iPhone photos - how does it group 'related' photos/videos together Watch

21ForEva
Badges: 17
Rep:
?
#1
Report Thread starter 2 weeks ago
#1
Hey there

So I had been to 2 concerts this week. And was checking out my videos from last night. Then I scrolled down and it showed me 4 thumbnails each related to a different time/date since I last synced my phone. It came under 'related' and basically all 4 of these groups of images/videos are of concerts.

How the heck did it know to group these photos together?

Is it because they all share a similar burst of lighting in the videos? Or there's loads of figures in the shot? Or there's a lot of movement in the clips?

I'm just really curious cuz it's quite cool.
0
reply
Maple_Leaf
Badges: 10
Rep:
?
#2
Report 1 week ago
#2
Yeah I was wondering about this as well! I think it must somehow know what kind of shapes/colours are in each picture and if it's people/scenery/an event and group them together based on this- quite scary but clever!
1
reply
0le
Badges: 21
Rep:
?
#3
Report 1 week ago
#3
Using Machine Learning and possibly meta data embedded within the photo.
1
reply
21ForEva
Badges: 17
Rep:
?
#4
Report Thread starter 1 week ago
#4
Gosh I can't wait to get to grips with programming so I can move into Machine Learning !!!!! It's actually so exciting.
0
reply
0le
Badges: 21
Rep:
?
#5
Report 1 week ago
#5
(Original post by 21ForEva)
Gosh I can't wait to get to grips with programming so I can move into Machine Learning !!!!! It's actually so exciting.
Porgramming is a lot of fun. The only advice I can offer is build your codes in pieces. Start off with basic tasks/features and gradually improve the code by adding more and more features. For example in image analysis, fist task is to scan the image, second task is to do some pre-processing (color to greyscale conversion, noise removal etc), third task is to process the image etc.
0
reply
21ForEva
Badges: 17
Rep:
?
#6
Report Thread starter 1 week ago
#6
Thanks a lot for the advice dude -- it sounds like a hella exciting side project. I'll research any Github projects to see if someone has started off with something similar to image analysis.
0
reply
winterscoming
Badges: 19
Rep:
?
#7
Report 1 week ago
#7
The apps on your phone almost certainly use something like Google Vision to analyse the image - you can read about how it works here: https://cloud.google.com/vision/
Or it might use Amazon Rekognition, which does a very similar thing: https://aws.amazon.com/rekognition/
(Or since it's an iPhone, then Apple Vision - https://developer.apple.com/documentation/vision )

Although the main reason these work so well is down to the unimaginably vast quantities of images and data which companies like Google and Amazon have got stored in their databases - Apple/Google/etc. have hundreds of millions of images, each with tonnes of tags/categories/metadata about what they depict and contain, so they've been able to train all of their statistical models to a really high degree of accuracy.

Luckily, you wouldn't need to do any of that in order to write an app which can do all of this because you can just use the Google/Apple Vision or Amazon Rekognition APIs yourself, then build an app around those. These kinds of AI/Machine Learning services are actually a lot more accessible than you might think; not just for images, but also for audio/speech, etc. For example, Amazon Echo has got an API which lets you build your own Alexa skills without too much effort.
Last edited by winterscoming; 1 week ago
1
reply
21ForEva
Badges: 17
Rep:
?
#8
Report Thread starter 1 week ago
#8
(Original post by winterscoming)
The apps on your phone almost certainly use something like Google Vision to analyse the image - you can read about how it works here: https://cloud.google.com/vision/
Or it might use Amazon Rekognition, which does a very similar thing: https://aws.amazon.com/rekognition/
(Or since it's an iPhone, then Apple Vision - https://developer.apple.com/documentation/vision )

Although the main reason these work so well is down to the unimaginably vast quantities of images and data which companies like Google and Amazon have got stored in their databases - Apple/Google/etc. have hundreds of millions of images, each with tonnes of tags/categories/metadata about what they depict and contain, so they've been able to train all of their statistical models to a really high degree of accuracy.

Luckily, you wouldn't need to do any of that in order to write an app which can do all of this because you can just use the Google/Apple Vision or Amazon Rekognition APIs yourself, then build an app around those. These kinds of AI/Machine Learning services are actually a lot more accessible than you might think; not just for images, but also for audio/speech, etc. For example, Amazon Echo has got an API which lets you build your own Alexa skills without too much effort.
Mind blown. Completely mind blown. I’m sick right now with the flu, bored out of my head so I will definitely read those articles. Thanks a lot for the info!
0
reply
winterscoming
Badges: 19
Rep:
?
#9
Report 1 week ago
#9
(Original post by 21ForEva)
Mind blown. Completely mind blown. I’m sick right now with the flu, bored out of my head so I will definitely read those articles. Thanks a lot for the info!
You're welcome! I had a look through some of the tutorials and example code they've got on Google Vision API, it all looks fairly beginner-friendly - if you get get a chance to try out some of their tutorials and example code then there's loads of interesting stuff to tinker with. Something like this is great for a spare-time coding project
0
reply
21ForEva
Badges: 17
Rep:
?
#10
Report Thread starter 1 week ago
#10
(Original post by winterscoming)
You're welcome! I had a look through some of the tutorials and example code they've got on Google Vision API, it all looks fairly beginner-friendly - if you get get a chance to try out some of their tutorials and example code then there's loads of interesting stuff to tinker with. Something like this is great for a spare-time coding project
Thanks - I will definitely check it out. Sounds way more interesting than the coding projects I have to do in my team at work haha.
0
reply
21ForEva
Badges: 17
Rep:
?
#11
Report Thread starter 1 week ago
#11
https://cloud.google.com/vision/ -- omg so I dragged images of myself and also fav KPOP band posters and wowww I am seriously blown away by how much the API can predict features and properties.... like it can tell if my face is full of joy or sadness. It's maddddddd
0
reply
winterscoming
Badges: 19
Rep:
?
#12
Report 1 week ago
#12
(Original post by 21ForEva)
Thanks - I will definitely check it out. Sounds way more interesting than the coding projects I have to do in my team at work haha.
Haha, I'm sure it is! What kinds of projects are you doing? it's far more interesting than a lot of the database webapp projects my team are doing at work too


(Original post by 21ForEva)
https://cloud.google.com/vision/ -- omg so I dragged images of myself and also fav KPOP band posters and wowww I am seriously blown away by how much the API can predict features and properties.... like it can tell if my face is full of joy or sadness. It's maddddddd
Yeah, it's really cool, especially that you can do that sort of thing with video too. Also a bit scary when you think about what kinds of things companies or police are probably already doing with this stuff - e.g. connected to CCTV feeds, also using facial recognition, etc.
0
reply
X

Quick Reply

Attached files
Write a reply...
Reply
new posts
Back
to top
Latest
My Feed

See more of what you like on
The Student Room

You can personalise what you see on TSR. Tell us a little about yourself to get started.

Personalise

Did you vote in the 2019 general election?

Yes (413)
43.84%
No (98)
10.4%
I'm not old enough (431)
45.75%

Watched Threads

View All
Latest
My Feed