiple scan recordings and provide dense camera trajectories in a globally referenced coordinate system as used for drone-based mappings, ecological studies, object tracking and land surveys.

While general object detection has seen tremendous progress, localization of elliptical objects has received little attention in the literature. Our motivating application is the detection of knots in sawn timber images, which is an important problem since the number and types of knots are visual characteristics that adversely affect the quality of sawn timber. We demonstrate how models can be tailored to the elliptical shape and thereby improve on general purpose detectors; more generally, elliptical defects are common in industrial production, such as enclosed air bubbles when casting glass or plastic. In this paper, we adapt the Faster R-CNN with its Region Proposal Network (RPN) to model elliptical objects with a Gaussian function, and extend the existing Gaussian Proposal Network (GPN) architecture by adding the region-of-interest pooling and regression branches, as well as using the Wasserstein distance as the loss function to predict the precise locations of elliptical objects. Our proposed method has promising results on the lumber knot dataset: knots are detected with an average intersection over union of 73.05%, compared to 63.63% for general purpose detectors. Specific to the lumber application, we also propose an algorithm to correct any misalignment in the raw timber images during scanning, and contribute the first open-source lumber knot dataset by labeling the elliptical knots in the preprocessed images.

We introduce a unified framework for generic video annotation with bounding boxes. Video annotation is a longstanding problem, as it is a tedious and time-consuming process. We tackle two important challenges of video annotation: (1) automatic temporal interpolation and extrapolation of bounding boxes provided by a human annotator on a subset of all frames, and (2) automatic selection of frames to annotate manually. Our contribution is two-fold: first, we propose a model that has both interpolating and extrapolating capabilities; second, we propose a guiding mechanism that sequentially generates suggestions for what frame to annotate next, based on the annotations made previously. We extensively evaluate our approach on several challenging datasets in simulation and demonstrate a reduction in terms of the number of manual bounding boxes drawn by 60% over linear interpolation and by 35% over an off-the shelf tracker. Moreover, we also show 10% annotation time improvement over a state-of-the-art method for video annotation with bounding boxes [25]. Finally, we run human annotation experiments and provide extensive analysis of the results, showing that our approach reduces actual measured annotation time by 50% compared to commonly used linear interpolation.

In this paper we present a novel algorithm for onboard radial distortion correction for unmanned aerial vehicles (UAVs) equipped with an inertial measurement unit (IMU), that runs in real-time. This approach makes calibration procedures redundant, thus allowing for exchange of optics extemporaneously. By utilizing the IMU data, the cameras can be aligned with the gravity direction. This allows us to work with fewer degrees of freedom, and opens up for further intrinsic calibration. We propose a fast and robust minimal solver for simultaneously estimating the focal length, radial distortion profile and motion parameters from homographies. The proposed solver is tested on both synthetic and real data, and perform better or on par with state-of-the-art methods relying on pre-calibration procedures. Code available at: https://github.com/marcusvaltonen/HomLib.

Dot-product attention has wide applications in computer vision and natural language processing. However, its memory and computational costs grow quadratically with the input size. Such growth prohibits its application on high-resolution inputs. To remedy this drawback, this paper proposes a novel efficient attention mechanism equivalent to dot-product attention but with substantially less memory and computational costs. Its resource efficiency allows more widespread and flexible integration of attention modules into a network, which leads to better accuracies. Empirical evaluations demonstrated the effectiveness of its advantages. Models with efficient attention achieved state-of-the-art accuracies on MS-COCO 2017. Further, the resource efficiency democratizes attention to complex models, where high costs prohibit the use of dot-product attention. As an exemplar, a model with efficient attention achieved state-of-the-art accuracies for stereo depth estimation on the Scene Flow dataset. Code is available at https://github.com/cmsflash/efficient-attention.

Traditional 3d video understanding methods process videos frame by frame. We argue that a lot of computation in this mechanism is redundant based on a key observation - adjacent frames in 3D videos have visually similar geometry structure. To handle the redundancy, we propose the Efficient 3D Video Engine (EVE), aiming to avoid the computation of redundant points. It consists of two modules: 1) redundancy removing module designed to detect redundancy and remove it; 2) residual learning module to extract features on non-redundant points. As a simple plug and play framework, EVE can be easily incorporated in main-stream 3D models. Experiments demonstrate that EVE can significantly reduce computation without performance loss on large scale datasets. On the other hand, with similar computation, EVE outperforms the strong baseline by up to 4.1 mIoU on SemanticKITTI. The code is available on https://github.com/ecr23xx/eve.

