WACV 2024 Papers

Visual as well as genetic biometrics are routinely employed to identify species and individuals in biological applications. However, no attempts have been made in this domain to computationally enhance visual classification of rare classes with little image data via genetics. In this paper, we thus propose aligned visual-genetic learning as a new application domain with the aim to implicitly encode cross-modality associations for improved performance. We demonstrate for the first time that such alignment can be achieved via deep embedding models and that the approach is directly applicable to boosting long-tailed recognition (LTR), particularly for rare species. We experimentally demonstrate the efficacy of the concept via application to microscopic imagery of 30k+ planktic foraminifer shells across 32 species when used together with independent genetic data samples. Most importantly for practitioners, we show that visual-genetic alignment can significantly benefit visual-only recognition of the rarest species. Technically, we pre-train a visual ResNet50 deep learning model using triplet loss formulations to create an initial embedding space. We re-structure this space based on genetic anchors embedded via a Sequence Graph Transform (SGT) and linked to visual data by cross-domain cosine alignment. We show that an LTR approach improves the state-of-the-art across all benchmarks and that adding our visual-genetic alignment improves per-class and particularly rare tail class benchmarks significantly further. We conclude that visual-genetic alignment can be a highly effective tool for complementing visual biological data containing rare classes. The concept proposed may serve as an important future tool for integrating genetics and imageomics towards a more complete scientific representation of taxonomic spaces and life itself. Code, weights, and data splits are published for full reproducibility.

Unsupervised domain adaptation (UDA), which aims to transfer knowledge learned from a labeled source domain to an unlabeled target domain, is useful for various cross-domain image classification scenarios. A commonly used approach for UDA is to minimize the distribution differences between two domains, and subdomain alignment is found to be an effective method. However, most of the existing subdomain alignment methods are based on adversarial learning and focus on subdomain alignment procedures without considering the discriminability among individual subdomains, resulting in slow convergence and unsatisfactory adaptation results. To address these issues, we propose a novel deep subdomain alignment method for UDA in image classification, which consists of a Union Subdomain Contrastive Learning (USCL) module and a Multi-view Subdomain Alignment (MvSA) strategy. USCL can create discriminative and dispersed subdomains by bringing samples from the same subdomain closer while pushing away samples from different subdomains. MvSA makes use of labeled source domain data and easy target domain data to perform target-to-source and target-to-target alignment. Experimental results on three image classification datasets (Office-31, Office-Home, Visda-17) demonstrate that our proposed method is effective for UDA and achieves promising results in several cross-domain image classification tasks.

Due to the setting of shutter speeds, over-exposed blurry images can often be seen in nighttime photography. Although image deblurring is a classic problem in image restoration, state-of-the-art methods often fail in nighttime cases with saturated pixels. The primary reason is that those pixels are out of the sensor range and thus violate the assumption of the linear blur model. To address this issue, we propose a new nighttime non-blind deblurring algorithm with saturated pixel handling schemes, including a pixel stretching mask, an image segment mask, and a saturation awareness mechanism (SAM). Our algorithm achieves superior results by strategically adjusting mask configurations, making our method robust to various saturation levels. We formulate our task into two new optimization problems and introduce a unified framework based on the plug-and-play alternating direction method of multipliers (PnP-ADMM). We also evaluate our approach qualitatively and quantitatively to demonstrate its effectiveness. The results show that the proposed algorithm recovers sharp latent images with finer details and fewer artifacts than other state-of-the-art deblurring methods.

An emerging class of Fizeau optical telescopes have the potential to upend prior cost scaling models, substantially improving the angular resolution and contrast attainable by ground-based astronomical instruments. However, this design introduces a challenging visual control problem that must be solved to compensate for wavefront aberrations induced by the flexible substructure it employs. We subvert this problem with a deep optics approach to policy design and image recovery that exploits, rather than corrects, aberrations to obtain domain-specific object recovery performance exceeding that of more costly filled aperture designs.

Deep metric learning (DML) aims to minimize empirical expected loss of the pairwise intra-/inter- class proximity violations in the embedding space. We relate DML to feasibility problem of finite chance constraints. We show that minimizer of proxy-based DML satisfies certain chance constraints, and that the worst case generalization performance of the proxy-based methods can be characterized by the radius of the smallest ball around a class proxy to cover the entire domain of the corresponding class samples, suggesting multiple proxies per class helps performance. To provide a scalable algorithm as well as exploiting more proxies, we consider the chance constraints implied by the minimizers of proxy-based DML instances and reformulate DML as finding a feasible point in intersection of such constraints, resulting in a problem to be approximately solved by iterative projections. Simply put, we repeatedly train a regularized proxy-based loss and re-initialize the proxies with the embeddings of the deliberately selected new samples. We applied our method with 4 well-accepted DML losses and show the effectiveness with extensive evaluations on 4 popular DML benchmarks. Code is available at: https://github.com/yetigurbuz/ccp-dml

