Search Results for author: Aljosa Smolic

Found 29 papers, 10 papers with code

View Sub-sampling and Reconstruction for Efficient Light Field Compression

no code implementations12 Aug 2022 Yang Chen, Martin Alain, Aljosa Smolic

In this work, we study different sub-sampling and reconstruction strategies for light field compression.

KinePose: A temporally optimized inverse kinematics technique for 6DOF human pose estimation with biomechanical constraints

no code implementations26 Jul 2022 Kevin Gildea, Clara Mercadal-Baudart, Richard Blythman, Aljosa Smolic, Ciaran Simms

With frame-by-frame IK we obtain low errors in the case of bent elbows and knees, however, motion sequences with phases of extended/straight limbs results in ambiguity in twist angle.

3D Human Pose Estimation

Image Aesthetics Assessment Using Graph Attention Network

1 code implementation26 Jun 2022 Koustav Ghosal, Aljosa Smolic

Second, we propose a graph neural network architecture that takes this feature-graph and captures the semantic relationship between the different regions of the input image using visual attention.

Graph Attention

TEAM-Net: Multi-modal Learning for Video Action Recognition with Partial Decoding

1 code implementation17 Oct 2021 Zhengwei Wang, Qi She, Aljosa Smolic

Video compression (e. g., H. 264, MPEG-4) reduces superfluous information by representing the raw video stream using the concept of Group of Pictures (GOP).

Action Recognition Video Compression

Spectral analysis of re-parameterized light fields

no code implementations12 Oct 2021 Martin Alain, Aljosa Smolic

In this paper, we study the spectral properties of re-parameterized light field.

Frequency Domain Loss Function for Deep Exposure Correction of Dark Images

no code implementations22 Apr 2021 Ojasvi Yadav, Koustav Ghosal, Sebastian Lutz, Aljosa Smolic

We address the problem of exposure correction of dark, blurry and noisy images captured in low-light conditions in the wild.

Image Denoising Translation

Foreground color prediction through inverse compositing

no code implementations24 Mar 2021 Sebastian Lutz, Aljosa Smolic

In natural image matting, the goal is to estimate the opacity of the foreground object in the image.

Image Matting

ACTION-Net: Multipath Excitation for Action Recognition

1 code implementation CVPR 2021 Zhengwei Wang, Qi She, Aljosa Smolic

To this end, we propose a spAtio-temporal, Channel and moTion excitatION (ACTION) module consisting of three paths: Spatio-Temporal Excitation (STE) path, Channel Excitation (CE) path, and Motion Excitation (ME) path.

Action Recognition

Self-supervised Light Field View Synthesis Using Cycle Consistency

no code implementations12 Aug 2020 Yang Chen, Martin Alain, Aljosa Smolic

In order to enhance the angular resolution of light fields, view synthesis methods can be utilized to generate dense intermediate views from sparse light field input.

A Study of Efficient Light Field Subsampling and Reconstruction Strategies

no code implementations11 Aug 2020 Yang Chen, Martin Alain, Aljosa Smolic

Limited angular resolution is one of the main obstacles for practical applications of light fields.


Fast and Accurate Optical Flow based Depth Map Estimation from Light Fields

no code implementations11 Aug 2020 Yang Chen, Martin Alain, Aljosa Smolic

Depth map estimation is a crucial task in computer vision, and new approaches have recently emerged taking advantage of light fields, as this new imaging modality captures much more information about the angular direction of light rays compared to common approaches based on stereoscopic images or multi-view.

Depth Estimation Optical Flow Estimation

A Study on Visual Perception of Light Field Content

no code implementations7 Aug 2020 Ailbhe Gill, Emin Zerman, Cagri Ozcinar, Aljosa Smolic

We conducted perception experiments displaying them to users in various ways and collected corresponding visual attention data.

Sub-Pixel Back-Projection Network For Lightweight Single Image Super-Resolution

1 code implementation3 Aug 2020 Supratik Banerjee, Cagri Ozcinar, Aakanksha Rana, Aljosa Smolic, Michael Manzke

To tackle this problem, in this paper, we study reducing the number of parameters and computational cost of CNN-based SISR methods while maintaining the accuracy of super-resolution reconstruction performance.

