Search

The recent emergence of latent diffusion models such as SDXL and SD 1.5 has shown significant capability in generating highly detailed and realistic images. Despite their remarkable ability to produce images, generating accurate text within images still remains a challenging task. In this paper, we examine the validity of fine-tuning approaches in generating legible text within the image. We propose a low-cost approach by leveraging SDXL without any time-consuming training on large-scale datasets. The proposed strategy employs a fine-tuning technique that examines the effects of data refinement levels and synthetic captions. Moreover, our results demonstrate how our small scale fine-tuning approach can improve the accuracy of text generation in different scenarios without the need of additional multimodal encoders. Our experiments show that with the addition of random letters to our raw dataset, our model's performance improves in producing well-formed visual text., Comment: 9 pages, 12 figures

Electroencephalography (EEG) motor imagery (MI) classification is a fundamental, yet challenging task due to the variation of signals between individuals i.e., inter-subject variability. Previous approaches try to mitigate this using task-specific (TS) EEG signals from the target subject in training. However, recording TS EEG signals requires time and limits its applicability in various fields. In contrast, resting state (RS) EEG signals are a viable alternative due to ease of acquisition with rich subject information. In this paper, we propose a novel subject-adaptive transfer learning strategy that utilizes RS EEG signals to adapt models on unseen subject data. Specifically, we disentangle extracted features into task- and subject-dependent features and use them to calibrate RS EEG signals for obtaining task information while preserving subject characteristics. The calibrated signals are then used to adapt the model to the target subject, enabling the model to simulate processing TS EEG signals of the target subject. The proposed method achieves state-of-the-art accuracy on three public benchmarks, demonstrating the effectiveness of our method in cross-subject EEG MI classification. Our findings highlight the potential of leveraging RS EEG signals to advance practical brain-computer interface systems. The code is available at https://github.com/SionAn/MICCAI2024-ResTL., Comment: Early Accepted at MICCAI 2024

The emergence of various adapters, including Low-Rank Adaptation (LoRA) applied from the field of natural language processing, has allowed diffusion models to personalize image generation at a low cost. However, due to the various challenges including limited datasets and shortage of regularization and computation resources, adapter training often results in unsatisfactory outcomes, leading to the corruption of the backbone model's prior knowledge. One of the well known phenomena is the loss of diversity in object generation, especially within the same class which leads to generating almost identical objects with minor variations. This poses challenges in generation capabilities. To solve this issue, we present Contrastive Adapter Training (CAT), a simple yet effective strategy to enhance adapter training through the application of CAT loss. Our approach facilitates the preservation of the base model's original knowledge when the model initiates adapters. Furthermore, we introduce the Knowledge Preservation Score (KPS) to evaluate CAT's ability to keep the former information. We qualitatively and quantitatively compare CAT's improvement. Finally, we mention the possibility of CAT in the aspects of multi-concept adapter and optimization., Comment: CVPRW 2024

Logical anomalies (LA) refer to data violating underlying logical constraints e.g., the quantity, arrangement, or composition of components within an image. Detecting accurately such anomalies requires models to reason about various component types through segmentation. However, curation of pixel-level annotations for semantic segmentation is both time-consuming and expensive. Although there are some prior few-shot or unsupervised co-part segmentation algorithms, they often fail on images with industrial object. These images have components with similar textures and shapes, and a precise differentiation proves challenging. In this study, we introduce a novel component segmentation model for LA detection that leverages a few labeled samples and unlabeled images sharing logical constraints. To ensure consistent segmentation across unlabeled images, we employ a histogram matching loss in conjunction with an entropy loss. As segmentation predictions play a crucial role, we propose to enhance both local and global sample validity detection by capturing key aspects from visual semantics via three memory banks: class histograms, component composition embeddings and patch-level representations. For effective LA detection, we propose an adaptive scaling strategy to standardize anomaly scores from different memory banks in inference. Extensive experiments on the public benchmark MVTec LOCO AD reveal our method achieves 98.1% AUROC in LA detection vs. 89.6% from competing methods., Comment: Accepted in AAAI2024

Artificial lattices have been used as a platform to extend the application of topological physics beyond electronic systems. Here, using the two-dimensional Lieb lattice as a prototypical example, we show that an array of disks which each support localized plasmon modes give rise to an analog of the quantum spin Hall state enforced by a synthetic time reversal symmetry. We find that an effective next-nearest-neighbor coupling mechanism intrinsic to the plasmonic disk array introduces a nontrivial $Z_2$ topological order and gaps out the Bloch spectrum. A faithful mapping of the plasmonic system onto a tight-binding model is developed and shown to capture its essential topological signatures. Full wave numerical simulations of graphene disks arranged in a Lieb lattice confirm the existence of propagating helical boundary modes in the nontrivial band gap., Comment: 14 pages, 5 figures, supplementary information included

