Computer Vision and Pattern Recognition (cs.CV)

Abstract: Since the beginning of world-wide COVID-19 pandemic, facial masks have been recommended to limit the spread of the disease. However, these masks hide certain facial attributes. Hence, it has become difficult for existing face recognition systems to perform identity verification on masked faces. In this context, it is necessary to develop masked Face Recognition (MFR) for contactless biometric recognition systems. Thus, in this paper, we propose Complementary Attention Learning and Multi-Focal Spatial Attention that precisely removes masked region by training complementary spatial attention to focus on two distinct regions: masked regions and backgrounds. In our method, standard spatial attention and networks focus on unmasked regions, and extract mask-invariant features while minimizing the loss of the conventional Face Recognition (FR) performance. For conventional FR, we evaluate the performance on the IJB-C, Age-DB, CALFW, and CPLFW datasets. We evaluate the MFR performance on the ICCV2021-MFR/Insightface track, and demonstrate the improved performance on the both MFR and FR datasets. Additionally, we empirically verify that spatial attention of proposed method is more precisely activated in unmasked regions.

Abstract: Object recognition and instance segmentation are fundamental skills in any robotic or autonomous system. Existing state-of-the-art methods are often unable to capture meaningful uncertainty in challenging or ambiguous scenes, and as such can cause critical errors in high-performance applications. In this paper, we explore a class of distributional instance segmentation models using latent codes that can model uncertainty over plausible hypotheses of object masks. For robotic picking applications, we propose a confidence mask method to achieve the high precision necessary in industrial use cases. We show that our method can significantly reduce critical errors in robotic systems, including our newly released dataset of ambiguous scenes in a robotic application. On a real-world apparel-picking robot, our method significantly reduces double pick errors while maintaining high performance.

Abstract: While the community of 3D point cloud generation has witnessed a big growth in recent years, there still lacks an effective way to enable intuitive user control in the generation process, hence limiting the general utility of such methods. Since an intuitive way of decomposing a shape is through its parts, we propose to tackle the task of controllable part-based point cloud generation. We introduce DiffFacto, a novel probabilistic generative model that learns the distribution of shapes with part-level control. We propose a factorization that models independent part style and part configuration distributions, and present a novel cross diffusion network that enables us to generate coherent and plausible shapes under our proposed factorization. Experiments show that our method is able to generate novel shapes with multiple axes of control. It achieves state-of-the-art part-level generation quality and generates plausible and coherent shape, while enabling various downstream editing applications such as shape interpolation, mixing and transformation editing. Code will be made publicly available.

Abstract: In this paper, we propose a new visual reasoning task, called Visual Transformation Telling (VTT). This task requires a machine to describe the transformation that occurred between every two adjacent states (i.e. images) in a series. Unlike most existing visual reasoning tasks that focus on state reasoning, VTT emphasizes transformation reasoning. We collected 13,547 samples from two instructional video datasets, CrossTask and COIN, and extracted desired states and transformation descriptions to create a suitable VTT benchmark dataset. Humans can naturally reason from superficial states differences (e.g. ground wetness) to transformations descriptions (e.g. raining) according to their life experience but how to model this process to bridge this semantic gap is challenging. We designed TTNet on top of existing visual storytelling models by enhancing the model's state-difference sensitivity and transformation-context awareness. TTNet significantly outperforms other baseline models adapted from similar tasks, such as visual storytelling and dense video captioning, demonstrating the effectiveness of our modeling on transformations. Through comprehensive diagnostic analyses, we found TTNet has strong context utilization abilities, but even with some state-of-the-art techniques such as CLIP, there remain challenges in generalization that need to be further explored.

Abstract: Automated detection of contraband items in X-ray images can significantly increase public safety, by enhancing the productivity and alleviating the mental load of security officers in airports, subways, customs/post offices, etc. The large volume and high throughput of passengers, mailed parcels, etc., during rush hours make it a Big Data analysis task. Modern computer vision algorithms relying on Deep Neural Networks (DNNs) have proven capable of undertaking this task even under resource-constrained and embedded execution scenarios, e.g., as is the case with fast, single-stage, anchor-based object detectors. This paper proposes a two-fold improvement of such algorithms for the X-ray analysis domain, introducing two complementary novelties. Firstly, more efficient anchors are obtained by hierarchical clustering the sizes of the ground-truth training set bounding boxes; thus, the resulting anchors follow a natural hierarchy aligned with the semantic structure of the data. Secondly, the default Non-Maximum Suppression (NMS) algorithm at the end of the object detection pipeline is modified to better handle occluded object detection and to reduce the number of false predictions, by inserting the Efficient Intersection over Union (E-IoU) metric into the Weighted Cluster NMS method. E-IoU provides more discriminative geometrical correlations between the candidate bounding boxes/Regions-of-Interest (RoIs). The proposed method is implemented on a common single-stage object detector (YOLOv5) and its experimental evaluation on a relevant public dataset indicates significant accuracy gains over both the baseline and competing approaches. This highlights the potential of Big Data analysis in enhancing public safety.

