WACV 2023 Papers

We introduce visual hints expansion for guiding stereo matching to improve generalization. Our work is motivated by the robustness of Visual Inertial Odometry (VIO) in computer vision and robotics, where a sparse and unevenly distributed set of feature points characterizes a scene. To improve stereo matching, we propose to elevate 2D hints to 3D points. These sparse and unevenly distributed 3D visual hints are expanded using a 3D random geometric graph, which enhances the learning and inference process. We evaluate our proposal on multiple widely adopted benchmarks and show improved performance without access to additional sensors other than the image sequence. To highlight practical applicability and symbiosis with visual odometry, we demonstrate how our methods run on embedded hardware.

Spatial transcriptomics (ST) is essential for understanding diseases and developing novel treatments. It measures gene expression of each fine-grained area (i.e., different windows) in the tissue slide with low throughput. This paper proposes an Exemplar Guided Network (EGN) to accurately and efficiently predict gene expression directly from each window of a tissue slide image. We apply exemplar learning to dynamically boost gene expression prediction from nearest/similar exemplars of a given tissue slide image window. Our EGN framework composes of three main components: 1) an extractor to structure a representation space for unsupervised exemplar retrievals; 2) a vision transformer (ViT) backbone to progressively extract representations of the input window; and 3) an Exemplar Bridging (EB) block to adaptively revise the intermediate ViT representations by using the nearest exemplars. Finally, we complete the gene expression prediction task with a simple attention-based prediction block. Experiments on standard benchmark datasets indicate the superiority of our approach when comparing with the past state-of-the-art (SOTA) methods.

This paper proposes a self-supervised learned local detector and descriptor, called EventPoint, for event stream/camera tracking and registration. Event-based cameras have grown in popularity because of their biological inspiration and low power consumption. Despite this, applying local features directly to the event stream is difficult due to its peculiar data structure. We propose a new time-surface-like event stream representation method called Tencode. The event stream data processed by Tencode can obtain the pixel-level positioning of interest points while also simultaneously extracting descriptors through a neural network. Instead of using costly and unreliable manual annotation, our network leverages the prior knowledge of local feature extraction on color images and conducts self-supervised learning via homographic and spatio-temporal adaptation. To the best of our knowledge, our proposed method is the first research on event-based local features learning using a deep neural network. We provide comprehensive experiments of feature point detection and matching, and three public datasets are used for evaluation (i.e. DSEC, N-Caltech101, and HVGA ATIS Corner Dataset). The experimental findings demonstrate that our method outperforms SOTA in terms of feature point detection and description.

To understand our surrounding world, our brain is continuously inundated with multisensory information and their complex interactions coming from the outside world at any given moment. While processing this information might seem effortless for human brains, it is challenging to build a machine that can perform similar tasks since complex interactions cannot be dealt with a single type of integration but require more sophisticated approaches. In this paper, we propose a new simple method to address the multisensory integration in video understanding. Unlike previous works where a single fusion type is used, we design a multi-head model with individual event-specific layers to deal with different audio-visual relationships, enabling different ways of audio-visual fusion. Experimental results show that our event-specific layers can discover unique properties of the audio-visual relationships in the videos, e.g., semantically matched moments, and rhythmic events. Moreover, although our network is trained with single labels, our multi-head design can inherently output additional semantically meaningful multi-labels for a video. As an application, we demonstrate that our proposed method can expose the extent of event-characteristics of popular benchmark datasets.

Event-based cameras (ECs) are bio-inspired sensors that asynchronously report pixel brightness changes. Due to their high dynamic range, pixel bandwidth, temporal resolution, low power consumption, and computational simplicity, they are beneficial for vision-based projects in challenging lighting conditions and they can detect fast movements with their microsecond response time. The first generation of ECs are monochrome, but color data is very useful and sometimes essential for certain vision-based applications. The latest technology enables manufacturers to build color ECs, trading off the size of the sensor and substantially reducing the resolution compared to monochrome models, despite having the same bandwidth. In addition, ECs only detect changes in light and do not show static or slowly moving objects. We introduce a method to detect full RGB events using a monochrome EC aided by a structured light projector. The projector emits rapidly changing RGB patterns of light beams on the scene, the reflection of which is captured by the EC. We combine the benefits of ECs and projection-based techniques and allow depth and color detection of static or moving objects with a commercial TI LightCrafter 4500 projector and a monocular monochrome EC, paving the way for frameless RGB-D sensing applications. Our code is available publicly: github.com/MISTLab/event_based_rgbd_ros

