Discussion in 'General Discussion' started by Lgel, Nov 5, 2017.
I didn't express well. For layout I mean top view (X,Y). For Z coordinate I agree Google is trash.
And I meant X,Y Is often very bad because of the elevations. It throws off the X,Y and often by large amounts.
Can you please share an example where we can compare the obtained xy layout?
I would do it myself but I dont know any laser scanned track for rF2 which I could use.
In my case the only real track I have modeled is OcioKart which is available as a WIP project for rF2. I stopped working at it until I could find a way to get proper elevation. My track is completely flat because I cannot find a proper way to do this. However, I have never noticed differences in XY between the real track and the virtual one.
For all my other fictional tracks I certainly don't have this problem.
This is an extreme example but it shows what is going on. Depending on how the photo was taken it can have a pretty severe angle to it. Remember close up GE images are not taken by satellites. They are taken by aircraft. This building is 230' tall and it measures 276' between the top of the building and the parking lot on the X axis. If this was a track with as little 20' of elevation change you can quickly see how the track can be "shifted" quite a bit.
Also note the road on the left. Again this is an extreme example. This image is taken with 3d terrain turned off so it is perfectly "flat"
For me showing this image is trying to prove nothing and confuse people. Google aerial textures are projected into the 3d model so that ground objects are correctly spotted. The fact that the views show perspective is an undesirable consequence of using aerial pictures for better resolution. The error is not so big as you try to prove with the tip of the building argument.
Perspective is the same whether you are talking about a 100' building or a 100' hill.
But just for the sake of argument here is the differences between lidar and a google image of NJMP.
Here is turn 7 over the lidar intensity data.
Same corner over GE image that has been properly scaled
Exit turn 9 over lidar
Turn 3-4 over lidar
While these indeed look small you have to remember there is overall fairly little elevation change at NJMP Lightning. At a track with much more elevation change it can be much worse. The only way to get these images to line up would be to scale and manipulate the GE image in various ways to make it fit. And without any other data to reference you wouldn't know which way to scale it. I didn't bother with this image as I only needed it for basic track object placement.
EDIT: A direct comparison of the track in motion.
Anyway I apologize for going more off topic.
What does Toban tell us?
Love me some Toban ..;-)
I remember way back racin rf1 on the BRO's server ! they did megains & clios @ toban ,toban reverse, toban long ..all of them !
& the new updated version is some of ISI's best work IMO!
@kro388th what name did you race under? I was in there lots for a good 6 months...
He said twice it was an extreme example, and it's a valid point. As you said yourself, the Z data is trash, so even if the image is projected onto that Z data, it's not a perfect basis for doing a track and making it difficult to distinguish from laser scanned. Which, if you're going to do it without laser scanning, should be the objective - close enough that any easily accessible data/footage doesn't show a problem.
There are other kind of cartography that describe elevation changes, they are freely avaiable and with some work can be used to make the overall landscape heightmap, yes they can't pick very small elevation changes, but if the track lay on a slope you could get the elevation changes, then to refine it, probably cameracar videos could help too.
In your second example both Lidar and Google data show a very good correlation.
When driving both versions nobody could tell the difference.
Actually, it should be noted that the apparent resolution of Lidar data is not very high. Tracing the road in it seems to be a bit subjective due to the lack of a sharp definition of the road.
What I meant with a good example would be Spa. I have found that there is a laser scanned version of it. I will compare it with the not laser scanned one and with google Earth images. I will focus in the details of Eau Rouge where the elevation change is the biggest we can have in a real track. That will give us a fair comparison between them.
Sorry but I still disagree about the building comparison. As I said in previous post Laser Scanned Spa vs Google will give us a fair comparison. Practical and real data are beyond any error theory. If we find big differences you will be right but if we find the same layout using Google or the Laser scanned version then please accept that Google is accurate enough for providing XY data.
Bros did run with rf2 for awhile early on but faded away. Megane where always crowded in rf1 , Mills ,Toban Essington and orchid lake !!
The only one they ever will drive on is the Nordschleife and even then the sim doesn't factor in g-forces and things like that.