Abstract: The paper describes a dataset that was collected by infrared thermography, which is a non-contact, non-intrusive technique to collect data and analyze the built environment in various aspects. While most studies focus on the city and building scales, the rooftop observatory provides high temporal and spatial resolution observations with dynamic interactions on the district scale. The rooftop infrared thermography observatory with a multi-modal platform that is capable of assessing a wide range of dynamic processes in urban systems was deployed in Singapore. It was placed on the top of two buildings that overlook the outdoor context of the campus of the National University of Singapore. The platform collects remote sensing data from tropical areas on a temporal scale, allowing users to determine the temperature trend of individual features such as buildings, roads, and vegetation. The dataset includes 1,365,921 thermal images collected on average at approximately 10 seconds intervals from two locations during ten months.

Abstract: Most deepfake detection methods focus on detecting spatial and/or spatio-temporal changes in facial attributes. This is because available benchmark datasets contain mostly visual-only modifications. However, a sophisticated deepfake may include small segments of audio or audio-visual manipulations that can completely change the meaning of the content. To addresses this gap, we propose and benchmark a new dataset, Localized Audio Visual DeepFake (LAV-DF), consisting of strategic content-driven audio, visual and audio-visual manipulations. The proposed baseline method, Boundary Aware Temporal Forgery Detection (BA-TFD), is a 3D Convolutional Neural Network-based architecture which efficiently captures multimodal manipulations. We further improve (i.e. BA-TFD+) the baseline method by replacing the backbone with a Multiscale Vision Transformer and guide the training process with contrastive, frame classification, boundary matching and multimodal boundary matching loss functions. The quantitative analysis demonstrates the superiority of BA- TFD+ on temporal forgery localization and deepfake detection tasks using several benchmark datasets including our newly proposed dataset. The dataset, models and code are available at https://github.com/ControlNet/LAV-DF.

Abstract: Existing datasets for autonomous driving (AD) often lack diversity and long-range capabilities, focusing instead on 360{\deg} perception and temporal reasoning. To address this gap, we introduce Zenseact Open Dataset (ZOD), a large-scale and diverse multimodal dataset collected over two years in various European countries, covering an area 9x that of existing datasets. ZOD boasts the highest range and resolution sensors among comparable datasets, coupled with detailed keyframe annotations for 2D and 3D objects (up to 245m), road instance/semantic segmentation, traffic sign recognition, and road classification. We believe that this unique combination will facilitate breakthroughs in long-range perception and multi-task learning. The dataset is composed of Frames, Sequences, and Drives, designed to encompass both data diversity and support for spatio-temporal learning, sensor fusion, localization, and mapping. Frames consist of 100k curated camera images with two seconds of other supporting sensor data, while the 1473 Sequences and 29 Drives include the entire sensor suite for 20 seconds and a few minutes, respectively. ZOD is the only large-scale AD dataset released under a permissive license, allowing for both research and commercial use. The dataset is accompanied by an extensive development kit. Data and more information are available online (https://zod.zenseact.com).

Abstract: Three-dimensional (3-D) synthetic aperture radar (SAR) is widely used in many security and industrial applications requiring high-resolution imaging of concealed or occluded objects. The ability to resolve intricate 3-D targets is essential to the performance of such applications and depends directly on system bandwidth. However, because high-bandwidth systems face several prohibitive hurdles, an alternative solution is to operate multiple radars at distinct frequency bands and fuse the multiband signals. Current multiband signal fusion methods assume a simple target model and a small number of point reflectors, which is invalid for realistic security screening and industrial imaging scenarios wherein the target model effectively consists of a large number of reflectors. To the best of our knowledge, this study presents the first use of deep learning for multiband signal fusion. The proposed network, called kR-Net, employs a hybrid, dual-domain complex-valued convolutional neural network (CV-CNN) to fuse multiband signals and impute the missing samples in the frequency gaps between subbands. By exploiting the relationships in both the wavenumber domain and wavenumber spectral domain, the proposed framework overcomes the drawbacks of existing multiband imaging techniques for realistic scenarios at a fraction of the computation time of existing multiband fusion algorithms. Our method achieves high-resolution imaging of intricate targets previously impossible using conventional techniques and enables finer resolution capacity for concealed weapon detection and occluded object classification using multiband signaling without requiring more advanced hardware. Furthermore, a fully integrated multiband imaging system is developed using commercially available millimeter-wave (mmWave) radars for efficient multiband imaging.

