WACV 2024 论文列表

3DMM conditioned face generation has gained traction due to its well-defined controllability; however, the trade-off is lower sample quality: Previous works such as DiscoFaceGAN and 3D-FM GAN show a significant FID gap compared to the unconditional StyleGAN, suggesting that there is a quality tax to pay for controllability. In this paper, we challenge the assumption that quality and controllability cannot coexist. To pinpoint the previous issues, we mathematically formalize the problem of 3DMM conditioned face generation. Then, we devise simple solutions to the problem under our proposed framework. This results in a new model that effectively removes the quality tax between 3DMM conditioned face GANs and the unconditional StyleGAN. Project webpage: https://visual.cs.brown.edu/taxfreegan

In deformable object manipulation, we often want to interact with specific segments of an object that are only defined in non-deformed models of the object. We thus require a system that can recognize and locate these segments in sensor data of deformed real world objects. This is normally done using deformable object registration, which is problem specific and complex to tune. Recent methods utilize neural occupancy functions to improve deformable object registration by registering to an object reconstruction. Going one step further, we propose a system that in addition to reconstruction learns segmentation of the reconstructed object. As the resulting output already contains the information about the segments, we can skip the registration process. Tested on a variety of deformable objects in simulation and the real world, we demonstrate that our method learns to robustly find these segments. We also introduce a simple sampling algorithm to generate better training data for occupancy learning.

Scene Graph Generation (SGG) plays an important role in enhancing visual image comprehension. However, existing approaches often struggle to represent implicit relationship features, resulting in a limited ability to distinguish predicates. Meanwhile, they are vulnerable to skewed instance distributions, which impairs effective training for fine-grained predicates. To address these problems, we propose a novel feature refinement and data redistribution framework (RAR). Specifically, a multi-domain fusion (MDF) module is designed to acquire comprehensive predicate representations, integrating global knowledge from the contextual domain and local details in the spatial-frequency domains. Then, we introduce a dynamic label assignment (DLA) strategy to tackle the long-tailed problem. Different predicate categories are adaptively grouped, accommodating varying training conditions. Guided by this strategy, we leverage a hierarchical auto-encoder to generate siamese samples, expanding the label cardinality. Furthermore, we explore the updated sample space to derive reliable samples and assign tailored labels, ultimately achieving the data rebalancing. Experiments on VG and GQA demonstrate that our model contributes to correcting prediction bias and achieves a significant improvement of approximately 10% in mean recall compared to baseline models.

Analog magnetic tapes have been the main video data storage device for several decades. Videos stored on analog videotapes exhibit unique degradation patterns caused by tape aging and reader device malfunctioning that are different from those observed in film and digital video restoration tasks. In this work, we present a reference-based approach for the resToration of digitized Analog videotaPEs (TAPE). We leverage CLIP for zero-shot artifact detection to identify the cleanest frames of each video through textual prompts describing different artifacts. Then, we select the clean frames most similar to the input ones and employ them as references. We design a transformer-based Swin-UNet network that exploits both neighboring and reference frames via our Multi-Reference Spatial Feature Fusion (MRSFF) blocks. MRSFF blocks rely on cross-attention and attention pooling to take advantage of the most useful parts of each reference frame. To address the absence of ground truth in real-world videos, we create a synthetic dataset of videos exhibiting artifacts that closely resemble those commonly found in analog videotapes. Both quantitative and qualitative experiments show the effectiveness of our approach compared to other state-of-the-art methods. The code, the model, and the synthetic dataset are publicly available at https://github.com/miccunifi/TAPE.

Quantized networks use less computational and memory resources and are suitable for deployment on edge devices. While quantization-aware training QAT is the well-studied approach to quantize the networks at low precision, most research focuses on over-parameterized networks for classification with limited studies on popular and edge device friendly single-shot object detection and semantic segmentation methods like YOLO. Moreover, majority of QAT methods rely on Straight-through Estimator (STE) approximation which suffers from an oscillation phenomenon resulting in sub-optimal network quantization. In this paper, we show that it is difficult to achieve extremely low precision (4-bit and lower) for efficient YOLO models even with SOTA QAT methods due to oscillation issue and existing methods to overcome this problem are not effective on these models. To mitigate the effect of oscillation, we first propose Exponentially Moving Average (EMA) based update to the QAT model. Further, we propose a simple QAT correction method, namely QC, that takes only a single epoch of training after standard QAT procedure to correct the error induced by oscillating weights and activations resulting in a more accurate quantized model. With extensive evaluation on COCO dataset using various YOLO5 and YOLO7 variants, we show that our correction method improves quantized YOLO networks consistently on both object detection and segmentation tasks at low-precision (4-bit and 3-bit).

