Learnable Visual Markers
Oleg Grinchuk, Vadim Lebedev, Victor Lempitsky

TL;DR
This paper introduces a novel method for designing visual markers using deep generative networks, enabling adaptive, stylized, and robust markers for augmented reality and robotics that retain information despite distortions.
Contribution
The authors develop a joint training framework for synthesizing and recognizing visual markers with deep networks, incorporating stylization and distortion robustness, advancing marker design capabilities.
Findings
Markers retain long enough bit strings for practical use
Markers can be stylized to match texture prototypes
The approach offers insights into pattern recognition by ConvNets
Abstract
We propose a new approach to designing visual markers (analogous to QR-codes, markers for augmented reality, and robotic fiducial tags) based on the advances in deep generative networks. In our approach, the markers are obtained as color images synthesized by a deep network from input bit strings, whereas another deep network is trained to recover the bit strings back from the photos of these markers. The two networks are trained simultaneously in a joint backpropagation process that takes characteristic photometric and geometric distortions associated with marker fabrication and marker scanning into account. Additionally, a stylization loss based on statistics of activations in a pretrained classification network can be inserted into the learning in order to shift the marker appearance towards some texture prototype. In the experiments, we demonstrate that the markers obtained using…
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsAdvanced Image and Video Retrieval Techniques · Augmented Reality Applications · Face recognition and analysis
