2019
DOI: 10.48550/arxiv.1911.07783
|View full text |Cite
Preprint
|
Sign up to set email alerts
|

AIM 2019 Challenge on Real-World Image Super-Resolution: Methods and Results

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
1
1
1

Citation Types

0
5
0

Year Published

2019
2019
2021
2021

Publication Types

Select...
3

Relationship

1
2

Authors

Journals

citations
Cited by 3 publications
(5 citation statements)
references
References 0 publications
0
5
0
Order By: Relevance
“…For image perceptual quality evaluations we refer to the results of MGBPv1 and MGBPv2 in the first challenge on perceptual image SR, PIRM-SR 2018 [32], and the first extreme SR challenge, AIM Extreme-SR 2019 [40], respectively. Table V shows our best average scores and rankings in the PIRM-SR Challenge 2018 [32] for Region 1 (RM SE 11.5), Region 2 (11.5 < RM SE 12.5) and Region 3 (12.5 < RM SE 16).…”
Section: B Perceptual Evaluationmentioning
confidence: 99%
See 2 more Smart Citations
“…For image perceptual quality evaluations we refer to the results of MGBPv1 and MGBPv2 in the first challenge on perceptual image SR, PIRM-SR 2018 [32], and the first extreme SR challenge, AIM Extreme-SR 2019 [40], respectively. Table V shows our best average scores and rankings in the PIRM-SR Challenge 2018 [32] for Region 1 (RM SE 11.5), Region 2 (11.5 < RM SE 12.5) and Region 3 (12.5 < RM SE 16).…”
Section: B Perceptual Evaluationmentioning
confidence: 99%
“…14. Evolution of perceptual and fidelity metrics when moving the input noise amplitude from W = 0 to W = 1 in our MGBPv2 16× system used for AIM Extreme-SR 2019 [40]. optimal as we are not enforcing optimality.…”
Section: Analysis a Perception-distortion Trajectorymentioning
confidence: 99%
See 1 more Smart Citation
“…In particular, they extend the MGBPv2 network [36] that was designed to scale efficiently for the task of extreme image SR and was successfully used in the 2019-AIM Extreme Image SR competition [33] to win the Perceptual track of that challenge. For this challenge they redesigned the MGBPv2 network to use 3D-convolutions strided in space.…”
Section: Boe-iot-aibdmentioning
confidence: 99%
“…often leads to poor results [7,22]. The real-world SR problem is then to super-resolve LR images downsampled by unknown, realistic image degradations [24].…”
Section: Introductionmentioning
confidence: 99%