Despite the remarkable success of deep learning systems over the last decade, a key difference still remains between neural network and human decision-making: As humans, we can not only form a decision on the spot, but also ponder, revisiting an initial guess from different angles, distilling relevant information, arriving at a better decision. Here, we propose RecycleNet, a latent feature recycling method, instilling the pondering capability for neural networks to refine initial decisions over a number of recycling steps, where outputs are fed back into earlier network layers in an iterative fashion. This approach makes minimal assumptions about the neural network architecture and thus can be implemented in a wide variety of contexts. Using medical image segmentation as the evaluation environment, we show that latent feature recycling enables the network to iteratively refine initial predictions even beyond the iterations seen during training, converging towards an improved decision. We evaluate this across a variety of segmentation benchmarks and show consistent improvements even compared with top-performing segmentation methods. This allows trading increased computation time for improved performance, which can be beneficial, especially for safety-critical applications.

We exploit field guides to learn bird species recognition, in particular zero-shot recognition of unseen species. Illustrations contained in field guides deliberately focus on discriminative properties of each species, and can serve as side information to transfer knowledge from seen to unseen bird species. We study two approaches: (1) a contrastive encoding of illustrations, which can be fed into standard zero-shot learning schemes; and (2) a novel method that leverages the fact that illustrations are also images and as such structurally more similar to photographs than other kinds of side information. Our results show that illustrations from field guides, which are readily available for a wide range of species, are indeed a competitive source of side information for zero-shot learning. On a subset of the iNaturalist2021 dataset with 749 seen and 739 unseen species, we obtain a classification accuracy of unseen bird species of 12% @top-1 and 38% @top-10, which shows the potential of field guides for challenging real-world scenarios with many species. Our code is available at https://github.com/ac-rodriguez/zsl_billow.

Weakly supervised video anomaly detection is an important problem in many real-world applications where during training there are some anomalous videos, in addition to nominal videos, without labelled frames to indicate when the anomaly happens. State-of-the-art methods in this domain typically focus on offline anomaly detection without any concern for real-time detection. Most of these methods rely on ad hoc feature aggregation techniques and the use of metric learning losses, which limit the ability of the models to detect anomalies in real-time. In line with the premise of deep neural networks, there also has been a growing interest in developing end-to-end approaches that can automatically learn effective features directly from the raw data. We propose the first real-time and end-to-end trained algorithm for weakly supervised video anomaly detection. Our training procedure builds upon recent action recognition literature and uses a trainable video model to learn visual features. This is in contrast to existing approaches which largely depend on pre-trained feature extractors. The proposed method significantly improves the anomaly detection speed and AUC performance compared to the existing methods. Specifically, on the UCF-Crime dataset, our method achieves 86.94% AUC with a decision period of 6.4 seconds while the competing methods achieve at most 85.92% AUC with a decision period of 273 seconds.

Recently, the vision transformer (ViT) has achieved remarkable performance in computer vision tasks and has been actively utilized in colorization. Vision transformer uses multi-head self attention to effectively propagate user hints to distant relevant areas in the image. However, despite the success of vision transformers in colorizing the image, heavy underlying ViT architecture and the large computational cost hinder active real-time user interaction for colorization applications. Several research removed redundant image patches to reduce the computational cost of ViT in image classification tasks. However, the existing efficient ViT methods cause severe performance degradation in colorization task since it completely removes the redundant patches. Thus, we propose a novel efficient ViT architecture for real-time interactive colorization, AdaColViT determines which redundant image patches and layers to reduce in the ViT. Unlike existing methods, our novel pruning method alleviates performance drop and flexibly allocates computational resources of input samples, effectively achieving actual acceleration. In addition, we demonstrate through extensive experiments on ImageNet-ctest10k, Oxford 102flowers, and CUB-200 datasets that our method outperforms the baseline methods.

Colorectal cancer (CRC) represents a major global health challenge, and early detection of polyps is crucial in preventing its progression. Although colonoscopy is the gold standard for polyp detection, it has limitations, such as human error and missed detection rates. In response, computer-aided detection (CADe) systems have been developed to enhance the efficiency and accuracy of polyp detection. As deep learning gained prominence, the incorporation of Convolutional Neural Networks (CNNs) into CADe systems emerged as a breakthrough approach. However, CADe systems based on CNNs often demand significant computational resources, making them unsuitable for deployment in resource-constrained environments. To mitigate this, we propose a novel and lightweight polyp detection model that integrates a Transformer layer into the You Only Look Once (YOLO) architecture, focusing on optimizing the neck part responsible for feature fusion and rescaling. Our model demonstrates a substantial reduction in computational complexity and the number of parameters, without compromising detection performances. The lightweight model makes it accessible and feasibly deployable in medically underserved regions, serving a significant public interest by potentially expanding the reach of critical diagnostic tools for CRC prevention. By optimizing the architecture to reduce resource requirements while maintaining performance, our model becomes a practical solution to assist healthcare professionals in the real-time identification of polyps, even with resource-constraint devices.

