arXiv subjects

Jongchan Park

Publications and source records attributed to Jongchan Park.

31 records · Page 2Linked to original sources

arXiv2021

Exploiting Features with Split-and-Share Module

Deep convolutional neural networks (CNNs) have shown state-of-the-art performances in various computer vision tasks. Advances on CNN architectures have focused mainly on designing convolutional blocks of the feature extractors, but less on the classifiers that exploit extracted features. In this work, we propose Split-and-Share Module (SSM),a classifier that splits a given feature into parts, which are partially shared by multiple sub-classifiers. Our intuition is that the more the features are shared, the more common they will become, and SSM can encourage such structural characteristics in the split features. SSM can be easily integrated into any architecture without bells and whistles. We have extensively validated the efficacy of SSM on ImageNet-1K classification task, andSSM has shown consistent and significant improvements over baseline architectures. In addition, we analyze the effect of SSM using the Grad-CAM visualization.

cs.CV↗

arXiv2021

Reducing Domain Gap by Reducing Style Bias

Convolutional Neural Networks (CNNs) often fail to maintain their performance when they confront new test domains, which is known as the problem of domain shift. Recent studies suggest that one of the main causes of this problem is CNNs' strong inductive bias towards image styles (i.e. textures) which are sensitive to domain changes, rather than contents (i.e. shapes). Inspired by this, we propose to reduce the intrinsic style bias of CNNs to close the gap between domains. Our Style-Agnostic Networks (SagNets) disentangle style encodings from class categories to prevent style biased predictions and focus more on the contents. Extensive experiments show that our method effectively reduces the style bias and makes the model more robust under domain shift. It achieves remarkable performance improvements in a wide range of cross-domain tasks including domain generalization, unsupervised domain adaptation, and semi-supervised domain adaptation on multiple datasets.

cs.CV↗

arXiv2020

Learning Visual Context by Comparison

Finding diseases from an X-ray image is an important yet highly challenging task. Current methods for solving this task exploit various characteristics of the chest X-ray image, but one of the most important characteristics is still missing: the necessity of comparison between related regions in an image. In this paper, we present Attend-and-Compare Module (ACM) for capturing the difference between an object of interest and its corresponding context. We show that explicit difference modeling can be very helpful in tasks that require direct comparison between locations from afar. This module can be plugged into existing deep learning models. For evaluation, we apply our module to three chest X-ray recognition tasks and COCO object detection & segmentation tasks and observe consistent improvements across tasks. The code is available at https://github.com/mk-minchul/attend-and-compare.

cs.CV↗

arXiv2020

Sequential Feature Filtering Classifier

We propose Sequential Feature Filtering Classifier (FFC), a simple but effective classifier for convolutional neural networks (CNNs). With sequential LayerNorm and ReLU, FFC zeroes out low-activation units and preserves high-activation units. The sequential feature filtering process generates multiple features, which are fed into a shared classifier for multiple outputs. FFC can be applied to any CNNs with a classifier, and significantly improves performances with negligible overhead. We extensively validate the efficacy of FFC on various tasks: ImageNet-1K classification, MS COCO detection, Cityscapes segmentation, and HMDB51 action recognition. Moreover, we empirically show that FFC can further improve performances upon other techniques, including attention modules and augmentation techniques. The code and models will be publicly available.

cs.CV↗

arXiv2018

BAM: Bottleneck Attention Module

Recent advances in deep neural networks have been developed via architecture search for stronger representational power. In this work, we focus on the effect of attention in general deep neural networks. We propose a simple and effective attention module, named Bottleneck Attention Module (BAM), that can be integrated with any feed-forward convolutional neural networks. Our module infers an attention map along two separate pathways, channel and spatial. We place our module at each bottleneck of models where the downsampling of feature maps occurs. Our module constructs a hierarchical attention at bottlenecks with a number of parameters and it is trainable in an end-to-end manner jointly with any feed-forward models. We validate our BAM through extensive experiments on CIFAR-100, ImageNet-1K, VOC 2007 and MS COCO benchmarks. Our experiments show consistent improvement in classification and detection performances with various models, demonstrating the wide applicability of BAM. The code and models will be publicly available.

cs.CV↗

arXiv2018

CBAM: Convolutional Block Attention Module

We propose Convolutional Block Attention Module (CBAM), a simple yet effective attention module for feed-forward convolutional neural networks. Given an intermediate feature map, our module sequentially infers attention maps along two separate dimensions, channel and spatial, then the attention maps are multiplied to the input feature map for adaptive feature refinement. Because CBAM is a lightweight and general module, it can be integrated into any CNN architectures seamlessly with negligible overheads and is end-to-end trainable along with base CNNs. We validate our CBAM through extensive experiments on ImageNet-1K, MS~COCO detection, and VOC~2007 detection datasets. Our experiments show consistent improvements in classification and detection performances with various models, demonstrating the wide applicability of CBAM. The code and models will be publicly available.

