WACV 2024 论文列表

Most image classification tasks benefit from using pretrained feature stacks. In contrast, the discriminator for adversarial losses is trained at the same time as the model because using a pretrained feature stack yields a very poor model. Recent work has shown that an implicit regularization scheme allows using pretrained feature stacks to construct a discriminator, which improves both speed of training and quality of results. However, we observe that changes in hyperparameters can result in substantial changes in generator behavior. We show that using a modified version of the R1 regularization scheme that regularizes in the feature space instead of the image space results in a plug-and-play discriminator -- P2D. Our scheme results in a method that is highly stable across changes in architecture and framework; that significantly speeds up training; and that produces models that reliably beat SOTA in quality. The huge reduction in training resources required means that P2D could make training powerful generative models over specific datasets accessible to most researchers.

Age estimation is a challenging task that has numerous applications. In this paper, we propose a new direction for age classification that utilizes a video-based model to address challenges such as occlusions, low-resolution, and lighting conditions. To address these challenges, we propose AgeFormer which utilizes spatio-temporal information on the dynamics of the entire body dominating face-based methods for age classification. Our novel two-stream architecture uses TimeSformer and EfficientNet as backbones, to effectively capture both facial and body dynamics information for efficient and accurate age estimation in videos. Furthermore, to fill the gap in predicting age in real-world situations from videos, we construct a video dataset called Pexels Age (P-Age) for age classification. The proposed method achieves superior results compared to existing face-based age estimation methods and is evaluated in situations where the face is highly occluded, blurred, or masked. The method is also cross-tested on a variety of challenging video datasets such as Charades, Smarthome, and Thumos-14.

Despite the recent progress of class-incremental learning (CIL) methods, their capabilities in real-world scenarios such as multi-label settings remain unexplored. This paper focuses on a more practical CIL problem named multi-label class-incremental learning (MLCIL). MLCIL requires the vision models to overcome catastrophic forgetting of old knowledge while learning new classes from multi-label samples. Direct application of existing CIL methods to MLCIL leads to label absence, representative sample selection, and feature dilution problems. To address these problems, we present a novel AdaPtive Pseudo-Label-drivEn (APPLE) framework consisting of three components. First, the adaptive pseudo-label strategy is proposed to solve the label absence problem, which leverages the old model to annotate old classes for new samples. Second, a cluster sampling strategy is proposed to obtain more diverse samples to alleviate catastrophic forgetting under the MLCIL setting better. Finally, a class attention decoder is designed to mitigate the object feature dilution problem in multi-label samples. The extensive experiments on PASCAL VOC 2007 and MS-COCO demonstrate that our proposed method significantly outperforms other representative state-of-the-art CIL methods.

Machine Learning models often only generalize reliably to samples from the training distribution. Consequentially, detecting when input data is out-of-distribution (OOD) is crucial, especially in safety-critical applications. Current OOD detection methods, however, tend to be domain agnostic and often fail to incorporate valuable prior knowledge about the structure of the training distribution. To address this limitation, we introduce a novel, hybrid OOD detection algorithm that combines a deep learning-based perception system with a first-order logic-based knowledge representation. A logical reasoning system uses this knowledge base at run-time to infer whether inputs are consistent with prior knowledge about the training distribution. In contrast to purely neural systems, the structured knowledge representation allows humans to inspect and modify the rules that govern the OOD detectors' behavior. This not only enhances performance but also fosters a level of explainability that is particularly beneficial in safety-critical contexts. We demonstrate the effectiveness of our method through experiments on several datasets and discuss advantages and limitations. Our code is available online.

Age is one of the major known risk factors for Alzheimer's Disease (AD). Detecting AD early is crucial for effective treatment and preventing irreversible brain damage. Brain age, a measure derived from brain imaging reflecting structural changes due to aging, may have the potential to identify AD onset, assess disease risk, and plan targeted interventions. Deep learning-based regression techniques to predict brain age from magnetic resonance imaging (MRI) scans have shown great accuracy recently. However, these methods are subject to an inherent regression to the mean effect, which causes a systematic bias resulting in an overestimation of brain age in young subjects and underestimation in old subjects. This weakens the reliability of predicted brain age as a valid biomarker for downstream clinical applications. Here, we reformulate the brain age prediction task from regression to classification to address the issue of systematic bias. Recognizing the importance of preserving ordinal information from ages to understand aging trajectory and monitor aging longitudinally, we propose a novel ORdinal Distance Encoded Regularization (ORDER) loss that incorporates the order of age labels, enhancing the model's ability to capture age-related patterns. Extensive experiments and ablation studies demonstrate that this framework reduces systematic bias, outperforms state-of-art methods by statistically significant margins, and can better capture subtle differences between clinical groups in an independent AD dataset. Our implementation is publicly available at https://github.com/jaygshah/Robust-Brain-Age-Prediction.