Abstract: Many important computer vision tasks are naturally formulated to have a non-differentiable objective. Therefore, the standard, dominant training procedure of a neural network is not applicable since back-propagation requires the gradients of the objective with respect to the output of the model. Most deep learning methods side-step the problem sub-optimally by using a proxy loss for training, which was originally designed for another task and is not tailored to the specifics of the objective. The proxy loss functions may or may not align well with the original non-differentiable objective. An appropriate proxy has to be designed for a novel task, which may not be feasible for a non-specialist. This thesis makes four main contributions toward bridging the gap between the non-differentiable objective and the training loss function. Throughout the thesis, we refer to a loss function as a surrogate loss if it is a differentiable approximation of the non-differentiable objective. Note that we use the terms objective and evaluation metric interchangeably. The contributions of this thesis make the training of neural networks more scalable -- to new tasks in a nearly labor-free manner when the evaluation metric is decomposable, which will help researchers with novel tasks. For non-decomposable evaluation metrics, the differentiable components developed for the recall@k surrogate, such as sorting and counting, can also be used for creating new surrogates.

Abstract: Classification of gigapixel Whole Slide Images (WSIs) is an important prediction task in the emerging area of computational pathology. There has been a surge of research in deep learning models for WSI classification with clinical applications such as cancer detection or prediction of molecular mutations from WSIs. Most methods require expensive and labor-intensive manual annotations by expert pathologists. Weakly supervised Multiple Instance Learning (MIL) methods have recently demonstrated excellent performance; however, they still require large slide-level labeled training datasets that need a careful inspection of each slide by an expert pathologist. In this work, we propose a fully unsupervised WSI classification algorithm based on mutual transformer learning. Instances from gigapixel WSI (i.e., image patches) are transformed into a latent space and then inverse-transformed to the original space. Using the transformation loss, pseudo-labels are generated and cleaned using a transformer label-cleaner. The proposed transformer-based pseudo-label generation and cleaning modules mutually train each other iteratively in an unsupervised manner. A discriminative learning mechanism is introduced to improve normal versus cancerous instance labeling. In addition to unsupervised classification, we demonstrate the effectiveness of the proposed framework for weak supervision for cancer subtype classification as downstream analysis. Extensive experiments on four publicly available datasets show excellent performance compared to the state-of-the-art methods. We intend to make the source code of our algorithm publicly available soon.

Abstract: The success of the Segment Anything Model (SAM) demonstrates the significance of data-centric machine learning. However, due to the difficulties and high costs associated with annotating Remote Sensing (RS) images, a large amount of valuable RS data remains unlabeled, particularly at the pixel level. In this study, we leverage SAM and existing RS object detection datasets to develop an efficient pipeline for generating a large-scale RS segmentation dataset, dubbed SAMRS. SAMRS surpasses existing high-resolution RS segmentation datasets in size by several orders of magnitude, and provides object category, location, and instance information that can be used for semantic segmentation, instance segmentation, and object detection, either individually or in combination. We also provide a comprehensive analysis of SAMRS from various aspects. We hope it could facilitate research in RS segmentation, particularly in large model pre-training.

Abstract: In this paper, we investigate novel data collection and training techniques towards improving classification accuracy of non-moving (static) hand gestures using a convolutional neural network (CNN) and frequency-modulated-continuous-wave (FMCW) millimeter-wave (mmWave) radars. Recently, non-contact hand pose and static gesture recognition have received considerable attention in many applications ranging from human-computer interaction (HCI), augmented/virtual reality (AR/VR), and even therapeutic range of motion for medical applications. While most current solutions rely on optical or depth cameras, these methods require ideal lighting and temperature conditions. mmWave radar devices have recently emerged as a promising alternative offering low-cost system-on-chip sensors whose output signals contain precise spatial information even in non-ideal imaging conditions. Additionally, deep convolutional neural networks have been employed extensively in image recognition by learning both feature extraction and classification simultaneously. However, little work has been done towards static gesture recognition using mmWave radars and CNNs due to the difficulty involved in extracting meaningful features from the radar return signal, and the results are inferior compared with dynamic gesture classification. This article presents an efficient data collection approach and a novel technique for deep CNN training by introducing ``sterile'' images which aid in distinguishing distinct features among the static gestures and subsequently improve the classification accuracy. Applying the proposed data collection and training methods yields an increase in classification rate of static hand gestures from $85\%$ to $93\%$ and $90\%$ to $95\%$ for range and range-angle profiles, respectively.