As acquiring MRIs is expensive, neuroscience studies struggle to attain a sufficient number of them for properly training deep learning models. This challenge could be reduced by MRI synthesis, for which Generative Adversarial Networks (GANs) are popular. GANs, however, are commonly unstable and struggle with creating diverse and high-quality data. A more stable alternative is Diffusion Probabilistic Models (DPMs) with a fine-grained training strategy. To overcome their need for extensive computational resources, we propose a conditional DPM (cDPM) with a memory-efficient process that generates realistic-looking brain MRIs. To this end, we train a 2D cDPM to generate an MRI subvolume conditioned on another subset of slices from the same MRI. By generating slices using arbitrary combinations between condition and target slices, the model only requires limited computational resources to learn interdependencies between slices even if they are spatially far apart. After having learned these dependencies via an attention network, a new anatomy-consistent 3D brain MRI is generated by repeatedly applying the cDPM. Our experiments demonstrate that our method can generate high-quality 3D MRIs that share a similar distribution to real MRIs while still diversifying the training set. The code is available at https://github.com/xiaoiker/mask3DMRI_diffusion and also will be released as part of MONAI, at https://github.com/Project-MONAI/GenerativeModels.

Non-Hermitian degeneracies, also known as exceptional points (EPs), have been the focus of much attention due to their singular eigenvalue surface structure. Nevertheless, as pertaining to a non-Hermitian metasurface platform, the reduction of an eigenspace dimensionality at the EP has been investigated mostly in a passive repetitive manner. Here, we propose an electrical and spectral way of resolving chiral EPs and clarifying the consequences of chiral mode collapsing of a non-Hermitian gated graphene metasurface. More specifically, the measured non-Hermitian Jones matrix in parameter space enables the quantification of nonorthogonality of polarisation eigenstates and half-integer topological charges associated with a chiral EP. Interestingly, the output polarisation state can be made orthogonal to the coalesced polarisation eigenstate of the metasurface, revealing the missing dimension at the chiral EP. In addition, the maximal nonorthogonality at the chiral EP leads to a blocking of one of the cross-polarised transmission pathways and, consequently, the observation of enhanced asymmetric polarisation conversion. We anticipate that electrically controllable non-Hermitian metasurface platforms can serve as an interesting framework for the investigation of rich non-Hermitian polarisation dynamics around chiral EPs.

Whole slide image (WSI) classification is a fundamental task for the diagnosis and treatment of diseases; but, curation of accurate labels is time-consuming and limits the application of fully-supervised methods. To address this, multiple instance learning (MIL) is a popular method that poses classification as a weakly supervised learning task with slide-level labels only. While current MIL methods apply variants of the attention mechanism to re-weight instance features with stronger models, scant attention is paid to the properties of the data distribution. In this work, we propose to re-calibrate the distribution of a WSI bag (instances) by using the statistics of the max-instance (critical) feature. We assume that in binary MIL, positive bags have larger feature magnitudes than negatives, thus we can enforce the model to maximize the discrepancy between bags with a metric feature loss that models positive bags as out-of-distribution. To achieve this, unlike existing MIL methods that use single-batch training modes, we propose balanced-batch sampling to effectively use the feature loss i.e., (+/-) bags simultaneously. Further, we employ a position encoding module (PEM) to model spatial/morphological information, and perform pooling by multi-head self-attention (PSMA) with a Transformer encoder. Experimental results on existing benchmark datasets show our approach is effective and improves over state-of-the-art MIL methods., Comment: MICCAI 2022

Surface plasmons, which allow extreme confinement of light, suffer from high intrinsic electronic losses. It has been shown that stimulated emission of electrons can transfer energy to plasmons and compensate for the high intrinsic losses. To-date, these realizations have relied on introducing an external gain media coupled to the surface plasmon. Here, we propose that plasmons in two-dimensional materials with closely located electron and hole Fermi pockets can experience gain, when an electrical current bias is applied along the displaced electron-hole pockets, without the need for an external gain media. As a prototypical example, we consider WTe$_2$ from the family of 1T$'$-MX$_2$ materials, whose electronic structure can be described within a type-II tilted massive Dirac model. We find that the nonlocal plasmonic response experiences prominent gain for experimentally accessible currents on the order of mA$\mu$m$^{-1}$. Furthermore, the group velocity of the plasmon found from the isofrequency curves imply that the amplified plasmons are highly collimated along a direction perpendicular to the Dirac node tilt when the electrical current is applied along it., Comment: 12 pages, 4 figures

