Spiral Generative Network for Image Extrapolation

Guo, Dongsheng; Liu, Hongzhi; Zhao, Haoru; Cheng, Yunhao; Song, Qingwei; Gu, Zhaorui; Zheng, Haiyong; Zheng, Bing

doi:10.1007/978-3-030-58529-7_41

Cited by 21 publications

(10 citation statements)

References 40 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Another close line of research is image extrapolation (or image "outpainting" [72,67]), i.e., predicting the surrounding context of an image given only its part. The latest approaches in this field rely on using GANs to predict an outlying image patch [21,64,69]. The fundamental difference of these methods compared to our problem design is the reliance on an input image as a starting point of the generation process.…”

Section: Related Workmentioning

confidence: 99%

Aligning Latent and Image Spaces to Connect the Unconnectable

Skorokhodov¹,

Sotnikov²,

Elhoseiny³

2021

Preprint

View full text Add to dashboard Cite

show abstract

Section: Related Workmentioning

confidence: 99%

Aligning Latent and Image Spaces to Connect the Unconnectable

Skorokhodov¹,

Sotnikov²,

Elhoseiny³

2021

Preprint

View full text Add to dashboard Cite

show abstract

“…Image Outpainting Conventional methods extend an input image to a larger seamless one; however, they require manual guidance [4,7,87] or image sets of the same scene category [30,59,70]. By contrast, learning-based methods synthesize large images with novel textures that do not exist in the input perspective image [19,20,31,32,35,48,56,75,83]. Some approaches focus on driving scenes [74,85] or synthesize panorama-like landscapes with iterative extension or multiple perspective images [21,33,61,79].…”

Section: Related Workmentioning

confidence: 99%

BIPS: Bi-modal Indoor Panorama Synthesis via Residual Depth-aided Adversarial Learning

Changgyoon¹,

Cho²,

Park³

et al. 2021

Preprint

View full text Add to dashboard Cite

Providing omnidirectional depth along with RGB information is important for numerous applications, e.g., VR/AR. However, as omnidirectional RGB-D data is not always available, synthesizing RGB-D panorama data from limited information of a scene can be useful. Therefore, some prior works tried to synthesize RGB panorama images from perspective RGB images; however, they suffer from limited image quality and can not be directly extended for RGB-D panorama synthesis. In this paper, we study a new problem: RGB-D panorama synthesis under the arbitrary configurations of cameras and depth sensors. Accordingly, we propose a novel bi-modal (RGB-D) panorama synthesis (BIPS) framework. Especially, we focus on indoor environments where the RGB-D panorama can provide a complete 3D model for many applications. We design a generator that fuses the bi-modal information and train it with residualaided adversarial learning (RDAL). RDAL allows to synthesize realistic indoor layout structures and interiors by jointly inferring RGB panorama, layout depth, and residual depth. In addition, as there is no tailored evaluation metric for RGB-D panorama synthesis, we propose a novel metric to effectively evaluate its perceptual quality. Extensive experiments show that our method synthesizes high-quality indoor RGB-D panoramas and provides realistic 3D indoor models than prior methods. Code will be released upon acceptance.

show abstract

“…Previous works deal with this task from different aspects. For example, [17] introduces a semantic regeneration network that learns semantic features from a smallsize input and generates a full image; [3] proposes Spiral-Net which performs image outpainting in a spiral fashion, growing from an input sub-image along a spiral curve to an expanded full image. Although the above two research topics are highly correlated to our task of wide-range image blending, none of the existing approaches is able to generate intermediate region that bridges two different images with smooth transition and exquisite details.…”

Section: Related Workmentioning

confidence: 99%

“…On the other hand, even if we can apply the existing image outpainting model (e.g. [3,15,17,19]) respectively on the two input photos for generating the image content of the intermediate region, there is no guarantee to have seamless composition between those two extrapolation results. Later in this paper, we will provide experimental evidence to demonstrate that directly adopting inpainting or outpainting methods without any modification leads to poor results under the problem scenario of wide-range image blending.…”

Section: Introductionmentioning

confidence: 99%

Bridging the Visual Gap: Wide-Range Image Blending

Chang

Chiu

2021

Preprint

View full text Add to dashboard Cite

Figure 1: Given two different input photos (the first and the second columns), our wide-range image blending model is able to seamlessly blend them into a novel panoramic image by generating smooth transition in the intermediate region between them. Here we show several examples of the resultant panoramas, where the result in the third column is produced by putting the first input photo (highlighted by blue dashed lines) on the left and the second input photo (highlighted by red dashed lines) on the right, while the fourth column is obtained with using opposite spatial arrangement.

show abstract

Spiral Generative Network for Image Extrapolation

Cited by 21 publications

References 40 publications

Aligning Latent and Image Spaces to Connect the Unconnectable

Aligning Latent and Image Spaces to Connect the Unconnectable

BIPS: Bi-modal Indoor Panorama Synthesis via Residual Depth-aided Adversarial Learning

Bridging the Visual Gap: Wide-Range Image Blending

Contact Info

Product

Resources

About