Abstract: In this paper, we develop a novel super-resolution algorithm for near-field synthetic-aperture radar (SAR) under irregular scanning geometries. As fifth-generation (5G) millimeter-wave (mmWave) devices are becoming increasingly affordable and available, high-resolution SAR imaging is feasible for end-user applications and non-laboratory environments. Emerging applications such freehand imaging, wherein a handheld radar is scanned throughout space by a user, unmanned aerial vehicle (UAV) imaging, and automotive SAR face several unique challenges for high-resolution imaging. First, recovering a SAR image requires knowledge of the array positions throughout the scan. While recent work has introduced camera-based positioning systems capable of adequately estimating the position, recovering the algorithm efficiently is a requirement to enable edge and Internet of Things (IoT) technologies. Efficient algorithms for non-cooperative near-field SAR sampling have been explored in recent work, but suffer image defocusing under position estimation error and can only produce medium-fidelity images. In this paper, we introduce a mobile-friend vision transformer (ViT) architecture to address position estimation error and perform SAR image super-resolution (SR) under irregular sampling geometries. The proposed algorithm, Mobile-SRViT, is the first to employ a ViT approach for SAR image enhancement and is validated in simulation and via empirical studies.

Abstract: Due to its light and weather-independent sensing, millimeter-wave (MMW) radar is essential in smart environments. Intelligent vehicle systems and industry-grade MMW radars have integrated such capabilities. Industry-grade MMW radars are expensive and hard to get for community-purpose smart environment applications. However, commercially available MMW radars have hidden underpinning challenges that need to be investigated for tasks like recognizing objects and activities, real-time person tracking, object localization, etc. Image and video data are straightforward to gather, understand, and annotate for such jobs. Image and video data are light and weather-dependent, susceptible to the occlusion effect, and present privacy problems. To eliminate dependence and ensure privacy, commercial MMW radars should be tested. MMW radar's practicality and performance in varied operating settings must be addressed before promoting it. To address the problems, we collected a dataset using Texas Instruments' Automotive mmWave Radar (AWR2944) and reported the best experimental settings for object recognition performance using different deep learning algorithms. Our extensive data gathering technique allows us to systematically explore and identify object identification task problems under cross-ambience conditions. We investigated several solutions and published detailed experimental data.

Abstract: Representation learning of Satellite Image Time Series (SITS) presents its unique challenges, such as prohibitive computation burden caused by high spatiotemporal resolutions, irregular acquisition times, and complex spatiotemporal interactions, leading to highly-specialized neural network architectures for SITS analysis. Despite the promising results achieved by some pioneering work, we argue that satisfactory representation learning paradigms have not yet been established for SITS analysis, causing an isolated island where transferring successful paradigms or the latest advances from Computer Vision (CV) to SITS is arduous. In this paper, we develop a unique perspective of SITS processing as a direct set prediction problem, inspired by the recent trend in adopting query-based transformer decoders to streamline the object detection or image segmentation pipeline, and further propose to decompose the representation learning process of SITS into three explicit steps: collect--update--distribute, which is computationally efficient and suits for irregularly-sampled and asynchronous temporal observations. Facilitated by the unique reformulation and effective feature extraction framework proposed, our models pre-trained on pixel-set format input and then fine-tuned on downstream dense prediction tasks by simply appending a commonly-used segmentation network have attained new state-of-the-art (SoTA) results on PASTIS dataset compared to bespoke neural architectures such as U-TAE. Furthermore, the clear separation, conceptually and practically, between temporal and spatial components in the panoptic segmentation pipeline of SITS allows us to leverage the recent advances in CV, such as Mask2Former, a universal segmentation architecture, resulting in a noticeable 8.8 points increase in PQ compared to the best score reported so far.

Abstract: In this paper, we introduce an innovative super resolution approach to emerging modes of near-field synthetic aperture radar (SAR) imaging. Recent research extends convolutional neural network (CNN) architectures from the optical to the electromagnetic domain to achieve super resolution on images generated from radar signaling. Specifically, near-field synthetic aperture radar (SAR) imaging, a method for generating high-resolution images by scanning a radar across space to create a synthetic aperture, is of interest due to its high-fidelity spatial sensing capability, low cost devices, and large application space. Since SAR imaging requires large aperture sizes to achieve high resolution, super-resolution algorithms are valuable for many applications. Freehand smartphone SAR, an emerging sensing modality, requires irregular SAR apertures in the near-field and computation on mobile devices. Achieving efficient high-resolution SAR images from irregularly sampled data collected by freehand motion of a smartphone is a challenging task. In this paper, we propose a novel CNN architecture to achieve SAR image super-resolution for mobile applications by employing state-of-the-art SAR processing and deep learning techniques. The proposed algorithm is verified via simulation and an empirical study. Our algorithm demonstrates high-efficiency and high-resolution radar imaging for near-field scenarios with irregular scanning geometries.

