Cut-Out Player Faces Megapack
Our Cut-Out Faces Megapack is the biggest collection of Football Manager Player Faces available. All players are available in the default cut-out style. When you download our Cut-Out Faces Megapack, you'll receive over 480,000 player faces to spice up your copy of Football Manager.
- 482,917
- 2025.01 - Released on 04 Dec 2024
HRiddick
Download Test Pack 1 Here
I think we can all agree that the 3D faces in Football Manager aren't what they could be, and for many people take away from getting truly immersed into the game. And while unfortunately there has been no effective way of replacing newgen/regen faces since the system was changed to from 2D to 3D generated faces in FM18, as a community it would be amazing to come up with an effective alternative if possible!
The end goal here is the following:
To provide a sister pack to the Cut-Out Faces Megapack that automatically replaces the in game 3D generated newgen/regen faces with a collection of high quality 250x250 pixel cut out images.
Above is a link to a small pack of about 850 cut out images, sorted into folders for the following different ethnicities: African/Caribbean, Asian (Central), East Asian (China/Japan), Mediterranean/Hispanic, Mixed Race (Black/White), North African/Middle Eastern, Northern European, Pacific Islander and South East Asian. These make up the bulk of the ethnicity categories from the official game database.
These images are mostly taken from US college sports portraits, futsal, handball, esports and other lesser known sports, so no chance of finding LeBron James or anyone well known from NBA/NFL appearing as your in game wonderkid! At the moment I still have about 6100 images stored and still to be cut and still have a lot more sources images to sort through and save from US college sports websites, so this is where the most help is required - more on that below.
For the time being, while there is no way of automatically replacing the in game 3D faces, this megapack could be used as either:
- A database from which you can choose individual images and assign manually to chosen newgens, or
- A system like that mentioned in this post where newgen faces are replaced automatically but with no discerning by ethnicity.
Here's how you can help!- Cutters - If you want to help produce cut outs to add to this future megapack, great! Let me know here or by PM and I can send over however many images you would like to cut.
- Image Finders - The more the better really. As long as the image is a high quality portrait, not of a professional footballer that already exists in game (so essentially from other sports) and is not recognisable (so preferably not from top level American Football, Basketball, Rugby etc) then it would be perfect, so anyone willing to collect source images for others to cut or knows where to find good quality source images would be of immense help. Source images for Asian and Pacific Islander would be particularly useful.
- Image Croppers - remove.bg is a fantastic tool and is perfect to get as many images cut as quickly as possible. If you are not a cutter yourself but still want to help, you can do so by cropping images already collected to include just the head and shoulders and put them through remove.bg - this drastically reduces the work required from a cutter who as this point just needs to re-scale and tidy the image!
- Football Manager/coding experts? - If anyone has ideas as to a way of coding/assigning images from a folder to match a certain ethnicity in game then that would be a massive step forward - as far as I'm aware there are no current solutions to this and I wouldn't know where to start myself, so any expertise or ideas on this would be hugely appreciated.
If you want to help or have any questions, leave a comment here or send me a PM!mons
Interesting idea, not my cup of tea but best of luck with it
I don't think there's a way of doing what you're asking above other than manually by the end user, I'm afraid. To aid them, I would strongly suggest doing a brief video outlining the steps which need to be taken, to be sure that people can actually see what needs to be done
krissmed
As for image finding, thispersondoesnotexist.com is a great place for finding decent looking sources. It's basically an AI that makes a randomally generated portrait, so there is no chance of it being recognicable. Here is a google drive folder with 100k generated images,
columbuscrewdude
HRiddick
That's encouraging, hopefully there would be a way of incorporating a player's ethnicity or nationality to the config if that data can be exported. As for the AI generated portraits that's a fantastic idea. Could be a great way of working around sources if it could be trained specifically with footballer's/athlete's portraits, at the moment from that website it's a bit of a lucky dip with age/gender/too close a crop unfortunately.
The main problem I'd have with this would be the poor image quality from upscaling. Hopefully we can gather enough sources to not see repeats in game but without the need to upscale images!
krissmed
You should pitch the idea to the guys over at fmscout as they are responsable for genie scout. Maybe they can help you out with the coding. I have almost no knowledge on this stuff but I figure the hardest part is to extract the data. Once you've managed that you just need to generate a config based on age, role, nationality, etc.
As for the source collecting, thispersondoesntexist isn't perfect, but you won't run into problems where faces are recogniceable and you can get as many source as you'd like. The racial and age diveristy is not great though, it's mostly white or asian people between 30 and 40. Although, if you want all people ingame to have a real face, you'll also need staff pictures, which it can provide.
Alternatively you could use some like this to edit current cuts.
Edit:
You can train this AI on your own images. Although I'm very unfamiliar with both phyton and tensorflow so it's best to leave it someone else. Just don't delete the sources you already cut so that we can use those to train the AI.
HRiddick
Would love to also include real images for staff later down the line but think that depends on the interest/success for just the players first of all. Would also have no idea how to train the AI myself so hopefully someone would be able to lend a hand there. Would be great to get it up and running though, could also include the abundance of player portraits out there, recognisable or not, as the end result would eventually be indistinguishable I'd imagine! Sites like transfermarkt would be a good place to start with images for the AI training.
krissmed
Have you posted this over at sports interactive's forum? Maybe you could get some coding/training help from someone over there.