The generation of high-quality images has become widely accessible and is a rapidly evolving process. As a result, anyone can generate images that are indistinguishable from real ones. This leads to a wide range of applications, including malicious usage with deceptive intentions. Despite advances in detection techniques for generated images, a robust detection method still eludes us. Furthermore, model personalization techniques might affect the detection capabilities of existing methods. In this work, we utilize the architectural properties of convolutional neural networks (CNNs) to develop a new detection method. Our method can detect images from a known generative model and enable us to establish relationships between fine-tuned generative models. We tested the method on images produced by both Generative Adversarial Networks (GANs) and recent large text-to-image models (LTIMs) that rely on Diffusion Models. Our approach outperforms others trained under identical conditions and achieves comparable performance to state-of-the-art pre-trained detection methods on images generated by Stable Diffusion and MidJourney, with significantly fewer required train samples.

In this work, we present a method to address the problem of novel view and time synthesis of complex dynamic scenes considering the input video is subject to blurriness caused due to camera or object motion or out-of-focus blur. Neural Scene Flow Field (NSFF) has shown remarkable results by training a dynamic NeRF to capture motion in the scene, but this method is not robust to unstable camera handling which can lead to blurred renderings. We propose Deblur-NSFF, a method that learns spatially-varying blur kernels to simulate the blurring process and gradually learns a sharp time-conditioned NeRF representation. We describe how to optimize our representation for sharp space-time view synthesis. Given blurry input frames, we perform both quantitative and qualitative comparison with state-of-the-art methods on modified NVIDIA Dynamic Scene dataset. We also compare our method with Deblur-NeRF, a method that has been designed to handle blur in static scenes. The demonstrated results show that our method outperforms prior work.

Recent studies on semi-supervised learning (SSL) have achieved great success. Despite their promising performance, current state-of-the-art methods tend toward increasingly complex designs at the cost of introducing more network components and additional training procedures. In this paper, we propose a simple method named Ensemble Projectors Aided for Semi-supervised Learning (EPASS), which focuses mainly on improving the learned embeddings to boost the performance of the existing contrastive joint-training semi-supervised learning frameworks. Unlike standard methods, where the learned embeddings from one projector are stored in memory banks to be used with contrastive learning, EPASS stores the ensemble embeddings from multiple projectors in memory banks. As a result, EPASS improves generalization, strengthens feature representation, and boosts performance. For instance, EPASS improves strong baselines for semi-supervised learning by 39.47%/31.39%/24.70% top-1 error rate, while using only 100k/1%/10% of labeled data for SimMatch, and achieves 40.24%/32.64%/25.90% top-1 error rate for CoMatch on the ImageNet dataset. These improvements are consistent across methods, network architectures, and datasets, proving the general effectiveness of the proposed methods.

The recent works on Video Object Segmentation achieved remarkable results by matching dense semantic and instance-level features between the current and previous frames for long-time propagation. Nevertheless, global feature matching ignores scene motion context, failing to satisfy temporal consistency. Even though some methods introduce local matching branch to achieve smooth propagation, they fail to model complex appearance changes due to the constraints of the local window. In this paper, we present DeVOS (Deformable VOS), an architecture for Video Object Segmentation that combines memory-based matching with motion-guided propagation resulting in stable long-term modeling and strong temporal consistency. For short-term local propagation, we propose a novel attention mechanism ADVA (Adaptive Deformable Video Attention), allowing the adaption of similarity search region to query-specific semantic features, which ensures robust tracking of complex shape and scale changes. DeVOS employs an optical flow to obtain scene motion features which are further injected to deformable attention as strong priors to learnable offsets. Our method achieves top-rank performance on DAVIS 2017 val and test-dev (88.1%, 83.0%), YouTube-VOS 2019 val (86.6%) while featuring consistent run-time speed and stable memory consumption.

Deep neural networks can be unreliable in the real world when the training set does not adequately cover all the settings where they are deployed. Focusing on image classification, we consider the setting where we have an error distribution E representing a deployment scenario where the model fails. We have access to a small set of samples E_sample from E and it can be expensive to obtain additional samples. In the traditional model development framework, mitigating failures of the model in E can be challenging and is often done in an ad hoc manner. In this paper, we propose a general methodology for model debugging that can systemically improve model performance on E while maintaining its performance on the original test set. Our key assumption is that we have access to a large pool of weakly (noisily) labeled data F. However, naively adding F to the training would hurt model performance due to the large extent of label noise. Our Data-Centric Debugging (DCD) framework carefully creates a debug-train set by selecting images from F that are perceptually similar to the images in E_sample. To do this, we use the l_2 distance in the feature space (penultimate layer activations) of various models including ResNet, Robust ResNet and DINO where we observe DINO ViTs are significantly better at discovering similar images compared to Resnets. Compared to the baselines that maintain model performance on the test set, we achieve significantly (+9.45%) improved results on the debug-heldout sets.

