Final Leaderboard
The final leaderboard is based on the results that have been submitted until March 17, 23:59 (GMT 0+).The people's leaderboard is based on the people's opinion and the photographer's leaderboard is based on the professinal photographer's opinion, which is explained in more detail here.
The leaderboard was formed by including every submission in 2750 comparisons using the Toloka platform. The score mean multiplied by 2750 gives the number of actual votes that a particular submission received.
People's choice
Rank | Team | Member(s) | Affiliation | Score mean |
---|---|---|---|---|
1 | IVLTeam | Simone Zini, Claudio Rota, Marco Buzzelli, Simone Bianco, Raimondo Schettini | University of Milano - Bicocca | 0.67 |
2 | DH_ImageAlgo | Chunxia Lei, Tingniao Wang, Song Wang | Zhejiang Dahua Technology | 0.645 |
3 | MiAlgo | Shuai Liu | Xiaomi | 0.626 |
4 | BSSC | Lu Xu, Chao Zhang, Yasi Wang, Chaoqun Zhuang | SRCB | 0.606 |
5 | DH-AISP | Jin Guo, Yangfan Sun, Tianli Liu, Dejun Hao | Zhejiang Dahua Technology | 0.583 |
6 | Photoshop image enhancement | Maria Efimova | Organizers Team | 0.491 |
7 | OzUVGL | Furkan Kınlı, Barış Özcan, Furkan Kıraç | Özyeğin University | 0.453 |
8 | The Majestic Mavericks | Hyerin Chung, Nakyung Lee, Sung Keun Kwak | CJ OliveNetworks | 0.444 |
9 | JMUCVLAB | Marcos Conde, Tim Seizinger, Florin Vasluianu, Omar Elezabi | University of Wuerzburg, CVLAB | 0.439 |
10 | NTU607 | Wei-Ting Chen, Hao-Hsiang Yang, Zhi-Kai Huang, Hua-En Chang, Chia-Hsuan Hsieh, I-Hsiang Chen, Yi-Chung Chen, Yuan-Chun Chiang | Stanford University, Google, National Taiwan University, ServiceNow | 0.376 |
- | baseline | - | Organizers Team | 0.345 |
Photographer's choice
Small glossary of the photography terms used in the recent professional photographer's comments:
- ‘Punch’, ‘punchy’: a characterization of the visual impact of an image, it is one of the most powerful factors that drives the mood of the users. This term shows to what degree the tones and the contrasts between light and dark grab the viewer’s attention. The following illustrative comparisons could help in understanding: it is a difference between a tune being played nicely on one piano and between the same tune being played by a concert orchestra - they are both technically correct, but the latter one is significantly more impactful; or, it is a difference between driving down the same road in an old VW Beetle versus driving down in a brand new Porsche. They both get you to your destination, but one ride is more visceral!
- ‘Colour control’: a characterization of the quality of colour rendering. For example, when evaluating the colour control quality, in the second validation set there was an image with a green and yellow bin next to each other. The best entries rendered both colours well, but natural. The not so good entries either lacked colour here or had pushed the colours too far.
- ‘Flat’, ‘flat rendering’: it is meant describe what appears as the lack of contrast between highlights and shadow values.
Rank | Team | Member(s) | Affiliation | Photographer's comment |
---|---|---|---|---|
1 | MiAlgo | Shuai Liu | Xiaomi | Really excellent results. Great contrast, good balance of light and dark. Good colour control (kids playground a great test of this) and has the ‘punch’! |
2 | DH_ImageAlgo | Chunxia Lei, Tingniao Wang, Song Wang | Zhejiang Dahua Technology | Also very good in terms of exposure control and contrast. Just lacks a bit of colour saturation compared with the top ranked entry. |
3 | IVLTeam | Simone Zini, Claudio Rota, Marco Buzzelli, Simone Bianco, Raimondo Schettini | University of Milano - Bicocca | Also good results, but like number two it lacks a bit in colour saturation and is slightly too dark. |
4 | The Majestic Mavericks | Hyerin Chung, Nakyung Lee, Sung Keun Kwak | CJ OliveNetworks | Nice amount of ‘punch’ but maybe pushed too far in terms of colour saturation and too much contrast. Overall a bit too dark and looks processed. |
5 | BSSC | Lu Xu, Chao Zhang, Yasi Wang, Chaoqun Zhuang | SRCB | Decent contrast but overall too dark. |
6 | NTU607 | Wei-Ting Chen, Hao-Hsiang Yang, Zhi-Kai Huang, Hua-En Chang, Chia-Hsuan Hsieh, I-Hsiang Chen, Yi-Chung Chen, Yuan-Chun Chiang | Stanford University, Google, National Taiwan University, ServiceNow | Much too dark, but colour control is decent. |
7 | DH-AISP | Jin Guo, Yangfan Sun, Tianli Liu, Dejun Hao | Zhejiang Dahua Technology | Everything is too light and ‘washed out’, resulting in drab colours and lack of ‘punch’. |
8 | Photoshop image enhancement | Maria Efimova | Organizers Team | All images are far too dark. |
9 | OzUVGL | Furkan Kınlı, Barış Özcan, Furkan Kıraç | Özyeğin University | Also far too dark and overall lack of colour. |
10 | JMUCVLAB | Marcos Conde, Tim Seizinger, Florin Vasluianu, Omar Elezabi | University of Wuerzburg, CVLAB | Much too dark and colours don’t look real. |