Image Super-Resolution Single Image Super Resolution

CatNet: Class Incremental 3D ConvNets for Lifelong Egocentric Gesture Recognition

1 code implementation20 Apr 2020 Zhengwei Wang, Qi She, Tejo Chalasani, Aljosa Smolic

Egocentric gestures are the most natural form of communication for humans to interact with wearable devices such as VR/AR helmets and glasses.

Gesture Recognition Video Recognition

STaDA: Style Transfer as Data Augmentation

no code implementations3 Sep 2019 Xu Zheng, Tejo Chalasani, Koustav Ghosal, Sebastian Lutz, Aljosa Smolic

The success of training deep Convolutional Neural Networks (CNNs) heavily depends on a significant amount of labelled data.

Classification Data Augmentation +3

A Geometry-Sensitive Approach for Photographic Style Classification

no code implementations3 Sep 2019 Koustav Ghosal, Mukta Prasad, Aljosa Smolic

We further introduce a two-column CNN architecture that performs better than the state-of-the-art (SoA) in photographic style classification.

Classification Data Augmentation +2

Aesthetic Image Captioning From Weakly-Labelled Photographs

1 code implementation29 Aug 2019 Koustav Ghosal, Aakanksha Rana, Aljosa Smolic

Aesthetic image captioning (AIC) refers to the multi-modal task of generating critical textual feedbacks for photographs.

Aesthetic Image Captioning

ColorNet -- Estimating Colorfulness in Natural Images

1 code implementation22 Aug 2019 Emin Zerman, Aakanksha Rana, Aljosa Smolic

Measuring the colorfulness of a natural or virtual scene is critical for many applications in image processing field ranging from capturing to display.

Image Classification

Super-resolution of Omnidirectional Images Using Adversarial Learning

1 code implementation12 Aug 2019 Cagri Ozcinar, Aakanksha Rana, Aljosa Smolic

An omnidirectional image (ODI) enables viewers to look in every direction from a fixed point through a head-mounted display providing an immersive experience compared to that of a standard image.


Deep Tone Mapping Operator for High Dynamic Range Images

no code implementations12 Aug 2019 Aakanksha Rana, Praveer Singh, Giuseppe Valenzise, Frederic Dufaux, Nikos Komodakis, Aljosa Smolic

In this paper, we address this problem by proposing a fast, parameter-free and scene-adaptable deep tone mapping operator (DeepTMO) that yields a high-resolution and high-subjective quality tone mapped output.

Tone Mapping

A Fourier Disparity Layer representation for Light Fields

no code implementations21 Jan 2019 Mikael Le Pendu, Christine Guillemot, Aljosa Smolic

In this paper, we present a new Light Field representation for efficient Light Field processing and rendering called Fourier Disparity Layers (FDL).


Dynamic Environment Mapping for Augmented Reality Applications on Mobile Devices

no code implementations21 Sep 2018 Rafael Monroy, Matis Hudon, Aljosa Smolic

In this work, we present a novel pipeline that allows for coupled environment acquisition and virtual object rendering on a mobile device equipped with a depth sensor.

Egocentric Gesture Recognition for Head-Mounted AR devices

no code implementations16 Aug 2018 Tejo Chalasani, Jan Ondrej, Aljosa Smolic

Natural interaction with virtual objects in AR/VR environments makes for a smooth user experience.

Data Augmentation Gesture Recognition

A Pipeline for Lenslet Light Field Quality Enhancement

no code implementations16 Aug 2018 Pierre Matysiak, Mairéad Grogan, Mikaël Le Pendu, Martin Alain, Aljosa Smolic

In recent years, light fields have become a major research topic and their applications span across the entire spectrum of classical image processing.


AlphaGAN: Generative adversarial networks for natural image matting

1 code implementation26 Jul 2018 Sebastian Lutz, Konstantinos Amplianitis, Aljosa Smolic

We present the first generative adversarial network (GAN) for natural image matting.

Image Matting

SalNet360: Saliency Maps for omni-directional images with CNN

no code implementations19 Sep 2017 Rafael Monroy, Sebastian Lutz, Tejo Chalasani, Aljosa Smolic

The prediction of Visual Attention data from any kind of media is of valuable use to content creators and used to efficiently drive encoding algorithms.

Saliency Prediction

Cannot find the paper you are looking for? You can Submit a new open access paper.