Monitoring a fleet of robots requires stable long-term tracking with re-identification, which is yet an unsolved challenge in many scenarios. One application of this is the analysis of autonomous robotic soccer games at RoboCup. Tracking in these games requires handling of identically looking players, strong occlusions, and non-professional video recordings, but also offers state information estimated by the robots. In order to make effective use of the information coming from the robot sensors, we propose a robust tracking and identification pipeline. It fuses external non-calibrated camera data with the robots' internal states using quadratic optimization for tracklet matching. The approach is validated using game recordings from previous RoboCup World Cup tournaments.

Optical flows play an integral role for a variety of motion-related tasks such as action recognition, object segmentation, and tracking in videos. While state-of-the-art optical flow methods heavily rely on learning, the learned optical flow methods significantly degrade when applied to different domains, and the training datasets are very limited due to the extreme cost of flow-level annotation. To tackle the issue, we introduce a domain adaptation technique for optical flow estimation. Our method extracts diverse style statistics of the target domain and use them in training to generate synthetic features from the source features, which contain the contents of the source but the style of the target. We also impose motion consistency between the synthetic target and the source and deploy adversarial learning at the flow prediction to encourage domain-invariant features. Experimental results show that the proposed method achieves substantial and consistent improvements in different domain adaptation scenarios on VKITTI 2, Sintel, and KITTI 2015 benchmarks.

Scene flow estimation is a crucial component in the development of autonomous driving and 3D robotics, providing valuable information for environment perception and navigation. Despite the advantages of learning-based scene flow estimation techniques, their domain specificity and limited generalizability across varied scenarios pose challenges. In contrast, non-learning optimization-based methods, incorporating robust priors or regularization, offer competitive scene flow estimation performance, require no training, and show extensive applicability across datasets, but suffer from lengthy inference times. In this paper, we present OptFlow, a fast optimization-based scene flow estimation method. Without relying on learning or any labeled datasets, OptFlow achieves state-of-the-art performance for scene flow estimation on popular autonomous driving benchmarks. It integrates a local correlation weight matrix for correspondence matching, an adaptive correspondence threshold limit for nearest-neighbor search, and graph prior rigidity constraints, resulting in expedited convergence and improved point correspondence identification. Moreover, we demonstrate how integrating a point cloud registration function within our objective function bolsters accuracy and differentiates between static and dynamic points without relying on external odometry data. Consequently, OptFlow outperforms the baseline graph-prior method by approximately 20% and the Neural Scene Flow Prior method by 5%-7% in accuracy, all while offering the fastest inference time among all non-learning scene flow estimation methods.

Image quality assessment is a challenging computer vision task due to the lack of corresponding reference (pristine) images. This no-reference bottleneck has been tackled with the utilisation of subjective mean opinion scores (MOS) termed as supervised blind image quality assessment (BIQA) methods. However, inaccessible opinion score scenarios limits their applicability. To relieve these limitations, we propose to employ reconstruction based learning trained only on pristine images. This permits an implicit distribution learning of pristine images and the deviation from this learned feature distribution is subsequently utilised for unsupervised image quality assessment. Specifically, an adversarial convolutional variational auto-encoder framework is employed with KL divergence, perceptual and discriminator loss. With state-of-the-art results on four benchmark datasets, we demonstrate the effectiveness of our proposed framework. An ablation study has also been conducted to highlight the contribution of each module i.e. loss and quality metric for an efficient unsupervised BIQA.

Open Set Object Detection (OSOD) has emerged as a contemporary research direction to address the detection of unknown objects. Recently, few works have achieved remarkable performance in the OSOD task by employing contrastive clustering to separate unknown classes. In contrast, we propose a new semantic clustering-based approach to facilitate a meaningful alignment of clusters in semantic space and introduce a class decorrelation module to enhance inter-cluster separation. Our approach further incorporates an object focus module to predict objectness scores, which enhances the detection of unknown objects. Further, we employ i) an evaluation technique that penalizes low-confidence outputs to mitigate the risk of misclassification of the unknown objects and ii) a new metric called HMP that combines known and unknown precision using harmonic mean. Our extensive experiments demonstrate that the proposed model achieves significant improvement on the MS-COCO & PASCAL VOC dataset for the OSOD task.