Abstract: Birds are important indicators for monitoring both biodiversity and habitat health; they also play a crucial role in ecosystem management. Decline in bird populations can result in reduced eco-system services, including seed dispersal, pollination and pest control. Accurate and long-term monitoring of birds to identify species of concern while measuring the success of conservation interventions is essential for ecologists. However, monitoring is time consuming, costly and often difficult to manage over long durations and at meaningfully large spatial scales. Technology such as camera traps, acoustic monitors and drones provide methods for non-invasive monitoring. There are two main problems with using camera traps for monitoring: a) cameras generate many images, making it difficult to process and analyse the data in a timely manner; and b) the high proportion of false positives hinders the processing and analysis for reporting. In this paper, we outline an approach for overcoming these issues by utilising deep learning for real-time classi-fication of bird species and automated removal of false positives in camera trap data. Images are classified in real-time using a Faster-RCNN architecture. Images are transmitted over 3/4G cam-eras and processed using Graphical Processing Units (GPUs) to provide conservationists with key detection metrics therefore removing the requirement for manual observations. Our models achieved an average sensitivity of 88.79%, a specificity of 98.16% and accuracy of 96.71%. This demonstrates the effectiveness of using deep learning for automatic bird monitoring.

Abstract: Emerged as a biology-inspired method, Spiking Neural Networks (SNNs) mimic the spiking nature of brain neurons and have received lots of research attention. SNNs deal with binary spikes as their activation and therefore derive extreme energy efficiency on hardware. However, it also leads to an intrinsic obstacle that training SNNs from scratch requires a re-definition of the firing function for computing gradient. Artificial Neural Networks (ANNs), however, are fully differentiable to be trained with gradient descent. In this paper, we propose a joint training framework of ANN and SNN, in which the ANN can guide the SNN's optimization. This joint framework contains two parts: First, the knowledge inside ANN is distilled to SNN by using multiple branches from the networks. Second, we restrict the parameters of ANN and SNN, where they share partial parameters and learn different singular weights. Extensive experiments over several widely used network structures show that our method consistently outperforms many other state-of-the-art training methods. For example, on the CIFAR100 classification task, the spiking ResNet-18 model trained by our method can reach to 77.39% top-1 accuracy with only 4 time steps.

Abstract: It is challenging to remove rain-steaks from a single rainy image because the rain steaks are spatially varying in the rainy image. Although the CNN based methods have reported promising performance recently, there are still some defects, such as data dependency and insufficient interpretation. A single image deraining algorithm based on the combination of data-driven and model-based approaches is proposed. Firstly, an improved weighted guided image filter (iWGIF) is used to extract high-frequency information and learn the rain steaks to avoid interference from other information through the input image. Then, transfering the input image and rain steaks from the image domain to the feature domain adaptively to learn useful features for high-quality image deraining. Finally, networks with attention mechanisms is used to restore high-quality images from the latent features. Experiments show that the proposed algorithm significantly outperforms state-of-the-art methods in terms of both qualitative and quantitative measures.

Abstract: Vision in adverse weather conditions, whether it be snow, rain, or fog is challenging. In these scenarios, scattering and attenuation severly degrades image quality. Handling such inclement weather conditions, however, is essential to operate autonomous vehicles, drones and robotic applications where human performance is impeded the most. A large body of work explores removing weather-induced image degradations with dehazing methods. Most methods rely on single images as input and struggle to generalize from synthetic fully-supervised training approaches or to generate high fidelity results from unpaired real-world datasets. With data as bottleneck and most of today's training data relying on good weather conditions with inclement weather as outlier, we rely on an inverse rendering approach to reconstruct the scene content. We introduce ScatterNeRF, a neural rendering method which adequately renders foggy scenes and decomposes the fog-free background from the participating media-exploiting the multiple views from a short automotive sequence without the need for a large training data corpus. Instead, the rendering approach is optimized on the multi-view scene itself, which can be typically captured by an autonomous vehicle, robot or drone during operation. Specifically, we propose a disentangled representation for the scattering volume and the scene objects, and learn the scene reconstruction with physics-inspired losses. We validate our method by capturing multi-view In-the-Wild data and controlled captures in a large-scale fog chamber.

