MIUI News Team

MIUIer

Rank: 6Rank: 6

Credits
20134
Device
Redmi Note 5/Pro
Online time
532 Hours
Send PM

Unlock MedalMi WiFi RouterNinebot MiniNinety Degree LuggageSuper MondayMIUI Lucky FridayHappy Valentine's Day1st AnniversaryMid-Autumn FestivalMIUI World TourMIUI Happy MedalMerry ChristmasMIUI Marathon MedalFan Of MIUI IndiaPuzzle MasterMIUI SHappy 2019MIUI Tech Survey13 Million Forum MembersOreoMIUI FollowerDie-hard MIUI FanPOCO F1Mi DropHeart to HeartScroll Of WisdomHumor MasterIndia Independence DayMi TalentXiaomi 8th Anniversary10 Million Forum MembersMIUI 8th AnniversaryMIUI 10Discuss HereMaster Of TechnologyMIUI Weekly Devices Quiz6 Million Forum Members

[Xiaomi] Xiaomi’s AI restores details and enhances colors in poorly exposed photos

[ Promote]
470 5 |
18:50, Dec-05-2018 | From PC
|
|
Go to reply no.
This post was edited by Yaswanth_T at 16:19, Dec-05-2018

image.png

Hello MIUIers,

Researchers at Xiaomi described a solution to the exposure dilemma in a new paper (“DeepExposure: Learning to Expose Photos with Asynchronously Reinforced Adversarial Learning“) accepted at NeurIPS 2018 in Montreal. In it, they describe an AI system capable of segmenting an image into multiple “sub-images,” each associated with a local exposure, that it subsequently uses to retouch the original input photo.

“The accurate exposure is the key of capturing high-quality photos in computational photography, especially for mobile phones that are limited by sizes of camera modules,” the researchers wrote. “Inspired by luminosity masks usually applied by professional photographers, in this paper, we develop a novel algorithm for learning local exposures with deep reinforcement adversarial learning.”
xiaomi_Ai.png

The AI pipeline — dubbed DeepExposure — kicks things off with image segmentation. Next comes an “action-generating” stage during which the input low-resolution, sub-images, and direct fusion of the images are concatenated and processed by a policy network that computes each’s local and global exposures. After the images pass through local and global value filters, the model completes a finishing step in which a value function evaluates the overall quality. Finally, the sub-images are blended together with the input image.

The neural networks at play here are of the generative adversarial network (GAN) variety. Broadly speaking, GANs are two-part neural networks consisting of generators that produce samples and discriminators that attempt to distinguish between the generated samples and real-world samples. To train the discriminator, the researchers randomly chose small batches of machine-retouched and expert-retouched photos; the contrast, illumination, and saturation features are extracted and concatenated with the RGB image to form an input.

They “taught” the AI system in Google’s TensorFlow framework on a Nvidia P40 Tesla GPU, and their corpus of choice was MIT-Adobe FiveK, a dataset containing 5,000 RAW photos — i.e., files with minimally processed data from the image sensor — and corresponding retouched ones edited by five experts for each photo. Specifically, they used 2,000 unretouched images, 2,000 retouched images, and 1,000 RAW images for testing.

DeepExposure outperformed state-of-the-art algorithms in key metrics, managing to consistently restore most details and styles in original images while enhancing brightness and colors.

“[Our] method bridges deep-learning methods and traditional methods of filtering: Deep-learning methods serve to learn parameters of filters, which makes more precise filtering of traditional methods,” the team wrote. “And traditional methods reduce the training time of deep-learning methods because filtering pixels is much faster than generating pixels with neural networks.”

Source

Favorite Share

Rated by 1 person   Experience Prestige Reason  

VikuBalupura + 10 + 1 Fantastic!

Experience +10  Prestige +1  View Rating Log

MIUI News Team

Rank: 6Rank: 6

Credits
12386
Device
Redmi Note 5/Pro
Online time
341 Hours
Send PM

Ninety Degree LuggageAwesome!MIUI Happy MedalMerry ChristmasSuper MondayAlpaca Mi Bunny SurpriseMIUI Halloween 2018MIUI Weekly Devices Quiz13 Million Forum MembersMIUI Tech SurveyHappy 2019MIUI Medals CollectorOreoMIUI FollowerDie-hard MIUI Fan6 Million Forum MembersMIUI 10Humor MasterMIUI Halloween 2017Mi DropHeart to HeartPOCO F1