HRiddick
Good idea, I've mirrored the top post over to a thread there
krissmed
Edit:
A exported html file looks for like this:
The code used to generate the site is fairly simple so you can easily seperate the players from each other.
krissmed
@Footygamer Is this possible at all? If so how difficult would it be to make it?
Footygamer
Then each individual who wants to use the pack would upload their personal csv of regen ids/age/ethnicities and then a config.xml would be generated.
An alternative might be a program that works on your computer and you just download the images once then the program generates a config.xml based on the regens in your game and the images you've downloaded (the images would come with a config stating age/ethnicity etc).
One of the problems with this is that the regens will be different in every single game, so if you have two or three saved games going you'd have to swap out the configs all the time.
krissmed
The cuts can be hosted at an external site to limit the bandwidth usage and the site would just be used to generate a config.
Would training an AI based on the cutout requests and other sites like transfermakt be possible with a mid-range desktop?
Almost the same as fmxml then, right? The only difference being that it works for regens rather than pre-existing players? Does this take ethnicity, age, etc. into account?
Footygamer
It would need to get data from the game so slightly different from fmxml.
And training an AI you'd use something like AWS (Amazon Web Services) rather than your personal desktop. I'm not sure on the costs involved.
krissmed
Pretty cheap it seems.
Footygamer
Haha machine learning/AI has the potential to escalate very quickly, i'd be surprised if you could keep it to $1
HRiddick
Of course the main problems with this would both what you previously mentioned @Footygamer that this would only work for one save file, and that you would need to repeat this process for every youth intake year in game when more newgens are generated, but this would all still be preferable to any existing alternative!
Are you able to specifically export the assigned ethnicity of a player using Genie Scout, rather than their nationality? I'm on mac so can't use the software myself unfortunately - I assume not as the ethnicity parameter holds no relevance for what Genie Scout does, but hopefully from the same way they extract the save game data the ethnicity parameter can also be extracted.
Footygamer
I'll contact him.
krissmed
I totally agree with you. A simple fmxml type program would be the best option. And no genie socut doesn't have a column for ethnicity. But you can filter by region, e.g south european, south asian, east african, etc. So you should probably seperate the folders in those same regions.
As for training, I found this software. @Footygamer Are you fimiliar with this at all? It seems really easy to and does exactly what we want. It does require a subscription though.
Footygamer
I wonder if it would work doing a face in the game type thing. I figure we could get a few thousand.
Just also looked at age progression stuff, i.e. if we could age all 20 year olds 10 years and deage all 30 year old 10 years it'd be like getting 100,000s of new faces
krissmed
Yes, training an AI for each region. It says that we'll need 500-5000 images for the dataset. That age progression stuff is also available in that program. In theory we could also age the regens as they age ingame.
laurentius
dpendao
https://sortitoutsi.net/uploads/mirrored_images/ub9ajs977VNpZJbxQTupfxb4UBhpziCbnvD4BeXB.png
krissmed
Edit: Forgot to link to the actual thread.
Here you go: https://fm-base.co.uk/resources/zealands-real-faces-on-regens.205/
Edit 2: "Hey guys,
The reality about the ethnicity identifier, I can't for the life of me figure it out.
I looked into using conditions like if skin colour = (skin colour ID) then display specific faces. No luck whatsoever.
If anyone has any ideas, let me know!"
They've given up (for now) it looks like. Still worth a shot
HRiddick
Yeah I messaged the guy that uploaded that pack a while ago and I pretty much got the same message, think they're in the same situation of not knowing how to get past that hurdle of discerning by ethnicity. Either way good to know that other people are keen on the idea and have thought of working on it!
the_hdk
laurentius
HRiddick
Would you be able to PM me further details on this?
krissmed
@Footygamer I found these frameworks which can read FM savefiles. Could be helpfull if you want to make a FMxml type program.
https://code.google.com/archive/p/fmsx/downloads
https://community.sigames.com/topic/191264-fm11sx/
https://www.fmscout.com/a-fmscout-framework.html
https://github.com/ThanosSiopoudis/FMScoutFramework
Edit: Just to clarify, these are the regions I'm talking about
HRiddick
Neither region nor skin tone would be a good way to sort the images. To use examples of real players, both Antoine Griezmann and Paul Pogba have the same region of birth assigned, whereas in game Griezmann has the same skin tone assigned as Shinji Kagawa, David Silva and Henrikh Mkhitaryan; all these players are from different ethnic backgrounds and all look very different. The ethnicity variable isn't great as you said (another example is how the East Asian category bundles in China, Japan and Korea together), but unfortunately it's still by far the best variable the game provides to categorise everything.
I know that you can't view a player's assigned ethnicity in game, but every player in the database as far as I know has one assigned that you can view in the official Editor. I've always assumed this would also be the case for any newgen/regen generated, but I've never known for sure as you can't check this in game. If it isn't the case that regens have an assigned ethnicity then some sort of system that assigns images based on both region *and* skin tone would probably be necessary, although I imagine that would be a nightmare to get right.