I love Silverstone. Going through maggots-becketts is fun in any car, especially when you're going for that all important pole lap and feeling those front tyres just start to slide away from you. As with any old airfield track, it's going to be flat.
Toban is a great track and people might not be aware it's fictional because it's so good. But you don't NEED fancy laser scanned tracks to have a good time.
I never said it is always bad but it can be bad. Mosport for example was pretty dead on accurate as well as Watkins Glen. But unless you have other data to compare it to you have no idea if it is right or not. If you are OK with that level of error then that is fine. I personally am not. And as far as the examples above If you don't think 5-10' difference is enough to drastically change a corner then I can't help you. Yes it is "close" but it isn't close enough for me.
Guys, we've had two guys who create high-quality tracks including one who's arguably the best track maker from the Assetto Corsa community, posting here with supporting argument regarding quality and accuracy, and there is still this attitude against facts?
Edit: How do you feel about the quality from GT Legends, GTR, and SimRaceWay compared to ISI tracks from rFactor 1? All of the former titles had their tracks converted and available for comparison within the rF platform.
1) You have fully understood what is at stake for RF2: make smart priorities and stick to it because S397 is a small team and can absolutely not assume the secondary priorities without impacting TOP priorities.
2) But you make a huge mistake !!
You deliberately contrast physics with an extremely detailed track .... these two elements are totally complementary.
Ok at the end of 2017, we do not have the same two on RF2 .... but it is indeed the future of simulation.
The essence of the physics is the tires in contact with the ground and suspensions reacting with the ground: quality of the tire model is crucial, the same for other technologies of physics ... and you'll understand (otherwise I can not do anything for you) that the details of the track were of crucial importance.
An opposition such as physics VS beautiful graphics or new UI or visual damage of cars would be more appropriate for example. In this case it is to do: realism as high as possible VS maximum immersion / ergonomics of the UI etc .... Personnaly I have my preference beetween max realism and maximum immersion............... This is just an example.
I do not know if some of you are lovers of sound / music but just to make a very fair metaphor because I know myself in this area:
- You can have the best speakers of the earth, the best DAC (converter digital to analog audio), better cables cut at the right distance, etc. :
If you listen to music mp3 format even in 320kbs, the sound in the end will be VERY BAD.
- With an intermediate sound system and with a CD sound or a studio sound, the sound will have a very much higher quality than that of ultra high-end sono with mp3 audio source (I know what I'm talking about). And why ? Because the source sound is of poor quality (NOT ENOUGH *** DATA***).
Very high quality sound system = the PHYSICS (with all technologies)
Audio track IN QUALITY CD OR STUDIO and very detailed = the SCANNED LASER TRACK.
It is a metaphor of course.
I'm talking rF1, circa 2007
Aerial shots obviously aren't taken from directly above every point on the ground, so perspective will skew the edges of each 'run'. Easy to compensate for by warping the image, of course, except that doesn't take elevation into account. Hence potential errors, and those errors of even 3m can completely change the nature of a corner.
As for comparisons, to paraphrase you I'm amazed that you would think a single section of one track proves or disproves anything.
Anyway, even if google doesn't automatically arrive at perfect results, any prominent track (and those tend to be the most driven-on, and most often compared, across sims) will have plenty of footage and pictures to allow for cross-checking and correcting of the initial google layout. For that reason I don't see laser scanning as a necessity if care is taken.
And I'm sorry I waded into this, I was trying to just stick to the topic.
In 2007-08 I did many a lap in the BROS megane server, around the ISI fictional tracks (plus Silverstone and nuerburgring). For a time I was the guy running the eurocup config against all the endurance cars and having guys asking me how I could go so fast in a slower car... and saying "is it slower?"... people are too quick to judge based purely on horsepower...
Its clear that a bad MP3 sounds bad over a good Stereo - and a good Vinyl over a bad Stereo sounds bad too...
I think there should be found a compromise - nothing else.
But, things I don't understand in the Sim-World-Communitiy: why is everyone talking so much about Sims while they could drive these things and progress?
What about a Challenge on Toban similar to something like this?
Anyway, i talked too much and wish everyone a nice evening
Separate names with a comma.