We develop a measure for evaluating the performance of generative networks given two sets of images. A popular performance measure currently used to do this is the Frechet Inception Distance (FID). FID assumes that images featurized using the penultimate layer of Inception-v3 follow a Gaussian distribution, an assumption which cannot be violated if we wish to use FID as a metric. However, we show that Inception-v3 features of the ImageNet dataset are not Gaussian; in particular, every single marginal is not Gaussian. To remedy this problem, we model the featurized images using Gaussian mixture models (GMMs) and compute the 2-Wasserstein distance restricted to GMMs. We define a performance measure, which we call WaM, on two sets of images by using Inception-v3 (or another classifier) to featurize the images, estimate two GMMs, and use the restricted 2-Wasserstein distance to compare the GMMs. We experimentally show the advantages of WaM over FID, including how FID is more sensitive than WaM to imperceptible image perturbations. By modelling the non-Gaussian features obtained from Inception-v3 as GMMs and using a GMM metric, we can more accurately evaluate generative network performance.

We present Ev-NeRF, a Neural Radiance Field derived from event data. While event cameras can measure subtle brightness changes in high frame rates, the measurements in low lighting or extreme motion suffer from significant domain discrepancy with complex noise. As a result, the performance of event-based vision tasks does not transfer to challenging environments, where the event cameras are expected to thrive over normal cameras. We find that the multi-view consistency of NeRF provides a powerful self-supervision signal for eliminating spurious measurements and extracting the consistent underlying structure despite highly noisy input. Instead of posed images of the original NeRF, the input to Ev-NeRF is the event measurements accompanied by the movements of the sensors. Using the loss function that reflects the measurement model of the sensor, Ev-NeRF creates an integrated neural volume that summarizes the unstructured and sparse data points captured for about 2-4 seconds. The generated neural volume can also produce intensity images from novel views with reasonable depth estimates, which can serve as a high-quality input to various vision-based tasks. Our results show that Ev-NeRF achieves competitive performance for intensity image reconstruction under extreme noise conditions and high-dynamic-range imaging.

Recent transformer-based super-resolution (SR) methods have achieved promising results against conventional CNN-based methods. However, these approaches suffer from essential shortsightedness created by only utilizing the standard self-attention-based reasoning. In this paper, we introduce an effective hybrid SR network to aggregate enriched features, including local features from CNNs and long-range multi-scale dependencies captured by transformers. Specifically, our network comprises transformer and convolutional branches, which synergetically complement each representation during the restoration procedure. Furthermore, we propose a cross-scale token attention module, allowing the transformer branch to exploit the informative relationships among tokens across different scales efficiently. Our proposed method achieves state-of-the-art SR results on numerous benchmark datasets.

We propose a simple yet effective algorithm for motion-based video frame interpolation. Existing motion-based interpolation methods typically rely on an off-the-shelf optical flow model or a U-Net based pyramid network for motion estimation, which either suffer from large model size or limited capacity in handling various challenging motion cases. In this work, we present a novel compact model to simultaneously estimate the bi-directional motions between input frames. It is designed by carefully adapting the ingredients (e.g., warping, correlation) in optical flow research for simultaneous bi-directional motion estimation within a flexible pyramid recurrent framework. Our motion estimator is extremely lightweight (15x smaller than PWC-Net), yet enables reliable handling of large and complex motion cases. Based on estimated bi-directional motions, we employ a synthesis network to fuse forward-warped representations and predict the intermediate frame. Our method achieves excellent performance on a broad range of frame interpolation benchmarks. Code and trained models are available at: https://github.com/srcn-ivl/EBME.

This paper considers photography of high dynamic range scenes containing mixtures of shadows and highlights on mobile phones. Multi-frame merging constructs a high-quality image at the cost of capturing multiple frames of the same scene. Contrarily, end-to-end optimized image signal processing (E2EISP) produces an enhanced image from a single-frame Bayer array. This paper combines the merits of the two approaches by using labels of high-quality multi-frame merged images to train E2EISP with a novel neural network architecture composed of a multi-head mixture of brightness enhancement for accurately processing shadows/highlights and a multi-head mixture of image processing featured camera settings of white balance and color correction for a proper color generation. We also proposed a combination of supervised, unsupervised, and generative adversarial losses for brightness, edge, and detail enhancement. Experimental results show that the proposed single-frame ISP produces enhanced images and outperforms state-of-the-art methods.

