DDraceNetwork - developer

DDraceNetwork

Development / developer

Development discussion. Logged to https://ddnet.tw/irclogs/ Connected with DDNet's IRC channel, Matrix room and GitHub repositories — IRC: #ddnet on Quakenet | Matrix: #ddnet-developer:matrix.org GitHub: https://github.com/ddnet

Between 2020-01-22 00:00:00Z and 2020-01-23 00:00:00Z

[ddnet-maps:master] 1 new commit

93af14d M Rest to Die - ddnet-maps

Someone published a DDNet paper: https://arxiv.org/abs/1911.11582

DDNet: Dual-path Decoder Network for Occlusion Relationship Reasoning

Occlusion relationship reasoning based on convolution neural networks consists of two subtasks: occlusion boundary extraction and occlusion orientation inference. Due to the essential differences...

07:09

Seems like a popular name recently: https://www.researchgate.net/publication/332522533_DDNet_Cartesian-polar_Dual-domain_Network_for_the_Joint_Optic_Disc_and_Cup_Segmentation

(PDF) DDNet: Cartesian-polar Dual-domain Network for the Joint Opt...

PDF | Existing joint optic disc and cup segmentation approaches are developed either in Cartesian or polar coordinate system. However, due to the subtle... | Find, read and cite all the research you need on ResearchGate

lmaoo hopefully the ddnet website gets some of the juice

lol nice

12:54

where did u find that @deen ?

12:55

ah wait

12:55

its not ddracenetwork xxD

12:55

https://arxiv.org/pdf/1911.11582.pdf

12:56

:/

[ddnet/ddnet] Pull request closed: #2042 Check snap

13:14

[ddnet:master] 5 new commits

ef3bce2 added some checks to snap handling - oy 49168da fixed typo - oy ee825f4 Add snap check for dummy - ChillerDragon 610976e Fix max snapshot size - ChillerDragon a47a94e Merge #2042 - bors[bot]

@ChillerDragon just googled for ddnet

13:55

also found the discussion of elxreno trying to get us into Fedora/RedHat: https://bugzilla.redhat.com/show_bug.cgi?id=1787069

oh nice

Image attachment

15:26

wrong link. it leads to the previous tournament discussion

Image attachment

[ddnet/ddnet-web] Pull request opened: #82 Fix forum link on tournament 49

15:40

[ddnet/ddnet-web] Pull request opened: #83 Ignore tags

[ddnet/ddnet-web] Pull request closed: #82 Fix forum link on tournament 49

16:32

[ddnet-web:master] 2 new commits

9f8d354 Fix forum link on tournament 49 - ChillerDragon baa4c56 Merge pull request #82 from ChillerDragon/pr_fix_tourn49_link - def-

when im in a situation like this where i fall a good amout before entering a single unfreeze, it doesnt register my shots most of the time and some times not even my double jump. this cost me rank1 on another map once, and its a problem with the connection

Image attachment

can we get some crowdsource to fix corneums connection pls

Not sure I get it, but as you go faster the resolution of the input gets smaller and smaller

18:02

if you are moving any more then 1 tile per tick, it'll get very hard to accurately do inputs

well on that other map i was always falling from the same height, and it took n9mkoik's shot input but not mine so i had to do a extra jump

18:03

its not the same for everybody

I don't quite get the part, so I don't quite get what happened. Were you moving fast?

wait i will search it

"doesn't register" isn't something that can really happen especially for shots, you'd need to drop so many packets

18:05

factoring in ping though, you'll never be able to get very consistent on a server anyway, it'd be nice if we had a way to verify teehistorian demos so the people speedrunning could do so on local servers where ping is not an issue

heres the part. im really not moving fast there

Image attachment

18:07

yeah. it worked on my lan server so i was pissed when it didnt work for real

18:08

actually... this part is different to what i said. but its the same problem for sure

18:09

btw i will legit pay you 50€ if you make a way to verify demos

so you need to rocket right before you go into freeze?

yes

this is one of those things that will always be a little inconsistent, depending on your falling speed and your ping, it actually can be impossible, I think

18:11

After a certain falling speed, it's definitely impossible as the rocket will not travel enough in the tick before you get frozen to hit the ground before you

what about double jumps tho. it just failed me too. unlucky ping spike?

verfiying demos would actually a cool thing

18:12

or submitting teehistorian files idk

@ChillerDragon 50€ to whoever does it

below that speed you need to be lucky to be close to the edge of the non freeze block in the last thing before you get frozen so the rocket doesn't have to travel too far

@Learath2 is there anything untrusted about local servers that would be a risk to accept the runs as official ddnet ranks?

