To investigate the feasibility of using a deep learning-based approach to detect an anterior cruciate ligament (ACL) tear within knee joint at MRI by arthroscopy as reference standard.A fully automated diagnosis system was developed two convolutional neural networks (CNNs) isolate ACL on MR images followed classification CNN structural abnormalities isolated ligament. With institutional review board approval, sagittal proton density-weighted and fat-suppressed T2-weighted fast spin-echo in...

Materials exhibiting negative differential resistance have important applications in technologies involving microwave generation, which range from motion sensing to radio astronomy. Despite their usefulness, there has been few physical mechanisms giving rise materials with such properties, i.e. GaAs employed the Gunn diode. In this work, we show that also generically arise Dirac ring systems, an example of experimentally observed surface states Topological Insulators. This novel realization...

Recently, there has been a drive towards the realization of topological phases beyond conventional electronic materials, including defined in more than three dimensions. We propose versatile and experimentally realistic approach realizing large variety multicomponent two-dimensional (2D) photonic crystals with quasiperiodically modulated defects. With length scale introduced by background resonator lattice, defects are found to host various effective orbitals $s$-, $p$-, $d$-type symmetries,...

The memory consumption of most Convolutional Neural Network (CNN) architectures grows rapidly with increasing depth the network, which is a major constraint for efficient network training on modern GPUs limited memory, embedded systems, and mobile devices. Several studies show that feature maps (as generated after convolutional layers) are main bottleneck in this problem. Often, these mimic natural photographs sense their energy concentrated spectral domain. Although embedding CNN do-main...

The memory consumption of most Convolutional Neural Network (CNN) architectures grows rapidly with increasing depth the network, which is a major constraint for efficient network training on modern GPUs limited memory, embedded systems, and mobile devices. Several studies show that feature maps (as generated after convolutional layers) are main bottleneck in this problem. Often, these mimic natural photographs sense their energy concentrated spectral domain. Although embedding CNN domain...

In this paper, we develop a RRAM compact model accounting for random telegraph noise (RTN) effect. particular, Monte Carlo method to effectively capture the behaviors of traps in tunneling gap, which can be used predict current fluctuation caused by RTN. The is validated with experimental data under various operating conditions. applied study circuit reliability efficient design space explorations.

Estimation of the frequency and duration logos in videos is important challenging advertisement industry as a way estimating impact ad purchases. Since occupy only small area videos, popular methods image retrieval could fail. This paper develops an algorithm called Video Logo Retrieval (VLR), which image-to-video based on spatial distribution local descriptors that measure distance between query (the logo) collection video images. VLR uses features to overcome weakness global feature models...

The Segment Anything Model (SAM) has exhibited outstanding performance in various image segmentation tasks. Despite being trained with over a billion masks, SAM faces challenges mask prediction quality numerous scenarios, especially real-world contexts. In this paper, we introduce novel prompt-driven adapter into SAM, namely Prompt Adapter (PA-SAM), aiming to enhance the of original SAM. By exclusively training prompt adapter, PA-SAM extracts detailed information from images and optimizes...

In the recent years, there has been a drive towards realization of topological phases beyond conventional electronic materials, including defined in more than three dimensions. We propose way to realize 2nd Chern number with photonic crystals simply made up defect resonators embedded within regular lattice resonators. particular, through novel quasiperiodic spatial modulations radii, possessing topologically nontrivial bands non-abelian berry curvature living four-dimensional synthetic space...

Deep models have demonstrated recent success in single-image dehazing. Most prior methods consider fully supervised training and learn from paired clean hazy images, where a image is synthesized based on its estimated depth map. This paradigm, however, can produce low-quality images due to inaccurate estimation, resulting poor generalization of the trained models. In this paper, we explore an alternative approach for generating clean-hazy by leveraging computer graphics. Using modern game...

Convolutional neural networks (CNNs) and vision transformers (ViTs) have achieved remarkable success in various tasks. However, many architectures do not consider interactions between feature maps from different stages scales, which may limit their performance. In this work, we propose a simple add-on attention module to overcome these limitations via multi-stage cross-scale interactions. Specifically, the proposed Multi-Stage Cross-Scale Attention (MSCSA) takes enable achieves by computing...

Recent approaches such as ControlNet offer users fine-grained spatial control over text-to-image (T2I) diffusion models. However, auxiliary modules have to be trained for each type of condition, model architecture, and checkpoint, putting them at odds with the diverse intents preferences a human designer would like convey AI models during content creation process. In this work, we present FreeControl, training-free approach controllable T2I generation that supports multiple conditions,...

Estimation of the frequency and duration logos in videos is important challenging advertisement industry as a way estimating impact ad purchases. Since occupy only small area videos, popular methods image retrieval could fail. This paper develops an algorithm called Video Logo Retrieval (VLR), which image-to-video based on spatial distribution local descriptors that measure distance between query (the logo) collection video images. VLR uses features to overcome weakness global feature-based...

For many years, the family of convolutional neural networks (CNNs) has been a workhorse in deep learning. Recently, novel CNN structures have designed to address increasingly challenging tasks. To make them work efficiently on edge devices, researchers proposed various structured network pruning strategies reduce their memory and computational cost. However, most only focus reducing number filter channels per layer without considering redundancy within individual channels. In this work, we...