1

In this paper we adapt a localization algorithm to work indoors with the help of panoptic segmentation and feature description. The method was tested in multiple real-world scenarios using a hand-held sensor.

In this paper we model accents by adding extra conditioning to a diffusion model. We also convert the accent by using saliency to achieve accent inpainting in melspectrograms.

In this paper we model accented speech by providing aditional data via voice augmentation. This leads to reliable accent generation without a TTS frontend.

In this paper we propose a new method for universal vocoding allowing us to substitute 43 individual vocoding models speaking 20 languages in 17 styles of different age and geneder.

In this paper we investigate state-of-the-art dense correspondence networks and their ability to process high-resolution images. This leads to discovery of the existence of an optimal resolution 𝕏.

In this paper we investigate state-of-the-art dense correspondence networks and their ability to process high-resolution images. This leads to discovery of the existence of an optimal resolution 𝕏.

In this paper we propose a new SLAM method with a novel risk alignment prediction method that verifies the registration.

In this paper we propose a novel LIDAR localization method that works in challenging natural environments.

In this paper we propose a novel method for assessing the risk of registration failure when aligning two point clouds.