cs.CV↗

arXiv2018

Distort-and-Recover: Color Enhancement using Deep Reinforcement Learning

Learning-based color enhancement approaches typically learn to map from input images to retouched images. Most of existing methods require expensive pairs of input-retouched images or produce results in a non-interpretable way. In this paper, we present a deep reinforcement learning (DRL) based method for color enhancement to explicitly model the step-wise nature of human retouching process. We cast a color enhancement process as a Markov Decision Process where actions are defined as global color adjustment operations. Then we train our agent to learn the optimal global enhancement sequence of the actions. In addition, we present a 'distort-and-recover' training scheme which only requires high-quality reference images for training instead of input and retouched image pairs. Given high-quality reference images, we distort the images' color distribution and form distorted-reference image pairs for training. Through extensive experiments, we show that our method produces decent enhancement results and our DRL approach is more suitable for the 'distort-and-recover' training scheme than previous supervised approaches. Supplementary material and code are available at https://sites.google.com/view/distort-and-recover/

cs.CV↗

arXiv2016

Time-reversing a monochromatic subwavelength optical focus by optical phase conjugation of multiply-scattered light

Due to its time-reversal nature, optical phase conjugation generates a monochromatic light wave which retraces its propagation paths. Here, we demonstrate the regeneration of a subwavelength optical focus by phase conjugation. Monochromatic light from a subwavelength source is scattered by random nanoparticles, and the scattered light is phase conjugated at the far-field region by coupling its wavefront into a single-mode optical reflector using a spatial light modulator. Then the conjugated beam retraces its propagation paths and forms a refocus on the source at the subwavelength scale. This is the first direct experimental realization of subwavelength focusing beyond the diffraction limit with far-field time reversal in the optical domain.

physics.optics↗

arXiv2016

Scattering optical elements: stand-alone optical elements exploiting multiple light scattering

Optical design and fabrication techniques are crucial for making optical elements. From conventional lenses to diffractive optical elements, and to recent metasurfaces, various types of optical elements have been proposed to manipulate light where optical materials are fabricated into desired structures. Here, we propose a scattering optical element (SOE) which exploits multiple light scattering and wavefront shaping. Instead of fabricating optical materials, the SOE consists of a disordered medium and a photopolymer-based wavefront recorder, with shapes the wavefront of impinging light on demand. With the proposed stand-alone SOEs, we experimentally demonstrate control of various properties of light, including intensity, polarisation, spectral frequency and near field. Due to the tremendous freedom brought about by disordered media, the proposed approach will provide unexplored routes to manipulate arbitrary optical fields in stand-alone optical elements.

physics.optics↗

arXiv2015

Optogenetic control of cell signaling pathway through scattering skull using wavefront shaping

We introduce a non-invasive approach for optogenetic regulation in biological cells through highly scattering skull tissue using wavefront shaping. The wavefront of the incident light was systematically controlled using a spatial light modulator in order to overcome multiple light-scattering in a mouse skull layer and to focus light on the target cells. We demonstrate that illumination with shaped waves enables spatiotemporal regulation of intracellular Ca2+ level at the individual-cell level.

physics.optics↗

arXiv2015

Recent advances in wavefront shaping techniques for biomedical applications

Due to the highly inhomogeneous distributions of refractive indexes, light propagation in complex media such as biological tissue experiences multiple light scattering events. The suppression and control of multiple light scattering events are investigated because they offer the possibility of optical focusing and imaging through biological tissues, and they may open new avenues for diagnosis and treatment of several human diseases. In order to provide insight into how new optical techniques can address the issues of multiple light scattering in biomedical applications, the recent progress in optical wavefront-shaping techniques is summarized.

physics.optics↗

arXiv2015

Focusing through turbid media by polarization modulation

We demonstrate that polarization modulation of an illumination beam can effectively control the spatial profile of the light transmitted through turbid media. Since the transmitted electric fields are completely mingled in turbid media, polarization states of an illumination beam can be effectively used to control the propagation of light through turbid media. Numerical simulations were performed which agree with experimental results obtained using a commercial in-plane switching liquid-crystal display for modulating the input polarization states.

physics.optics↗

arXiv2015

Measuring optical transmission matrices by wavefront shaping

We introduce a simple but practical method to measure the optical transmission matrix (TM) of complex media. The optical TM of a complex medium is obtained by modulating the wavefront of a beam impinging on the complex medium and imaging the transmitted full-field speckle intensity patterns. Using the retrieved TM, we demonstrate the generation and linear combination of multiple foci on demand through the complex medium. This method will be used as a versatile tool for coherence control of waves through turbid media.

physics.optics↗