DeepGIN: Deep Generative Inpainting Network for Extreme Image Inpainting

Chu-Tak Li (1) , Wan-Chi Siu (1) , Zhi-Song Liu (2) , Li-Wen Wang (1) , Daniel Pak-Kong Lun (1)

1 POLYU - The Hong Kong Polytechnic University [Hong Kong]
2 LIX - Laboratoire d'informatique de l'École polytechnique [Palaiseau]

The degree of difficulty in image inpainting depends on the types and sizes of the missing parts. Existing image inpainting approaches usually encounter difficulties in completing the missing parts in the wild with pleasing visual and contextual results as they are trained for either dealing with one specific type of missing patterns (mask) or unilaterally assuming the shapes and/or sizes of the masked areas. We propose a deep generative inpainting network, named DeepGIN, to handle various types of masked images. We design a Spatial Pyramid Dilation (SPD) ResNet block to enable the use of distant features for reconstruction. We also employ Multi-Scale Self-Attention (MSSA) mechanism and Back Projection (BP) technique to enhance our inpainting results. Our Deep-GIN outperforms the state-of-the-art approaches generally, including two publicly available datasets (FFHQ and Oxford Buildings), both quantitatively and qualitatively. We also demonstrate that our model is capable of completing masked images in the wild.

Mots clés: Image Inpainting; Attention; Back Projection

Domaines: Informatique [cs]

Chu-Tak Li, Wan-Chi Siu, Zhi-Song Liu, Li-Wen Wang, Daniel Pak-Kong Lun. DeepGIN: Deep Generative Inpainting Network for Extreme Image Inpainting. The 2020 European Conference on Computer Vision, Sep 2020, Online, United Kingdom. ⟨hal-03014865⟩