Super-Resolution Erlangen (SupER)

2020

Licence: CC BY (Attribution)

Details

Description

Capturing ground truth data to benchmark super-resolution (SR) is challenging. Therefore, current quantitative studies are mainly evaluated on simulated data artificially sampled from ground truth images. We argue that such evaluations overestimate the actual performance of SR methods compared to their behavior on real images. Toward bridging this simulated-to-real gap, we introduce the Super-Resolution Erlangen (SupER) database, the first comprehensive laboratory SR database of all-real acquisitions with pixel-wise ground truth. It consists of more than 80k images of 14 scenes combining different facets: CMOS sensor noise, real sampling at four resolution levels, nine scene motion types, two photometric conditions, and lossy video coding at five levels. As such, the database exceeds existing benchmarks by an order of magnitude in quality and quantity. This work also benchmarks 19 popular single-image and multi-frame algorithms on our data. The benchmark comprises a quantitative study by exploiting ground truth data and qualitative evaluations in a large-scale observer study. We also rigorously investigate agreements between both evaluations from a statistical perspective. One interesting result is that top-performing methods on simulated data may be surpassed by others on real data. Our insights can spur further algorithm development, and the publicy available dataset can foster future evaluations. For details, please refer to the ReadMe.md file.

Size: 733613 MB / MiB

Creators/Owners

Download

Name Size
truck-budha-duck.zip

md5sum: b224fc189e26ff8da082fd05191ca10d

2.02 GB Download
timeMeasurements.zip

md5sum: 8484c57ea4978d30bb63fbdc825ba34f

28.05 KB Download
tea-bottles.zip

md5sum: 0d1212f39b11622a65137b8799b19003

2.25 GB Download
SupER.zip

md5sum: e10bea0302dfab9b610f16b8c5a08475

338.31 MB Download
results_sr9.zip

md5sum: b3595716669ac7270a84d2971f5925af

110.04 GB Download
results_sr8.zip

md5sum: 6465c7782aaa636d480a02953d0afc65

43.47 GB Download
results_sr7.zip

md5sum: 74a228952fbcd2fe8a952f447476fca8

41.75 GB Download
results_sr6.zip

md5sum: 1317b9811fc9cb5f6889c6fb65fa13b0

43.76 GB Download
results_sr5.zip

md5sum: b520cc8682fa9b2f590be019613bb358

43.75 GB Download
results_sr4.zip

md5sum: 9e6d4c55ec1dae13be2a1395e3eb0a02

27.89 GB Download
results_sr3.zip

md5sum: 8ee282f5d273e31d34f316e3fd3b0ac8

20.87 GB Download
results_sr20.zip

md5sum: 59c235316a453414addaa81e623593ce

21.41 GB Download
results_sr2.zip

md5sum: ce9f7e17b5f157e9a39e3c65af13fd78

7.90 GB Download
results_sr19.zip

md5sum: ce49dba38436c64a479421dd4c0a8828

20.87 GB Download
results_sr18.zip

md5sum: 363fac181772d613a17bb500aeeb2ce3

20.86 GB Download
results_sr17.zip

md5sum: 884b192ec9710c1a334da11b54d030b7

46.43 GB Download
results_sr16.zip

md5sum: 20d15c97349732d898378f2ffb57fe5e

37.06 GB Download
results_sr15.zip

md5sum: 0802d751f06d2efdc71cd0b9f4d7708d

33.04 GB Download
results_sr14.zip

md5sum: 1726a2a228a38f2ee6e751cce6c375d9

36.75 GB Download
results_sr13.zip

md5sum: ee2ae2d673f703cd0cb4908e06f6d23a

33.25 GB Download
results_sr11.zip

md5sum: c1172c877178da84658bf77afb8db53c

37.00 GB Download
results_sr10.zip

md5sum: 621d5033710391939dc030491b351b41

36.95 GB Download
results_sr1.zip

md5sum: 19ff7e18928e6a3711128e59250c69de

7.35 GB Download
results_sr0.zip

md5sum: 49a76559df6e8ebc0973f69934b6e408

9.15 GB Download
ReadMe.md

md5sum: da28c50c9d94c6f5fe371bd4b700db6f

4.75 KB Download
pencils.zip

md5sum: b5c9f6d5f4ad56480feb0d8d48b650bf

1.84 GB Download
observerStudy.zip

md5sum: 4d84c57ef88793f0ad147034f4ec6f51