Abstract: We propose two deep learning models that fully automate shape parameterization for aerodynamic shape optimization. Both models are optimized to parameterize via deep geometric learning to embed human prior knowledge into learned geometric patterns, eliminating the need for further handcrafting. The Latent Space Model (LSM) learns a low-dimensional latent representation of an object from a dataset of various geometries, while the Direct Mapping Model (DMM) builds parameterization on the fly using only one geometry of interest. We also devise a novel regularization loss that efficiently integrates volumetric mesh deformation into the parameterization model. The models directly manipulate the high-dimensional mesh data by moving vertices. LSM and DMM are fully differentiable, enabling gradient-based, end-to-end pipeline design and plug-and-play deployment of surrogate models or adjoint solvers. We perform shape optimization experiments on 2D airfoils and discuss the applicable scenarios for the two models.

Abstract: We propose a real-time and lightweight single-image super-resolution (SR) network named Bicubic++. Despite using spatial dimensions of the input image across the whole network, Bicubic++ first learns quick reversible downgraded and lower resolution features of the image in order to decrease the number of computations. We also construct a training pipeline, where we apply an end-to-end global structured pruning of convolutional layers without using metrics like magnitude and gradient norms, and focus on optimizing the pruned network's PSNR on the validation set. Furthermore, we have experimentally shown that the bias terms take considerable amount of the runtime while increasing PSNR marginally, hence we have also applied bias removal to the convolutional layers. Our method adds ~1dB on Bicubic upscaling PSNR for all tested SR datasets and runs with ~1.17ms on RTX3090 and ~2.9ms on RTX3070, for 720p inputs and 4K outputs, both in FP16 precision. Bicubic++ won NTIRE 2023 RTSR Track 2 x3 SR competition and is the fastest among all competitive methods. Being almost as fast as the standard Bicubic upsampling method, we believe that Bicubic++ can set a new industry standard.

Abstract: In recent years, the increasing availability of personal data has raised concerns regarding privacy and security. One of the critical processes to address these concerns is data anonymization, which aims to protect individual privacy and prevent the release of sensitive information. This research focuses on the importance of face anonymization. Therefore, we introduce GANonymization, a novel face anonymization framework with facial expression-preserving abilities. Our approach is based on a high-level representation of a face which is synthesized into an anonymized version based on a generative adversarial network (GAN). The effectiveness of the approach was assessed by evaluating its performance in removing identifiable facial attributes to increase the anonymity of the given individual face. Additionally, the performance of preserving facial expressions was evaluated on several affect recognition datasets and outperformed the state-of-the-art method in most categories. Finally, our approach was analyzed for its ability to remove various facial traits, such as jewelry, hair color, and multiple others. Here, it demonstrated reliable performance in removing these attributes. Our results suggest that GANonymization is a promising approach for anonymizing faces while preserving facial expressions.

Abstract: Progressive compression allows images to start loading as low-resolution versions, becoming clearer as more data is received. This increases user experience when, for example, network connections are slow. Today, most approaches for image compression, both classical and learned ones, are designed to be non-progressive. This paper introduces ProgDTD, a training method that transforms learned, non-progressive image compression approaches into progressive ones. The design of ProgDTD is based on the observation that the information stored within the bottleneck of a compression model commonly varies in importance. To create a progressive compression model, ProgDTD modifies the training steps to enforce the model to store the data in the bottleneck sorted by priority. We achieve progressive compression by transmitting the data in order of its sorted index. ProgDTD is designed for CNN-based learned image compression models, does not need additional parameters, and has a customizable range of progressiveness. For evaluation, we apply ProgDTDto the hyperprior model, one of the most common structures in learned image compression. Our experimental results show that ProgDTD performs comparably to its non-progressive counterparts and other state-of-the-art progressive models in terms of MS-SSIM and accuracy.

Abstract: We propose to Transform Scene Graphs (TSG) into more descriptive captions. In TSG, we apply multi-head attention (MHA) to design the Graph Neural Network (GNN) for embedding scene graphs. After embedding, different graph embeddings contain diverse specific knowledge for generating the words with different part-of-speech, e.g., object/attribute embedding is good for generating nouns/adjectives. Motivated by this, we design a Mixture-of-Expert (MOE)-based decoder, where each expert is built on MHA, for discriminating the graph embeddings to generate different kinds of words. Since both the encoder and decoder are built based on the MHA, as a result, we construct a homogeneous encoder-decoder unlike the previous heterogeneous ones which usually apply Fully-Connected-based GNN and LSTM-based decoder. The homogeneous architecture enables us to unify the training configuration of the whole model instead of specifying different training strategies for diverse sub-networks as in the heterogeneous pipeline, which releases the training difficulty. Extensive experiments on the MS-COCO captioning benchmark validate the effectiveness of our TSG. The code is in: https://anonymous.4open.science/r/ACL23_TSG.