We focus on controllable disentangled representation learning (C-Dis-RL), where users can control the partition of the disentangled latent space to factorize dataset attributes (concepts) for downstream tasks. Two general problems remain under-explored in current methods: (1) They lack comprehensive disentanglement constraints, especially missing the minimization of mutual information between different attributes across latent and observation domains. (2) They lack convexity constraints, which is important for meaningfully manipulating specific attributes for downstream tasks. To encourage both comprehensive C-Dis-RL and convexity simultaneously, we propose a simple yet efficient method: Controllable Interpolation Regularization (CIR), which creates a positive loop where disentanglement and convexity can help each other. Specifically, we conduct controlled interpolation in latent space during training, and we reuse the encoder to help form a 'perfect disentanglement' regularization. In that case, (a) disentanglement loss implicitly enlarges the potential understandable distribution to encourage convexity; (b) convexity can in turn improve robust and precise disentanglement. CIR is a general module and we merge CIR with three different algorithms: ELEGANT, I2I-Dis, and GZS-Net to show the compatibility and effectiveness. Qualitative and quantitative experiments show improvement in C-Dis-RL and latent convexity by CIR. This further improves downstream tasks: controllable image synthesis, cross-modality image translation and zero-shot synthesis.

Current computer vision (CV) systems use an image signal processing (ISP) unit to convert the high resolution raw images captured by image sensors to visually pleasing RGB images. Typically, CV models are trained on these RGB images and have yielded state-of-the-art (SOTA) performance on a wide range of complex vision tasks, such as object detection. In addition, in order to deploy these models on resource-constrained low-power devices, recent works have proposed in-sensor and in-pixel computing approaches that try to partly/fully bypass the ISP and yield significant bandwidth reduction between the image sensor and the CV processing unit by downsampling the activation maps in the initial convolutional neural network (CNN) layers. However, direct inference on the raw images degrades the test accuracy due to the difference in covariance of the raw images captured by the image sensors compared to the ISP-processed images used for training. Moreover, it is difficult to train deep CV models on raw images, because most (if not all) large-scale open-source datasets consist of RGB images. To mitigate this concern, we propose to invert the ISP pipeline, which can convert the RGB images of any dataset to its raw counterparts, and enable model training on raw images. We release the raw version of the COCO dataset, a large-scale benchmark for generic high-level vision tasks. For ISP-less CV systems, training on these raw images result in a 7.1% increase in test accuracy on the visual wake works (VWW) dataset compared to relying on training with traditional ISP-processed RGB datasets. To further improve the accuracy of ISP-less CV models and to increase the energy and bandwidth benefits obtained by in-sensor/in-pixel computing, we propose an energy-efficient form of analog in-pixel demosaicing that may be coupled with in-pixel CNN computations. When evaluated on raw images captured by real sensors from the PASCALRAW dataset, our approach results in a 8.1% increase in mAP.

For autonomous vehicles and mobile robots to safely operate in the real world, i.e., the wild, scene understanding models should perform well in the many different scenarios that can be encountered. In reality, these scenarios are not all represented in the model's training data, leading to poor performance. To tackle this, current training strategies attempt to either exploit additional unlabeled data with unsupervised domain adaptation (UDA), or to reduce overfitting using the limited available labeled data with domain generalization (DG). However, it is not clear from current literature which of these methods allows for better generalization to unseen data from the wild. Therefore, in this work, we present an evaluation framework in which the generalization capabilities of state-of-the-art UDA and DG methods can be compared fairly. From this evaluation, we find that UDA methods, which leverage unlabeled data, outperform DG methods in terms of generalization, and can deliver similar performance on unseen data as fully-supervised training methods that require all data to be labeled. We show that semantic segmentation performance can be increased up to 30% for a priori unknown data without using any extra labeled data.

The timing of cell divisions in early embryos during the In-Vitro Fertilization (IVF) process is a key predictor of embryo viability. However, observing cell divisions in Time-Lapse Monitoring (TLM) is a time-consuming process and highly depends on experts. In this paper, we propose EmbryosFormer, a computational model to automatically detect and classify cell divisions from original time-lapse images. Our proposed network is designed as an encoder-decoder deformable transformer with collaborative heads. The transformer contracting path predicts per-image labels and is optimized by a classification head. The transformer expanding path models the temporal coherency between embryo images to ensure monotonic non-decreasing constraint and is optimized by a segmentation head. Both contracting and expanding paths are synergetically learned by a collaboration head. We have benchmarked our proposed EmbryosFormer on two datasets: a public dataset with mouse embryos with 8-cell stage and an in-house dataset with human embryos with 4-cell stage. Source code: https://github.com/UARK-AICV/Embryos.