Data augmentation is vital for object detection tasks that require expensive bounding box annotations. Recent successes in diffusion models have inspired the use of diffusion-based synthetic images for data augmentation. However, existing works have primarily focused on image classification, and their applicability to boost object detection's performance remains unclear. To address this gap, we propose a data augmentation pipeline based on controllable diffusion models and CLIP. Our approach involves generating appropriate visual priors to control the generation of synthetic data and implementing post-filtering techniques using category-calibrated CLIP scores. The evaluation of our approach is conducted under few-shot settings in MSCOCO, full PASCAL VOC dataset, and selected downstream datasets. We observe the performance increase using our augmentation pipeline. Specifically, the mAP improvement is +18.0%/+15.6%/+15.9% for COCO 5/10/30-shot, +2.9% on full PASCAL VOC dataset, and +12.4% on average for selected downstream datasets.

Typical text recognition methods rely on an encoder-decoder structure, in which the encoder extracts features from an image, and the decoder produces recognized text from these features. In this study, we propose a simpler and more effective method for text recognition, known as the Decoder-only Transformer for Optical Character Recognition (DTrOCR). This method uses a decoder-only Transformer to take advantage of a generative language model that is pre-trained on a large corpus. We examined whether a generative language model that has been successful in natural language processing can also be effective for text recognition in computer vision. Our experiments demonstrated that DTrOCR outperforms current state-of-the-art methods by a large margin in the recognition of printed, handwritten, and scene text in both English and Chinese.

In this work we present DREAM, an fMRI-to-image method for reconstructing viewed images from brain activities, grounded on fundamental knowledge of the human visual system. We craft reverse pathways that emulate the hierarchical and parallel nature of how humans perceive the visual world. These tailored pathways are specialized to decipher semantics, color, and depth cues from fMRI data, mirroring the forward pathways from visual stimuli to fMRI recordings. To do so, two components mimic the inverse processes within the human visual system: the Reverse Visual Association Cortex (R-VAC) which reverses pathways of this brain region, extracting semantics from fMRI data; the Reverse Parallel PKM (R-PKM) component simultaneously predicting color and depth from fMRI signals. The experiments indicate that our method outperforms the current state-of-the-art models in terms of the consistency of appearance, structure, and semantics. Code will be available at https://github.com/weihaox/DREAM.

Although substantial progress has been made in audio-driven talking video synthesis, there still remain two major difficulties: existing works 1) need a long sequence of training dataset (>1h) to synthesize co-speech gestures, which causes a significant limitation on their applicability; 2) usually fail to generate long sequences, or can only generate long sequences without enough diversity. To solve these challenges, we propose a Disentangled Recurrent Representation Learning framework to synthesize long diversified gesture sequences with a short training video of around 2 minutes. In our framework, we first make a disentangled latent space assumption to encourage unpaired audio and pose combinations, which results in diverse "one-to-many" mappings in pose generation. Next, we apply a recurrent inference module to feed back the last generation as initial guidance to the next phase, enhancing the long-term video generation of full continuity and diversity. Comprehensive experimental results verify that our model can generate realistic synchronized full-body talking videos with training data efficiency.

In this paper, we contrast the usage of two deep-learning approaches for the automatic grading of diabetic retinopathy (DR) and diabetic macular edema (DME) in retinal fundus photographs using a relatively small novel dataset. We developed a telemedicine system to collect and humanly grade 11,109 diabetic patients. The certified graders annotated the level of DR as well as the existence of a referable DME in the macula-centered fundus images only. We use EfficientNet to build an AI-based model for both problems. To examine the transfer learning validity, the model was trained on an external dataset (EyePacs) and then finetuned on the egyptian data for the DR and DME grading problems. Firstly, we use the macula-centered images only in fine-tuning. Secondly, we use optic-disc-centered images in addition to macula-centered images. We obtained the labels for the optic-disc-centered images directly from the corresponding macula-centered labels as weak labels. Then, both types of images are used in fine-tuning. We found an increase in the DR performance using the second approach in both accuracy and quadratic weighted kappa(QWK). Notably, QWK increased from 90.23% to 91.3% using additional weakly labeled optic-disc-centered fundus images.