In this paper, we address the challenge of decomposing Neural Radiance Fields (NeRF) into objects from an open vocabulary, a critical task for object manipulation in 3D reconstruction and view synthesis. Current techniques for NeRF decomposition involve a trade-off between the flexibility of processing open-vocabulary queries and the accuracy of 3D segmentation. We present, Open-vocabulary Embedded Neural Radiance Fields (Open-NeRF), that leverage large-scale, off-the-shelf, segmentation models like the Segment Anything Model (SAM) and introduce an integrate-and-distill paradigm with hierarchical embeddings to achieve both the flexibility of open-vocabulary querying and 3D segmentation accuracy. Open-NeRF first utilizes large-scale foundation models to generate hierarchical 2D mask proposals from varying viewpoints. These proposals are then aligned via tracking approaches and integrated within the 3D space and subsequently distilled into the 3D field. This process ensures consistent recognition and granularity of objects from different viewpoints, even in challenging scenarios involving occlusion and indistinct features. Our experimental results show that the proposed Open-NeRF outperforms state-of-the-art methods such as LERF and FFD in open-vocabulary scenarios. Open-NeRF offers a promising solution to NeRF decomposition, guided by open-vocabulary queries, enabling novel applications in robotics and vision-language interaction in open-world 3D scenes. Please find the code at https://github.com/haoz19/Open-NeRF

Food image classification is essential for monitoring health and tracking dietary in image-based dietary assessment methods. However, conventional systems often rely on static datasets with fixed classes and uniform distribution. In contrast, real-world food consumption patterns, shaped by cultural, economic, and personal influences, involve dynamic and evolving data. Thus, require the classification system to cope with continuously evolving data. Online Class Incremental Learning (OCIL) addresses the challenge of learning continuously from a single-pass data stream while adapting to the new knowledge and reducing catastrophic forgetting. Experience Replay (ER) based OCIL methods store a small portion of previous data and have shown encouraging performance. However, most existing OCIL works assume that the distribution of encountered data is perfectly balanced, which rarely happens in real-world scenarios. In this work, we explore OCIL for real-world food image classification by first introducing a probabilistic framework to simulate realistic food consumption scenarios. Subsequently, we present an attachable Dynamic Model Update (DMU) module designed for existing ER methods, which enables the selection of relevant images for model training, addressing challenges arising from data repetition and imbalanced sample occurrences inherent in realistic food consumption patterns within the OCIL framework. Our performance evaluation demonstrates significant enhancements compared to established ER methods, showing great potential for lifelong learning in real-world food image classification scenarios. The code of our method is publicly accessible at https://gitlab.com/viper-purdue/OCIL-real-world-food-image-classification

In this paper, we propose a style-based conditional video generative model. We introduce a novel temporal generator based on a set of learned sinusoidal bases. Our method learns dynamic representations of various actions that are independent of image content and can be transferred between different actors. Beyond the significant enhancement of video quality compared to prevalent methods, we demonstrate that the disentangled dynamic and content permit their independent manipulation, as well as temporal GAN-inversion to retrieve and transfer a video motion from one content or identity to another without further preprocessing such as landmark points.

Computational imaging plays a pivotal role in determining hidden information from sparse measurements. A robust inverse solver is crucial to fully characterize the uncertainty induced by these measurements, as it allows for the estimation of the complete posterior of unrecoverable targets. This, in turn, facilitates a probabilistic interpretation of observational data for decision-making. In this study, we propose a deep variational framework that leverages a deep generative model to learn an approximate posterior distribution to effectively quantify image reconstruction uncertainty without the need for training data. We parameterize the target posterior using a flow-based model and minimize their Kullback-Leibler (KL) divergence to achieve accurate uncertainty estimation. To bolster stability, we introduce a robust flow-based model with bi-directional regularization and enhance expressivity through gradient boosting. Additionally, we incorporate a space-filling design to achieve substantial variance reduction on both latent prior space and target posterior space. We validate our method on several benchmark tasks and two real-world applications, namely fastMRI and black hole image reconstruction. Our results indicate that our method provides reliable and high-quality image reconstruction with robust uncertainty estimation.