One of the most relevant tasks for augmented and virtual reality applications is the interaction of virtual objects with real humans which requires accurate 3D human pose predictions. Obtaining accurate 3D human poses requires careful camera calibration which is difficult for non-technical personal or in a pop-up scenario. Recent markerless motion capture approaches require accurate camera calibration at least for the final triangulation step. Instead, we solve this problem by presenting ElliPose, Stereoscopic 3D Human Pose Estimation by Fitting Ellipsoids, where we jointly estimate the 3D human as well as the camera pose. We exploit the fact that bones do not change in length over the course of a sequence and thus their relative trajectories have to lie on the surface of a sphere which we can utilize to iteratively correct the camera and 3D pose estimation. As another use-case we demonstrate that our approach can be used as replacement for ground-truth 3D poses to train monocular 3D pose estimators. We show that our method produces competitive results even when comparing with state-of-the-art methods that use more cameras or ground-truth camera extrinsics.

Few-shot segmentation aims to devise a generalizing model that segments query images from unseen classes during training with the guidance of a few support images whose class tally with the class of the query. There exist two domain-specific problems mentioned in the previous works, namely spatial inconsistency and bias towards seen classes. Taking the former problem into account, our method compares the support feature map with the query feature map at multi scales to become scale-agnostic. As a solution to the latter problem, a supervised model, called as base learner, is trained on available classes to accurately identify pixels belonging to seen classes. Hence, subsequent meta learner has a chance to discard areas belonging to seen classes with the help of an ensemble learning model that coordinates meta learner with the base learner. We simultaneously address these two vital problems for the first time and achieve state-of-the-art performances on both PASCAL-5i and COCO-20i datasets.

In recent years, many automobiles have been equipped with cameras, which have accumulated an enormous amount of video footage of driving scenes. Autonomous driving demands the highest level of safety, for which even unimaginably rare driving scenes have to be collected in training data to improve the recognition accuracy for specific scenes. However, it is prohibitively costly to find very few specific scenes from an enormous amount of videos. In this article, we show that proper video-to-video distances can be defined by focusing on ego-vehicle actions. It is well known that existing methods based on supervised learning cannot handle videos that do not fall into predefined classes, though they work well in defining video-to-video distances in the embedding space between labeled videos. To tackle this problem, we propose a method based on semi-supervised contrastive learning. We consider two related but distinct contrastive learning: standard graph contrastive learning and our proposed SOIA-based contrastive learning. We observe that the latter approach can provide more sensible video-to-video distances between unlabeled videos. Next, the effectiveness of our method is quantified by evaluating the classification performance of the ego-vehicle action recognition using HDD dataset, which shows that our method including unlabeled data in training significantly outperforms the existing methods using only labeled data in training.

Camera-based physiological measurement is a growing field with neural models providing state-of-the-art performance. Prior research has explored various end-to-end architectures; however these methods still require several preprocessing steps and are not able to run directly on mobile and edge devices. The operations are often non-trivial to implement, making replication and deployment difficult and can even have a higher computational budget than the core network itself. In this paper, we propose two novel and efficient neural models for camera-based physiological measurement called EfficientPhys that remove the need for face detection, segmentation, normalization, color space transformation or any other preprocessing steps. Using an input of raw video frames, our models achieve strong accuracy on three public datasets. We show that this is the case whether using a transformer or convolutional backbone. We further evaluate the latency of the proposed networks and show that our most lightweight network also achieves a 33% improvement in efficiency.

The design of more complex and powerful neural network models has significantly advanced the state-of-the-art in visual object tracking. These advances can be attributed to deeper networks, or the introduction of new building blocks, such as transformers. However, in the pursuit of increased tracking performance, runtime is often hindered. Furthermore, efficient tracking architectures have received surprisingly little attention. In this paper, we introduce the Exemplar Transformer, a transformer module utilizing a single instance level attention layer for realtime visual object tracking. E.T.Track, our visual tracker that incorporates Exemplar Transformer modules, runs at 47 FPS on a CPU. This is up to 8x faster than other transformer-based models. When compared to lightweight trackers that can operate in realtime on standard CPUs, E.T.Track consistently outperforms all other methods on the LaSOT, OTB-100, NFS, TrackingNet, and VOT-ST2020 datasets. Code and models are available at https://github.com/pblatter/ettrack.