For in-building networks, the potential of transparent antennas, which are used as windows of a building, is presented in this paper. In this scenario, a transparent window antenna communicates with outdoor devices or base stations, and the indoor repeaters act as relay stations of the transparent window antenna for indoor devices. At indoor, back lobe waves of the transparent window antenna are defined as interference to in-building networks. Hence, we analyze different SIR and SINR results according to the location of an indoor repeater through 3D ray tracing system-level simulation. Furthermore, a link-level simulation through a full-duplex software-defined radio platform with the fabricated transparent antenna is presented to examine the feasibility of the transparent antenna., Comment: 2 pages, 3 figures

Few-shot classification is a challenging problem that aims to learn a model that can adapt to unseen classes given a few labeled samples. Recent approaches pre-train a feature extractor, and then fine-tune for episodic meta-learning. Other methods leverage spatial features to learn pixel-level correspondence while jointly training a classifier. However, results using such approaches show marginal improvements. In this paper, inspired by the transformer style self-attention mechanism, we propose a strategy to cross-attend and re-weight discriminative features for few-shot classification. Given a base representation of support and query images after global pooling, we introduce a single shared module that projects features and cross-attends in two aspects: (i) query to support, and (ii) support to query. The module computes attention scores between features to produce an attention pooled representation of features in the same class that is later added to the original representation followed by a projection head. This effectively re-weights features in both aspects (i & ii) to produce features that better facilitate improved metric-based meta-learning. Extensive experiments on public benchmarks show our approach outperforms state-of-the-art methods by 3%~5%., Comment: CVPR2022

In this article, we discuss the potential of radio frequency (RF) lens for ultra-wideband millimeter-wave (mmWave) systems. In terms of the beam squint, we compare the proposed RF lens antenna with the phase shifter-based array for hybrid beamforming. To reduce the complexities for fully digital beamforming, researchers have come up with RF lens-based hybrid beamforming. The use of mmWave systems, however, causes an increase in bandwidth, which gives rise to the beam squint phenomenon. We first find the causative factors for beam squint in the dielectric RF lens antenna. Based on the beamforming gain at each frequency, we verify that, in a specific situation, RF lens can be free of the beam squint effect. We use 3D electromagnetic analysis software to numerically interpret the beam squint of each antenna type. Based on the results, we present the degraded spectral efficiency by system-level simulations with 3D indoor ray tracing. Finally, to verify our analysis, we fabricate an actual RF lens antenna and demonstrate the real performance using a mmWave, NI PXIe, software-defined radio system., Comment: 8 pages, 4 figures, 2 tables

Few shot segmentation (FSS) aims to learn pixel-level classification of a target object in a query image using only a few annotated support samples. This is challenging as it requires modeling appearance variations of target objects and the diverse visual cues between query and support images with limited information. To address this problem, we propose a semi-supervised FSS strategy that leverages additional prototypes from unlabeled images with uncertainty guided pseudo label refinement. To obtain reliable prototypes from unlabeled images, we meta-train a neural network to jointly predict segmentation and estimate the uncertainty of predictions. We employ the uncertainty estimates to exclude predictions with high degrees of uncertainty for pseudo label construction to obtain additional prototypes based on the refined pseudo labels. During inference, query segmentation is predicted using prototypes from both support and unlabeled images including low-level features of the query images. Our approach is end-to-end and can easily supplement existing approaches without the requirement of additional training to employ unlabeled samples. Extensive experiments on PASCAL-$5^i$ and COCO-$20^i$ demonstrate that our model can effectively remove unreliable predictions to refine pseudo labels and significantly improve upon state-of-the-art performances., Comment: 9 pages

Models trained on datasets with texture bias usually perform poorly on out-of-distribution samples since biased representations are embedded into the model. Recently, various image translation and debiasing methods have attempted to disentangle texture biased representations for downstream tasks, but accurately discarding biased features without altering other relevant information is still challenging. In this paper, we propose a novel framework that leverages image translation to generate additional training images using the content of a source image and the texture of a target image with a different bias property to explicitly mitigate texture bias when training a model on a target task. Our model ensures texture similarity between the target and generated images via a texture co-occurrence loss while preserving content details from source images with a spatial self-similarity loss. Both the generated and original training images are combined to train improved classification or segmentation models robust to inconsistent texture bias. Evaluation on five classification- and two segmentation-datasets with known texture biases demonstrates the utility of our method, and reports significant improvements over recent state-of-the-art methods in all cases.