21:43, Dec-05-2018 | From Redmi Note 5
|
Thanks for sharing

Diamond Member

Rank: 5Rank: 5

Credits
1169
Device
0
Online time
0 Hours
Send PM

App Star

21:56, Dec-05-2018 | From Redmi 4A
|
Thank's

Platinum Member

Rank: 4

Credits
2285
Device
POCO F1
Online time
16 Hours
Send PM

Merry ChristmasMIUI Happy Medal13 Million Forum MembersPOCO F1MIUI FollowerMIUI Tech SurveyMIUI Weekly Devices Quiz

12:19, Dec-16-2018 | From PC
|
Awesome Post ! thanks for sharing

Elite Member

Rank: 6Rank: 6

Credits
9292
Device
Redmi 5 Plus
Online time
198 Hours
Send PM

Merry ChristmasMid-Autumn FestivalMIUI Fan From IndonesiaMIUI Lucky FridaySuper MondayMIUI Marathon MedalMIUI Happy MedalSummer Mi BunnyAlpaca Mi Bunny Redmi 5/Plus6 Million Forum MembersPOCO F1OreoMIUI FollowerHeart to HeartMIUI 10MIUI Weekend GiveawayMaster Of TechnologyMIUI 8th AnniversaryMIUI Weekly Devices QuizMi DropDiscuss HereSurprise10 Million Forum Members

09:53, Dec-18-2018 | From PC
|
Honestly, Xiaomi's camera is still behind Samsung, Google and Huawei but give them sometimes to pick up!

Supreme Member

Rank: 7Rank: 7Rank: 7

Credits
48638
Device
Redmi 4X
Online time
31 Hours
Send PM

Mi  BunnyMi MIXMi BoxMi TV 70Unlock MedalMagic CubeMi Band 2Mi HomeMi ScaleMi HeadphonesMijia Walkie TalkieMijia Smart PTZ CameraMijia 360° Panoramic CameraYi Action CameraMi T-ShirtMi WiFi RouterNinebot MiniMi Power BankMijia Smart ShoesMi Notebook AirNinety Degree LuggageMi Bluetooth HeadsetMi BandMi Kids WatchMIUI Devices ContestFantastic!AnTuTuAwesome!I Love MIUIFlashholicMarshmallowRedmi 4XSuper SundayMIUI SMIUI 100 Million UsersSuper MondayHappy Valentine's DayMIUI 2017MIUI Fan from TurkeyMIUI P300K Forum MembersSuper SaturdayWomen's Day MedalMIUI Weekly PollHappy Singles' DayFan Of MIUI India2 Million Forum MembersXiaomi 6th Anniversary1st AnniversaryHappy New YearMIUI Fan From Indonesia10 Million Forum Posts5 Million Forum App InstallsDiwali Merry ChristmasRocket to FutureMIUI 9Happy DiwaliMIUI 7th AnniversaryMIUI World TourMIUI Lucky FridayMid-Autumn FestivalFan Of MIUI GreeceMIUI 50 Million UsersPuzzle MasterMIUI Forum AppRooster Year of 2017MIUI Happy MedalMIUI Fan From PhilippinesMiPopGamer Mi BunnyAlpaca Mi Bunny Summer Mi BunnyMi DropMIUI 8th AnniversaryDie-hard MIUI Fan2017 Annual AwardsMIUI Tech Survey9 Million Forum MembersIndia Independence DayMIUI 10Android ElfDiscuss HereMIUI KingMi Talent6 Million Forum MembersHumor MasterMIUI Halloween 201810 Million Forum MembersHeart to Heart7 Million Forum MembersHappy 2019Star WarsMIUI FollowerStar Of Technology25 Million StrongBlack SharkAndroid GuruMi Lover MedalScroll Of WisdomTheme TalentRingtone ExpertDebate MasterMaster Of TechnologyMIUI Halloween 2017Mr. DependableMIUI Medals CollectorTechnology GuruMost Active MIUIerXiaomi 8th AnniversaryQuick SurveyRedmi Quiz Master5 Million Forum MembersMIUI SuperstarMIUI FB FollowerSurpriseOreoMIUI Weekly Devices QuizMIUI Quick Discussion8 Million Forum MembersNougatMi Story TellerForum Contest Lover

23:41, Dec-18-2018 | From Redmi 4X
|
a step forward into improving xiaomi's camera images, especially redmis.. 🙂

Credits Management

Quick Reply Top Return to the list