Knowledge Distillation is an effective method to transfer the learning across deep neural networks. Typically, the dataset originally used for training the Teacher model is chosen as the "Transfer Set" to conduct the knowledge transfer to the Student. However, this original training data may not always be freely available due to privacy or sensitivity concerns. In such scenarios, existing approaches either iteratively compose a synthetic set representative of the original training dataset, one sample at a time or learn a generative model to compose such a transfer set. However, both these approaches involve complex optimization (GAN training or several backpropagation steps to synthesize one sample) and are often computationally expensive. In this paper, as a simple alternative, we investigate the effectiveness of "arbitrary transfer sets" such as random noise, publicly available synthetic, and natural datasets, all of which are completely unrelated to the original training dataset in terms of their visual or semantic contents. Through extensive experiments on multiple benchmark datasets such as MNIST, FMNIST, CIFAR-10 and CIFAR-100, we discover and validate surprising effectiveness of using arbitrary data to conduct knowledge distillation when this dataset is "target-class balanced". We believe that this important observation can potentially lead to designing baselines for the data-free knowledge distillation task.

FPN-based detectors have made significant progress in general object detection,e.g., MS COCO and CityPersons.However, these detectors fail in certain application scenarios,e.g., tiny object detection. In this paper, we argue that the top-down connections between adjacent layers in FPN bring two-side influences for tiny object detection, not only positive. We propose a novel concept, fusion factor, to control information that deep layers deliver to shallow layers,for adapting FPN to tiny object detection. After series of experiments and analysis, we explore how to estimate an effective value of fusion factor for a particular dataset by a statistical method. The estimation is dependent on the number of objects distributed to each layer. Comprehensive experiments are conducted on tiny object detection datasets,e.g., TinyPerson and Tiny CityPersons. Our results show that when configuring FPN with a proper fusion factor, the network is able to achieve significant performance gains over the baseline on tiny object detection datasets. Codes and models will be released.

Numerous interpretability methods have been developed to visually explain the behavior of complex machine learning models by estimating parts of the input image that are critical for the model's prediction. We propose a general pipeline of enhancing visual explanations using image transformations (EVET). EVET considers transformations of the original input image to refine the critical input region based on an intuitive rationale that the region estimated to be important in variously transformed inputs is more important. Our proposed EVET is applicable to existing visual explanation methods without modification. We validate the effectiveness of the proposed method qualitatively and quantitatively to show that the resulting explanation method outperforms the original in terms of faithfulness, localization, and stability. We also demonstrate that EVET can be used to achieve desirable performance with a low computational cost. For example, EVET-applied Grad-CAM achieves performance comparable to Score-CAM, which is the state-ofthe-art activation-based explanation method, while reducing execution time by more than 90% on VOC, COCO, and ImageNet.

Multimodal large-scale datasets for outdoor scenes are mostly designed for urban driving problems. The scenes are highly structured and semantically different from scenarios seen in nature-centered scenes such as gardens or parks. To promote machine learning methods for nature-oriented applications, such as agriculture and gardening, we propose the multimodal synthetic dataset for Enclosed garDEN scenes (EDEN). The dataset features more than 300K images captured from more than 100 garden models. Each image is annotated with various low/high-level vision modalities, including semantic segmentation, depth, surface normals, intrinsic colors, and optical flow. Experimental results on the state-of-the-art methods for semantic segmentation and monocular depth prediction, two important tasks in computer vision, show positive impact of pre-training deep networks on our dataset for unstructured natural scenes. The dataset and related materials will be available at https://lhoangan.github.io/eden.

Measuring the quality of a sports action entails attending to the execution of the short-term components as well as the overall impression of the whole program. In this assessment, both appearance clues and pose dynamics features should be involved. Current approaches often treat a sports routine as a simple fine-grained action, while taking little heed of its complex temporal structure. Besides, most of them rely solely on either appearance or pose features to score the performance. In this paper, we present JCA and ADA blocks that are responsible for reasoning about the coordination among the joints and appearance dynamics throughout the performance. We build our two-stream network upon the separate stack of these blocks. The early blocks capture the fine-grained temporal dependencies while the last ones reason about the long-term coarse-grained relations. We further introduce an annotated dataset of sports images with unusual pose configurations to boost the performance of pose estimation in such scenarios. Our experiments show that the proposed method not only outperforms the previous works in short- term action assessment but also is the first to generalize well to minute-long figure-skating scoring.

The deployment of deep neural networks in real-world applications is mostly restricted by their high inference costs. Extensive efforts have been made to improve the accuracy with expert-designed or algorithm-searched architectures. However, the incremental improvement is typically achieved with increasingly more expensive models that only a small portion of input instances really need. Inference with a static architecture that processes all input instances via the same transformation would thus incur unnecessary computational costs. Therefore, customizing the model capacity in an instance-aware manner is much needed for higher inference efficiency. In this paper, we propose Dynamic Routing Networks (DRNets), which support efficient instance-aware inference by routing the input instance to only necessary transformation branches selected from a candidate set of branches for each connection between transformation nodes. The branch selection is dynamically determined via the corresponding branch importance weights, which are first generated from lightweight hypernetworks (RouterNets) and then recalibrated with Gumbel-Softmax before the selection. Extensive experiments show that DRNets can reduce a substantial amount of parameter size and FLOPs during inference with prediction performance comparable to state-of-the-art architectures.