Abstract: We present CLUSTSEG, a general, transformer-based framework that tackles different image segmentation tasks (i.e., superpixel, semantic, instance, and panoptic) through a unified neural clustering scheme. Regarding queries as cluster centers, CLUSTSEG is innovative in two aspects:1) cluster centers are initialized in heterogeneous ways so as to pointedly address task-specific demands (e.g., instance- or category-level distinctiveness), yet without modifying the architecture; and 2) pixel-cluster assignment, formalized in a cross-attention fashion, is alternated with cluster center update, yet without learning additional parameters. These innovations closely link CLUSTSEG to EM clustering and make it a transparent and powerful framework that yields superior results across the above segmentation tasks.

Abstract: Inverse rendering methods that account for global illumination are becoming more popular, but current methods require evaluating and automatically differentiating millions of path integrals by tracing multiple light bounces, which remains expensive and prone to noise. Instead, this paper proposes a radiometric prior as a simple alternative to building complete path integrals in a traditional differentiable path tracer, while still correctly accounting for global illumination. Inspired by the Neural Radiosity technique, we use a neural network as a radiance function, and we introduce a prior consisting of the norm of the residual of the rendering equation in the inverse rendering loss. We train our radiance network and optimize scene parameters simultaneously using a loss consisting of both a photometric term between renderings and the multi-view input images, and our radiometric prior (the residual term). This residual term enforces a physical constraint on the optimization that ensures that the radiance field accounts for global illumination. We compare our method to a vanilla differentiable path tracer, and more advanced techniques such as Path Replay Backpropagation. Despite the simplicity of our approach, we can recover scene parameters with comparable and in some cases better quality, at considerably lower computation times.

Abstract: Language identification describes the task of recognizing the language of written text in documents. This information is crucial because it can be used to support the analysis of a document's vocabulary and context. Supervised learning methods in recent years have advanced the task of language identification. However, these methods usually require large labeled datasets, which often need to be included for various domains of images, such as documents or scene images. In this work, we propose DocLangID, a transfer learning approach to identify the language of unlabeled historical documents. We achieve this by first leveraging labeled data from a different but related domain of historical documents. Secondly, we implement a distance-based few-shot learning approach to adapt a convolutional neural network to new languages of the unlabeled dataset. By introducing small amounts of manually labeled examples from the set of unlabeled images, our feature extractor develops a better adaptability towards new and different data distributions of historical documents. We show that such a model can be effectively fine-tuned for the unlabeled set of images by only reusing the same few-shot examples. We showcase our work across 10 languages that mostly use the Latin script. Our experiments on historical documents demonstrate that our combined approach improves the language identification performance, achieving 74% recognition accuracy on the four unseen languages of the unlabeled dataset.

Abstract: In this work we present Low-rank Deconvolution, a powerful framework for low-level feature-map learning for efficient signal representation with application to signal recovery. Its formulation in multi-linear algebra inherits properties from convolutional sparse coding and low-rank approximation methods as in this setting signals are decomposed in a set of filters convolved with a set of low-rank tensors. We show its advantages by learning compressed video representations and solving image in-painting problems.

Abstract: The issue of Automatic License Plate Recognition (ALPR) has been one of the most challenging issues in recent years. Weather conditions, camera angle of view, lighting conditions, different characters written on license plates, and many other factors are among the challenges for the issue of ALPR. Given the advances that have been made in recent years in the field of deep neural networks, some types of neural networks and models based on them can be used to perform the task of Iranian license plate recognition. In the proposed method presented in this paper, the license plate recognition is done in two steps. The first step is to detect the rectangles of the license plates from the input image. In the second step, these license plates are cropped from the image and their characters are recognized. For the first step, 3065 images including license plates and for the second step, 3364 images including characters of license plates have been prepared and considered as the desired datasets. In the first step, license plates are detected using the YOLOv4-tiny model, which is based on Convolutional Neural Network (CNN). In the next step, the characters of these license plates are recognized using Convolutional Recurrent Neural Network (CRNN), and Connectionist Temporal Classification (CTC). In the second step, there is no need to segment and label the characters separately, only one string of numbers and letters is enough for the labels.

Abstract: We consider the problem of reconstructing a dynamic scene observed from a stereo camera. Most existing methods for depth from stereo treat different stereo frames independently, leading to temporally inconsistent depth predictions. Temporal consistency is especially important for immersive AR or VR scenarios, where flickering greatly diminishes the user experience. We propose DynamicStereo, a novel transformer-based architecture to estimate disparity for stereo videos. The network learns to pool information from neighboring frames to improve the temporal consistency of its predictions. Our architecture is designed to process stereo videos efficiently through divided attention layers. We also introduce Dynamic Replica, a new benchmark dataset containing synthetic videos of people and animals in scanned environments, which provides complementary training and evaluation data for dynamic stereo closer to real applications than existing datasets. Training with this dataset further improves the quality of predictions of our proposed DynamicStereo as well as prior methods. Finally, it acts as a benchmark for consistent stereo methods.

