Best Paper Awards

Best Paper

Optical Flow and Mode Selection for Learning-based Video Coding
by Théo Ladune (Orange, Rennes, France), Pierrick Philippe (Orange, Rennes, France), Wassim Hamidouche (CNRS, IETR – UMR 6164 Rennes, France), Lu Zhang (CNRS, IETR – UMR 6164 Rennes, France) and Oliviér Deforges (CNRS, IETR – UMR 6164 Rennes, France)

Awards Chairs’ comment:
“Showing a performance similar to HEVC, this paper shows a successful combination of neural-network-based motion compensation and neural-network-based signal coding avoiding straight prediction error coding.”

Best Paper Runner-up

Blind reverberation time estimation from ambisonic recordings
by Andrés Pérez-López (Universitat Pompeu Fabra, Barcelona, Spain; Centre Tecnologic de Catalunya, Barcelona, Spain), Archontis Politis (Tampere University, Finland) Emilia Gómez (Universitat Pompeu Fabra, Barcelona, Spain, Centre for Advanced Studies, Joint Research Centre, European Commission, Seville, Spain)

Awards Chairs’ comment:
“Verified by experiments on synthetic data, this paper extends an algorithm from single channel reverberation time estimation to modern applications using acoustic immersion.”

Best Student Paper Awards

Best Student Paper

Key Point Agnostic Frequency-Selective Mesh-to-Grid Image Resampling using Spectral Weighting
by Viktoria Heimann, Nils Genser, André Kaup (Friedrich-Alexander University, Erlangen-Nuremberg, Germany)

Awards Chairs’ comment:
“Using an adaptive image model, this paper enhances the state of the art for image resampling required for image rotation, zoom, sheer etc. in terms of speed as well as subjective quality.”

Best Student Paper Runner-up

Convolution Autoencoder-Based Sparse Representation Wavelet for Image Classification
by Tan-Sy Nguyen, Long H. Ngo, Marie Luong, Mounir Kaaniche and Azeddine Beghdadi (Universite Sorbonne Paris Nord, France)

Awards Chairs’ comment:
“Feeding the wavelet coefficients of an image into an autoencoder for image classification instead of the image reduces the required degrees of freedom of the autoencoder and may reduce the amount of training data required without sacrificing performance.”

All awards are sponsored by YouTube

List of nominated papers

NB. The papers are visible only for registered users.