Masked Image Modeling (MIM) has achieved impressive representative performance with the aim of reconstructing randomly masked images. Despite the empirical success, most previous works have neglected the important fact that it is unreasonable to force the model to reconstruct something beyond recovery, such as those masked objects. In this work, we show that uniformly random masking widely used in previous works unavoidably loses some key objects and changes original semantic information, resulting in a misalignment problem and hurting the representative learning eventually. To address this issue, we augment MIM with a new masking strategy namely the DPPMask by substituting the random process with Determinantal Point Process (DPPs) to reduce the semantic change of the image after masking. Our method is simple yet effective and requires no extra learnable parameters when implemented within various frameworks. In particular, we evaluate our method on two representative MIM frameworks, MAE and iBOT. We show that DPPMask surpassed random sampling under both lower and higher masking ratios, indicating that DPPMask makes the reconstruction task more reasonable. We further test our method on the background challenge and multi-class classification tasks, showing that our method is more robust at various tasks.

Deep neural networks (DNNs) have great potential to solve many real-world problems, but they usually require an extensive amount of computation and memory. It is of great difficulty to deploy a large DNN model to a single resource-limited device with small memory capacity. Distributed computing is a common approach to reduce single-node memory consumption and to accelerate the inference of DNN models. In this paper, we explore the "within-layer model parallelism", which distributes the inference of each layer into multiple nodes. In this way, the memory requirement can be distributed to many nodes, making it possible to use several edge devices to infer a large DNN model. Due to the dependency within each layer, data communications between nodes during this parallel inference can be a bottleneck when the communication bandwidth is limited. We propose a framework to train DNN models for Distributed Inference with Sparse Communications (DISCO). We convert the problem of selecting which subset of data to transmit between nodes into a model optimization problem, and derive models with both computation and communication reduction when each layer is inferred on multiple nodes. We show the benefit of the DISCO framework on a variety of CV tasks such as image classification, object detection, semantic segmentation, and image super resolution. The corresponding models include important DNN building blocks such as convolutions and transformers. For example, each layer of a ResNet-50 model can be distributively inferred across two nodes with 5x less data communications, almost half overall computations and less than half memory requirement for a single node, and achieve comparable accuracy to the original ResNet-50 model.

The performance of optical character recognition (OCR) heavily relies on document image quality, which is crucial for automatic document processing and document intelligence. However, most existing document enhancement methods require supervised data pairs, which raises concerns about data separation and privacy protection, and makes it challenging to adapt these methods to new domain pairs. To address these issues, we propose DECDM, an end-to-end document-level image translation method inspired by recent advances in diffusion models. Our method overcomes the limitations of paired training by independently training the source (noisy input) and target (clean output) models, making it possible to apply domain-specific diffusion models to other pairs. DECDM trains on one dataset at a time, eliminating the need to scan both datasets concurrently, and effectively preserving data privacy from the source or target domain. We also introduce simple data augmentation strategies to improve character-glyph conservation during translation. We compare DECDM with state-of-the-art methods on multiple synthetic data and benchmark datasets, such as document denoising and shadow removal, and demonstrate the superiority of performance quantitatively and qualitatively.

Person search (PS) is a challenging computer vision problem where the objective is to achieve joint optimization for pedestrian detection and re-Person search (PS) is a challenging computer vision problem where the objective is to achieve joint optimization for pedestrian detection and re-identification (ReID). Although previous advancements have shown promising performance in the field under fully and weakly supervised learning fashion, there exists a major gap in investigating the domain adaptation ability of PS models. In this paper, we propose a diligent domain adaptive mixer (DDAM) for person search (DDAP-PS) framework that aims to bridge a gap to improve knowledge transfer from the labeled source domain to the unlabeled target domain. Specifically, we introduce a novel DDAM module that generates moderate mixed-domain representations by combining source and target domain representations. The proposed DDAM module encourages domain mixing to minimize the distance between the two extreme domains, thereby enhancing the ReID task. To achieve this, we introduce two bridge losses and a disparity loss. The objective of the two bridge losses is to guide the moderate mixed-domain representations to maintain an appropriate distance from both the source and target domain representations. The disparity loss aims to prevent the moderate mixed-domain representations from being biased towards either the source or target domains, thereby avoiding overfitting. Furthermore, we address the conflict between the two subtasks, localization and ReID, during domain adaptation. To handle this cross-task conflict, we forcefully decouple the norm-aware embedding, which aids in better learning of the moderate mixed-domain representation. We conduct experiments to validate the effectiveness of our proposed method. Our approach demonstrates favorable performance on the challenging PRW and CUHK-SYSU datasets. Our code is publicly available at https://github.com/mustansarfiaz/DDAM.