Object detection models, a prominent class of machine learning algorithms, aim to identify and precisely locate objects in images or videos. However, the task of accurately localizing objects within images yields uneven performances sometimes caused by the objects sizes and the quality of the images and labels. In this paper, we highlight the importance of large objects in learning features that are critical for all sizes. Given these findings, we propose to address this by introducing a weighting term into the loss during training. This term is a function of the object area size. We show that giving more weight to large objects leads to improvement in detection scores across all sizes and so an overall improvement in Object Detectors performances (+2% mAP on small objects, +2% on medium and +4% on large on COCO val 2017 with InternImage-T). Additional experiments and ablation studies with different models and on different dataset further confirm the robustness of our findings.

Achieving high accuracy on data from domains unseen during training is a fundamental challenge in domain generalization (DG). While state-of-the-art DG classifiers have demonstrated impressive performance across various tasks, they have shown a bias towards domain-dependent information, such as image styles, rather than domain-invariant information, such as image content. This bias renders them unreliable for deployment in risk-sensitive scenarios such as autonomous driving where a misclassification could lead to catastrophic consequences. To enable risk-averse predictions from a DG classifier, we propose a novel inference procedure, Test-Time Neural Style Smoothing (TT-NSS), that uses a "style-smoothed" version of the DG classifier for prediction at test time. Specifically, the style-smoothed classifier classifies a test image as the most probable class predicted by the DG classifier on random re-stylizations of the test image. TT-NSS uses a neural style transfer module to stylize a test image on the fly, requires only black-box access to the DG classifier, and crucially, abstains when predictions of the DG classifier on the stylized test images lack consensus. Additionally, we propose a neural style smoothing (NSS) based training procedure that can be seamlessly integrated with existing DG methods. This procedure enhances prediction consistency, improving the performance of TT-NSS on non-abstained samples. Our empirical results demonstrate the effectiveness of TT-NSS and NSS at producing and improving risk-averse predictions on unseen domains from DG classifiers trained with SOTA training methods on various benchmark datasets and their variations.

Diffusion models have gained attention for image editing yielding impressive results in text-to-image tasks. On the downside, one might notice that generated images of stable diffusion models suffer from deteriorated details. This pitfall impacts image editing tasks that require information preservation e.g., scene text editing. As a desired result, the model must show the capability to replace the text on the source image to the target text while preserving the details e.g., color, font size, and background. To leverage the potential of diffusion models, in this work, we introduce Diffusion-BasEd Scene Text manipulation network so-called DBEST. Specifically, we design two adaptation strategies, namely one-shot style adaptation and text-recognition guidance. In experiments, we thoroughly assess and compare our proposed method against state-of-the-arts on various scene text datasets, then provide extensive ablation studies for each granularity to analyze our performance gain. Also, we demonstrate the effectiveness of our proposed method to synthesize scene text indicated by competitive Optical Character Recognition (OCR) accuracy. Our method achieves 94.15% and 98.12% on COCO-text and ICDAR2013 datasets for character-level evaluation.

Majority of research in learning based methods has been towards designing and training networks for specific tasks. However, many of the learning based tasks, across modalities, share commonalities and could be potentially tackled in a joint framework. We present an approach in such direction, to learn multiple tasks, in multiple modalities, with a unified architecture. The proposed network is composed of task specific encoders, a common trunk in the middle, followed by task specific prediction heads. We first pre-train it by self-supervised masked training, followed by sequential training for the different tasks. We train the network on all major modalities, e.g. visual, audio, text and 3D, and report results on 22 diverse and challenging public benchmarks. We demonstrate empirically that, using a joint network to train across modalities leads to meaningful information sharing and this allows us to achieve state-of-the-art results on most of the benchmarks. We also show generalization of the trained network on cross-modal tasks as well as unseen datasets and tasks.

In this paper, we present offline-to-online knowledge distillation (OOKD) for video instance segmentation (VIS), which transfers a wealth of video knowledge from an offline model to an online model for consistent prediction. Unlike previous methods that have adopted either an online or offline model, our single online model takes advantage of both models by distilling offline knowledge. To transfer knowledge correctly, we propose query filtering and association (QFA), which filters irrelevant queries to exact instances. Our KD with QFA increases the robustness of feature matching by encoding object-centric features from a single frame supplemented by long-range global information. We also propose a simple data augmentation scheme for knowledge distillation in the VIS task that fairly transfers the knowledge of all classes into the online model. Extensive experiments show that our method significantly improves the performance in video instance segmentation, especially for challenging datasets, including long, dynamic sequences. Our method also achieves state-of-the-art performance on YTVIS-21, YTVIS-22, and OVIS datasets, with mAP scores of 46.1%, 43.6%, and 31.1%, respectively.