Abstract: Large-scale visual language models are widely used as pre-trained models and then adapted for various downstream tasks. While humans are known to efficiently learn new tasks from a few examples, deep learning models struggle with adaptation from few examples. In this work, we look into task adaptation in the low-data regime, and provide a thorough study of the existing adaptation methods for generative Visual Language Models. And we show important benefits of self-labelling, i.e. using the model's own predictions to self-improve when having access to a larger number of unlabelled images of the same distribution. Our study demonstrates significant gains using our proposed task adaptation pipeline across a wide range of visual language tasks such as visual classification (ImageNet), visual captioning (COCO), detailed visual captioning (Localised Narratives) and visual question answering (VQAv2).

Abstract: Existing fashion datasets do not consider the multi-facts that cause a consumer to like or dislike a fashion image. Even two consumers like a same fashion image, they could like this image for total different reasons. In this paper, we study the reason why a consumer like a certain fashion image. Towards this goal, we introduce an interpretability dataset, Fashionpedia-taste, consist of rich annotation to explain why a subject like or dislike a fashion image from the following 3 perspectives: 1) localized attributes; 2) human attention; 3) caption. Furthermore, subjects are asked to provide their personal attributes and preference on fashion, such as personality and preferred fashion brands. Our dataset makes it possible for researchers to build computational models to fully understand and interpret human fashion taste from different humanistic perspectives and modalities.

Abstract: We present a one-shot method to infer and render a photorealistic 3D representation from a single unposed image (e.g., face portrait) in real-time. Given a single RGB input, our image encoder directly predicts a canonical triplane representation of a neural radiance field for 3D-aware novel view synthesis via volume rendering. Our method is fast (24 fps) on consumer hardware, and produces higher quality results than strong GAN-inversion baselines that require test-time optimization. To train our triplane encoder pipeline, we use only synthetic data, showing how to distill the knowledge from a pretrained 3D GAN into a feedforward encoder. Technical contributions include a Vision Transformer-based triplane encoder, a camera data augmentation strategy, and a well-designed loss function for synthetic data training. We benchmark against the state-of-the-art methods, demonstrating significant improvements in robustness and image quality in challenging real-world settings. We showcase our results on portraits of faces (FFHQ) and cats (AFHQ), but our algorithm can also be applied in the future to other categories with a 3D-aware image generator.

Abstract: While progress in 2D generative models of human appearance has been rapid, many applications require 3D avatars that can be animated and rendered. Unfortunately, most existing methods for learning generative models of 3D humans with diverse shape and appearance require 3D training data, which is limited and expensive to acquire. The key to progress is hence to learn generative models of 3D avatars from abundant unstructured 2D image collections. However, learning realistic and complete 3D appearance and geometry in this under-constrained setting remains challenging, especially in the presence of loose clothing such as dresses. In this paper, we propose a new adversarial generative model of realistic 3D people from 2D images. Our method captures shape and deformation of the body and loose clothing by adopting a holistic 3D generator and integrating an efficient and flexible articulation module. To improve realism, we train our model using multiple discriminators while also integrating geometric cues in the form of predicted 2D normal maps. We experimentally find that our method outperforms previous 3D- and articulation-aware methods in terms of geometry and appearance. We validate the effectiveness of our model and the importance of each component via systematic ablation studies.

Abstract: The oscillations of climatic parameters of North Atlantic Ocean play important role in various events in North America and Europe. Several climatic indices are associated with these oscillations. The long term Atlantic temperature anomalies are described by the Atlantic Multidecadal Oscillation (AMO). The Atlantic Multidecadal Oscillation also known as Atlantic Multidecadal Variability (AMV), is the variability of the sea surface temperature (SST) of the North Atlantic Ocean at the timescale of several decades. The AMO is correlated to air temperatures and rainfall over much of the Northern Hemisphere, in particular in the summer climate in North America and Europe. The long-term variations of surface temperature are driven mainly by the cycles of solar activity, represented by the variations of the Total Solar Irradiance (TSI). The frequency and amplitude dependences between the TSI and AMO are analyzed by wavelet coherence of millennial time series since 800 AD till now. The results of wavelet coherence are compared with the detected common solar and climate cycles in narrow frequency bands by the method of Partial Fourier Approximation. The long-term coherence between TSI and AMO can help to understand better the recent climate change and can improve the long term forecast.