Showing 1–1 of 1 results for author: Werman, L K M

DeePaste -- Inpainting for Pasting

Authors: Levi Kassel Michael Werman

Abstract: One of the challenges of supervised learning training is the need to procure an substantial amount of tagged data. A well-known method of solving this problem is to use synthetic data in a copy-paste fashion, so that we cut objects and paste them onto relevant backgrounds. Pasting the objects naively results in artifacts that cause models to give poor results on real data. We present a new method… ▽ More One of the challenges of supervised learning training is the need to procure an substantial amount of tagged data. A well-known method of solving this problem is to use synthetic data in a copy-paste fashion, so that we cut objects and paste them onto relevant backgrounds. Pasting the objects naively results in artifacts that cause models to give poor results on real data. We present a new method for cleanly pasting objects on different backgrounds so that the dataset created gives competitive performance on real data. The main emphasis is on the treatment of the border of the pasted object using inpainting. We show state-of-the-art results both on instance detection and foreground segmentation △ Less

Submitted 26 December, 2021; v1 submitted 20 December, 2021; originally announced December 2021.