WordStylist: Styled Verbatim Handwritten Text Generation with Latent Diffusion Models
Authors:
Konstantina Nikolaidou,
George Retsinas,
Vincent Christlein,
Mathias Seuret,
Giorgos Sfikas,
Elisa Barney Smith,
Hamam Mokayed,
Marcus Liwicki
Abstract:
Text-to-Image synthesis is the task of generating an image according to a specific text description. Generative Adversarial Networks have been considered the standard method for image synthesis virtually since their introduction. Denoising Diffusion Probabilistic Models are recently setting a new baseline, with remarkable results in Text-to-Image synthesis, among other fields. Aside its usefulness…
▽ More
Text-to-Image synthesis is the task of generating an image according to a specific text description. Generative Adversarial Networks have been considered the standard method for image synthesis virtually since their introduction. Denoising Diffusion Probabilistic Models are recently setting a new baseline, with remarkable results in Text-to-Image synthesis, among other fields. Aside its usefulness per se, it can also be particularly relevant as a tool for data augmentation to aid training models for other document image processing tasks. In this work, we present a latent diffusion-based method for styled text-to-text-content-image generation on word-level. Our proposed method is able to generate realistic word image samples from different writer styles, by using class index styles and text content prompts without the need of adversarial training, writer recognition, or text recognition. We gauge system performance with the Fréchet Inception Distance, writer recognition accuracy, and writer retrieval. We show that the proposed model produces samples that are aesthetically pleasing, help boosting text recognition performance, and get similar writer retrieval score as real data. Code is available at: https://github.com/koninik/WordStylist.
△ Less
Submitted 17 May, 2023; v1 submitted 29 March, 2023;
originally announced March 2023.
Beta decay of molecular tritium
Authors:
Y. -T. Lin,
T. H. Burritt,
C. Claessens,
G. Holman,
M. Kallander,
E. Machado,
L. I. Minter,
R. Ostertag,
D. S. Parno,
J. Pedersen,
D. A. Peterson,
R. G. H. Robertson,
E. B. Smith,
T. D. Van Wechel,
A. P. Vizcaya Hernández
Abstract:
The beta decay of tritium in the form of molecular TT is the basis of sensitive experiments to measure neutrino mass. The final-state electronic, vibrational, and rotational excitations modify the beta spectrum significantly, and are obtained from theory. We report measurements of the branching ratios to specific ionization states for the isotopolog HT. Two earlier, concordant measurements gave br…
▽ More
The beta decay of tritium in the form of molecular TT is the basis of sensitive experiments to measure neutrino mass. The final-state electronic, vibrational, and rotational excitations modify the beta spectrum significantly, and are obtained from theory. We report measurements of the branching ratios to specific ionization states for the isotopolog HT. Two earlier, concordant measurements gave branching ratios of HT to the bound HHe$^+$ ion of 89.5% and 93.2%, in sharp disagreement with the theoretical prediction of 55-57%, raising concerns about the theory's reliability in neutrino mass experiments. Our result, 56.5(6)%, is compatible with the theoretical expectation and disagrees strongly with the previous measurements.
△ Less
Submitted 24 May, 2020; v1 submitted 31 January, 2020;
originally announced January 2020.