Real-time applications for autonomous operations depend largely on fast and robust vision-based localization systems. Since image processing tasks require processing large amounts of data, the computational resources often limit the performance of other processes. To overcome this limitation, traditional marker-based localization systems are widely used since they are easy to integrate and achieve reliable accuracy. However, classical marker-based localization systems significantly depend on standard cameras with low frame rates, which often lack accuracy due to motion blur. In contrast, event-based cameras provide high temporal resolution and a high dynamic range, which can be utilized for fast localization tasks, even under challenging visual conditions. This paper proposes a simple but effective event-based pose estimation system using active LED markers (ALM) for fast and accurate pose estimation. The proposed algorithm is able to operate in real time with a latency below 0.5 ms while maintaining output rates of 3 kHz. Experimental results in static and dynamic scenarios are presented to demonstrate the performance of the proposed approach in terms of computational speed and absolute accuracy, using the OptiTrack system as the basis for measurement. Moreover, we demonstrate the feasibility of the proposed approach by deploying the hardware, i.e., the event-based camera and ALM, and the software in a real quadcopter application.

Eliminating time-consuming post-production processes and delivering high-quality videos in today's fast-paced digital landscape are the key advantages of real-time approaches. To address these needs, we present Real Time GAZED: a real-time adaptation of the GAZED framework integrated with CineFilter, a novel real-time camera trajectory stabilization approach. It enables users to create professionally edited videos in real-time. Comparative evaluations against baseline methods, including the non-real-time GAZED, demonstrate that Real Time GAZED achieves similar editing results, ensuring high-quality video output. Furthermore, a user study confirms the aesthetic quality of the video edits produced by the Real Time GAZED approach. With these advancements in real-time camera trajectory optimization and video editing presented, the demand for immediate and dynamic content creation in industries such as live broadcasting, sports coverage, news reporting, and social media content creation can be met more efficiently.

Anomaly detection is crucial to the advanced identification of product defects such as incorrect parts, misaligned components, and damages in industrial manufacturing. Due to the rare observations and unknown types of defects, anomaly detection is considered to be challenging in machine learning. To overcome this difficulty, recent approaches utilize the common visual representations pre-trained from natural image datasets and distill the relevant features. However, existing approaches still have the discrepancy between the pre-trained feature and the target data, or require the input augmentation which should be carefully designed, particularly for the industrial dataset. In this paper, we introduce ReConPatch, which constructs discriminative features for anomaly detection by training a linear modulation of patch features extracted from the pre-trained model. ReConPatch employs contrastive representation learning to collect and distribute features in a way that produces a target-oriented and easily separable representation. To address the absence of labeled pairs for the contrastive learning, we utilize two similarity measures between data representations, pairwise and contextual similarities, as pseudo-labels. Our method achieves the state-of-the-art anomaly detection performance (99.72%) for the widely used and challenging MVTec AD dataset. Additionally, we achieved a state-of-the-art anomaly detection performance (95.8%) for the BTAD dataset.

Large-scale pre-training vision-language models (VLM) such as CLIP has demonstrated outstanding performance in zero-shot classification, e.g. achieving 76.3% top-1 accuracy on ImageNet without seeing any example, which leads to potential benefits to many tasks that have no labeled data. However, while applying CLIP to a downstream target domain, the presence of visual and text domain gaps and cross-modality misalignment can greatly impact the model performance. To address such challenges, we propose ReCLIP, a novel source-free domain adaptation method for vision-language models, which does not require any source data or target labeled data. ReCLIP first learns a projection space to mitigate the misaligned visual-text embeddings and learns pseudo labels, and then deploys cross-modality self-training with the pseudo labels, to update visual and text encoders, refine labels and reduce domain gaps and misalignment iteratively. With extensive experiments, we demonstrate that ReCLIP outperforms all the baselines with significant margin and improves the averaged accuracy of CLIP from 69.83% to 74.94% on 22 image classification benchmarks.