@Im 'corneum double jumps suffer from the same problem at high speeds, but less so at below critical speed, I'd guess that was an unlucky spike indeed

hm ok

@ChillerDragon if we get teehistorian files we can run it through a server of our own to check that they didn't modify the server

but then you could do easy record bot ranks because they dont fail on homeserver

Yeah, there is that, we'd need to start working on proper bot detection aswell

18:14

some analysis of the movement of the tee to make sure it's human-ish

replay bot is only human input

noby would be the guy but i dont think he cares much about ddrace

@Im 'corneum thats the thing i can record any run locally and then replay it on official servers anyways

18:15

not that i recommend you doing that xd

18:15

but since thats possible already we could also make a legit similar solution

we could cross compare with other runs aswell

well ur internet might be too bad for that but mine works supberb with replay bots ❤️

no chuller. if you just replay online they will fail at some point due to laggs. so lags are literaly making the servers half secure now

18:16

chuller lol

i tested it

oh

and there are certain things that give away the usage of a replaybot, things that I don't feel like giving away here

yea ik

assuming you dont mean going into corners before every part?

"we shouldn't mention bot detection signs"

I guess there goes that one

Im 'corneum goes ahead and tries to guess bot detection signs

@Learath2 so local servers can record teehistorian without too much huzzle right? Even on windows. So we only need a way to replay teehistorian on official servers and then insert the ranks?

just making sure. see? it wasnt a secret xd

wat

18:17

"something better stay secret"

18:17

"is it <xyz>?"

xd

sounds like such a bad strategy

18:18

I'm not even commenting on whether or not that is what learath2 means

18:18

but the idea

18:18

to respond to something like that

18:18

in such a way

18:18

is thoughtless

18:18

i dont get what i did wrong tbh.

sounds like a fun machine learning project

7:17 PM] heinrich5991: "something better stay secret" [7:17 PM] heinrich5991: "is it <xyz>?"

18:19

you did that wrong

that was the first and most obvious way to detect record bot

nice. and then someone posts the second most obvious way to detect a record bot

that would be dumb if a ddnet official posted it.

18:20

but if i post it you know that you need better protection

security through obscurity is atleast a little helpful

I'm not saying this is the secret I want to keep hidden

18:21

I don't even know what secret Learath2 wants to keep hidden

18:21

but why do you think it is a good idea to guess the secret

18:21

that's just helping bot writers

18:21

and I'm pretty sure you don't like bots

well i sure hope he has something solid that i dont know about. thats why i said i assumed he didnt mean said method because only having that would be insecure

then tell your guess in private

Well any clever guess you have is a nice idea for bot writers, is the point @heinrich5991 is trying to make I guess

good point. but i am pretty sure it is/was very known stuff. especially bot writers would know about this thing

Yeah, I guess this one wasn't too bad, just don't guess more

i will guess. but then i pm

18:24

but how do you know im not telling a botwriter

18:25

you would have to force me to stop thinking about ot xd (edited)

well if you are helping write bots, then your guesses are all known to the people who do write bots, so putting them here doesn't change much

18:26

but if you aren't and you don't want them to get ideas, don't post guesses publicly

good for you that botwriters are even lazier that you guys

tbh i dont even have contact to one

[ddnet/ddnet-web] Pull request closed: #83 Ignore tags

21:53

[ddnet-web:master] 2 new commits

4cac0b1 Ignore tags - ChillerDragon de177d6 Merge pull request #83 from ChillerDragon/pr_ignore_tags - def-

[12:15 PM] Im 'corneum: noby would be the guy but i dont think he cares much about ddrace yeah i dont care about ddr really but i would be interested to work with any big server on anti cheat stuff

23:39

i have asked before about a way to replay teehist files thru a server that can run different types of anti bot checks im not sure what happened to this

Exported 110 message(s)