The vulnerabilities of global navigation satellite systems (GNSSs) to radio frequency jamming and spoofing have attracted significant research attention. In particular, the large-scale jamming incidents that occurred in South Korea substantiate the practical importance of implementing a complementary navigation system. This letter briefly summarizes the efforts of South Korea to deploy an enhanced long-range navigation (eLoran) system, which is a terrestrial low-frequency radio navigation system that can complement GNSSs. After four years of research and development, the Korean eLoran testbed system has been recently deployed and is operational since June 1, 2021. Although its initial performance at sea is satisfactory, navigation through a narrow waterway is still challenging because a complete survey of the additional secondary factor (ASF), which is the largest source of error for eLoran, is practically difficult in a narrow waterway. This letter proposes an alternative way to survey the ASF in a narrow waterway and improve the ASF map generation methods. Moreover, the performance of the proposed approach was validated experimentally., Comment: Submitted to IEEE Transactions on Aerospace and Electronic Systems

Non-rigid registration is a necessary but challenging task in medical imaging studies. Recently, unsupervised registration models have shown good performance, but they often require a large-scale training dataset and long training times. Therefore, in real world application where only dozens to hundreds of image pairs are available, existing models cannot be practically used. To address these limitations, we propose a novel unsupervised registration model which is integrated with a gradient-based meta learning framework. In particular, we train a meta learner which finds an initialization point of parameters by utilizing a variety of existing registration datasets. To quickly adapt to various tasks, the meta learner was updated to get close to the center of parameters which are fine-tuned for each registration task. Thereby, our model can adapt to unseen domain tasks via a short fine-tuning process and perform accurate registration. To verify the superiority of our model, we train the model for various 2D medical image registration tasks such as retinal choroid Optical Coherence Tomography Angiography (OCTA), CT organs, and brain MRI scans and test on registration of retinal OCTA Superficial Capillary Plexus (SCP). In our experiments, the proposed model obtained significantly improved performance in terms of accuracy and training time compared to other registration models.

Recently, Self-supervised learning methods able to perform image denoising without ground truth labels have been proposed. These methods create low-quality images by adding random or Gaussian noise to images and then train a model for denoising. Ideally, it would be beneficial if one can generate high-quality CT images with only a few training samples via self-supervision. However, the performance of CT denoising is generally limited due to the complexity of CT noise. To address this problem, we propose a novel self-supervised learning-based CT denoising method. In particular, we train pre-train CT denoising and noise models that can predict CT noise from Low-dose CT (LDCT) using available LDCT and Normal-dose CT (NDCT) pairs. For a given test LDCT, we generate Pseudo-LDCT and NDCT pairs using the pre-trained denoising and noise models and then update the parameters of the denoising model using these pairs to remove noise in the test LDCT. To make realistic Pseudo LDCT, we train multiple noise models from individual images and generate the noise using the ensemble of noise models. We evaluate our method on the 2016 AAPM Low-Dose CT Grand Challenge dataset. The proposed ensemble noise model can generate realistic CT noise, and thus our method significantly improves the denoising performance existing denoising models trained by supervised- and self-supervised learning.

Following the pandemic outbreak, several works have proposed to diagnose COVID-19 with deep learning in computed tomography (CT); reporting performance on-par with experts. However, models trained/tested on the same in-distribution data may rely on the inherent data biases for successful prediction, failing to generalize on out-of-distribution samples or CT with different scanning protocols. Early attempts have partly addressed bias-mitigation and generalization through augmentation or re-sampling, but are still limited by collection costs and the difficulty of quantifying bias in medical images. In this work, we propose Mixing-AdaSIN; a bias mitigation method that uses a generative model to generate de-biased images by mixing texture information between different labeled CT scans with semantically similar features. Here, we use Adaptive Structural Instance Normalization (AdaSIN) to enhance de-biasing generation quality and guarantee structural consistency. Following, a classifier trained with the generated images learns to correctly predict the label without bias and generalizes better. To demonstrate the efficacy of our method, we construct a biased COVID-19 vs. bacterial pneumonia dataset based on CT protocols and compare with existing state-of-the-art de-biasing methods. Our experiments show that classifiers trained with de-biased generated images report improved in-distribution performance and generalization on an external COVID-19 dataset.