Currently, widely employed LiDAR-based 3D object detectors adopt grid-based approaches to efficiently handle sparse point clouds. However, during this process, the down-sampled features inevitably lose spatial information, which can hinder the detectors from accurately predicting the location and size of objects. To address this issue, previous researches proposed sophisticatedly designed neck and head modules to effectively compensate for information loss. Inspired by the core insights of previous studies, we propose a novel voxel-based 3D object detector, named as Re-VoxelDet, which combines three distinct components to achieve both good detection capability and real-time performance. First, in order to learn features from diverse perspectives without additional computational costs during inference, we introduce Multi-view Voxel Backbone (MVBackbone). Second, to effectively compensate for abundant spatial and strong semantic information, we design Hierarchical Voxel-guided Auxiliary Neck (HVANeck), which attentively integrate hierarchically generated voxel-wise features with RPN blocks. Third, we present Rotation-based Group Head (RGHead), a simple yet effective head module that is designed with two groups according to the heading direction and aspect ratio of the objects. Through extensive experiments on the Argoverse2, nuScenes, and Waymo Open Dataset, we demonstrate the effectiveness of our approach. Our results significantly outperform existing state-of-the-art methods. We plan to release our model and code in the near future.

Popular benchmarks for self-supervised LiDAR scene flow (stereoKITTI, and FlyingThings3D) have unrealistic rates of dynamic motion, unrealistic correspondences, and unrealistic sampling patterns. As a result, progress on these benchmarks is misleading and may cause researchers to focus on the wrong problems. We evaluate a suite of top methods on a suite of real-world datasets (Argoverse 2.0, Waymo, and NuScenes) and report several conclusions. First, we find that performance on stereoKITTI is negatively correlated with performance on real-world data. Second, we find that one of this task's key components -- removing the dominant ego-motion -- is better solved by classic ICP than any tested method. Finally, we show that despite the emphasis placed on learning, most performance gains are caused by pre- and post-processing steps: piecewise- rigid refinement and ground removal. We demonstrate this through a baseline method that combines these processing steps with a learning-free test-time flow optimization. This baseline outperforms every evaluated method

Neural Radiance Fields (NeRF) have demonstrated exceptional capabilities in creating photorealistic novel views using volume rendering on a radiance field. However, the intrinsic assumption of straight light rays within NeRF becomes a limitation when dealing with transparent or translucent objects that exhibit refraction, and therefore have curved light paths. This hampers the ability of these approaches to accurately model the appearance of refractive objects, resulting in suboptimal novel view synthesis and geometry estimates. To address this issue, we propose an innovative solution using deformable networks to learn a tailored deformation field for refractive objects. Our approach predicts position and direction offsets, allowing NeRF to model the curved light paths caused by refraction and therefore the complex and highly view-dependent appearances of refractive objects. We also introduce a regularization strategy that encourages piece-wise linear light paths, since most physical systems can be approximated with a piece-wise constant index of refraction. By seamlessly integrating our deformation networks into the NeRF framework, our method achieves significant improvements in rendering refractive objects from novel views.

In recent years, deep learning techniques have shown significant potential for improving video quality assessment (VQA), achieving higher correlation with subjective opinions compared to conventional approaches. However, the development of deep VQA methods has been constrained by the limited availability of large-scale training databases and ineffective training methodologies. As a result, it is difficult for deep VQA approaches to achieve consistently superior performance and model generalization. In this context, this paper proposes new VQA methods based on a two-stage training methodology which motivates us to develop a large-scale VQA training database without employing human subjects to provide ground truth labels. This method was used to train a new transformer-based network architecture, exploiting quality ranking of different distorted sequences rather than minimizing the difference from the ground-truth quality labels. The resulting deep VQA methods (for both full reference and no reference scenarios), FR- and NR-RankDVQA, exhibit consistently higher correlation with perceptual quality compared to the state-of-the-art conventional and deep VQA methods, with average SROCC values of 0.8972 (FR) and 0.7791 (NR) over eight test sets without performing cross-validation. The source code of the proposed quality metrics and the large training database are available at https://chenfeng-bristol.github.io/RankDVQA.

The widespread adoption of commercial autonomous vehicles (AVs) and advanced driver assistance systems (ADAS) may largely depend on their acceptance by society, for which their perceived trustworthiness and interpretability to riders are crucial. In general, this task is challenging because modern autonomous systems software relies heavily on black-box artificial intelligence models. Towards this goal, this paper introduces a novel dataset, Rank2Tell, a multi-modal ego-centric dataset for Ranking the importance level and Telling the reason for the importance. Using various close and open-ended visual question answering, the dataset provides dense annotations of various semantic, spatial, temporal, and relational attributes of various important objects in complex traffic scenarios. The dense annotations and unique attributes of the dataset make it a valuable resource for researchers working on visual scene understanding and related fields. Furthermore, we introduce a joint model for joint importance level ranking and natural language captions generation to benchmark our dataset and demonstrate performance with quantitative evaluations.