21.85 MB Download
newspapers.zip

md5sum: 69c591b2b7d7780eccb205b5ed2bd11a

2.97 GB Download
loader.zip

md5sum: fb26926f52f5501cc16ce1cf4e57c1ae

2.09 GB Download
globe.zip

md5sum: ebca4b432445227445fcd2a50c885465

2.27 GB Download
globe-fast.zip

md5sum: 05d6cb2df97887b695672ff7eddcdf4a

2.31 GB Download
games.zip

md5sum: c366cfaf315ac4bc83b6f3f4335f5c8b

2.39 GB Download
dolls.zip

md5sum: 06f41ecf1a8a8f223a151c126cb2cf39

2.19 GB Download
coffee.zip

md5sum: e5a3c88373dac56b9ac7829735abe90f

1.87 GB Download
bookshelf.zip

md5sum: 276bab007f18a8e9714d132296868919

2.17 GB Download
books-and-papers.zip

md5sum: 4c234a4e28716d2affc1812474da0db8

2.22 GB Download
banknotes.zip

md5sum: aff30be9f43ae1c772e5c028f0fbb75a

3.06 GB Download
results_sr12.zip

md5sum: 9106422a614d90904710fc1f7819160f

4.63 GB Download
porsche.zip

md5sum: 84961e77d83031d3d1e102049dea1088

2.26 GB Download

Debug: Alles

Owner: T. Köhler, M. Bätz, F. Naderi Boldaji, A. Kaup, A. Maier, C. Riess, J. Seiler
Datum: None
Year: 2020
Beschreibung: Capturing ground truth data to benchmark super-resolution (SR) is challenging. Therefore, current quantitative studies are mainly evaluated on simulated data artificially sampled from ground truth images. We argue that such evaluations overestimate the actual performance of SR methods compared to their behavior on real images. Toward bridging this simulated-to-real gap, we introduce the Super-Resolution Erlangen (SupER) database, the first comprehensive laboratory SR database of all-real acquisitions with pixel-wise ground truth. It consists of more than 80k images of 14 scenes combining different facets: CMOS sensor noise, real sampling at four resolution levels, nine scene motion types, two photometric conditions, and lossy video coding at five levels. As such, the database exceeds existing benchmarks by an order of magnitude in quality and quantity. This work also benchmarks 19 popular single-image and multi-frame algorithms on our data. The benchmark comprises a quantitative study by exploiting ground truth data and qualitative evaluations in a large-scale observer study. We also rigorously investigate agreements between both evaluations from a statistical perspective. One interesting result is that top-performing methods on simulated data may be surpassed by others on real data. Our insights can spur further algorithm development, and the publicy available dataset can foster future evaluations. For details, please refer to the ReadMe.md file.
Subject:
Verf:
Publ-Datum: None
Lizenz: 248357759
Lizenz DE: CC BY (Namensnennung)
Lizenz Other: 2.0 International
Datentyp:
Anderer Datentyp:
Beschreibung Zugang:
Groesse: 733613
Einheit: MB / MiB
EOrgs: <QuerySet []>
FOBE: <QuerySet []>
Publications: <QuerySet []>
Projects: <QuerySet []>
Data Elements: <QuerySet [<DataElement: DataElement object (179)>, <DataElement: DataElement object (178)>, <DataElement: DataElement object (177)>, <DataElement: DataElement object (176)>, <DataElement: DataElement object (175)>, <DataElement: DataElement object (174)>, <DataElement: DataElement object (173)>, <DataElement: DataElement object (172)>, <DataElement: DataElement object (171)>, <DataElement: DataElement object (170)>, <DataElement: DataElement object (169)>, <DataElement: DataElement object (168)>, <DataElement: DataElement object (167)>, <DataElement: DataElement object (166)>, <DataElement: DataElement object (165)>, <DataElement: DataElement object (164)>, <DataElement: DataElement object (163)>, <DataElement: DataElement object (162)>, <DataElement: DataElement object (161)>, <DataElement: DataElement object (160)>, '...(remaining elements truncated)...']>