Search results

There are two main reasons why newer tiles (that I checked) are bigger. 1. More splines per tile (more roads mapped). 2. Splines are more detailed (more node points per spline / per mile). The process of localizing the car position on the spline involves finding the distances between the node...

Thanks for correction. I saw the discussion about stereo vision between main and narrow cameras, but I thought it was only about theoretical possibilities, not actual implementation.

There is definitely something going on around the maps at the moment, because the maps from "dev" source in the US got updated again recently (as in contrast to the more than 5 months of silence in version string previously). This time it looks like the maps in the whole US got updated, not just...

Thanks! Unfortunately there is nothing new in it, on the contrary, this one is even older. Its version string is "eu20160910: 2016-09-10 13:57" and its spline structure contains only 16 fields (radar zones is the first thing missing). Maybe some forgotten (not overwritten) files in the system?

Well, I don't know. Tapping directly into the Tesla servers is one of those things, that I would rather like to avoid, just to be safe. Besides, the architecture of the spline drawing is not really prepared to load too many tiles into it (just to preview a small region). All tiles from the dev...

Below is a new version of viewer tool. Here are some more notable changes in random order: Spline ID format changed from hexadecimal to decimal. Added "Go To" tool for finding coordinates and geohash boundaries. Widened chromatic scale for scale highlights. Added basic map legend (can be...

I downloaded the tile with Tesla HQ from "blacklist" branch to check and at this point current version from blacklist branch seems to be identical to current version from dev branch. Maybe at some point they were testing this feature and created a new source for it? However, while checking this...

This is a minor thing, but you are confusing two concepts - you are talking about spline nodes/points, but highlighting on the screenshots the number of elements in Local Speeds list. Those two does not have to be related. There are 18 and 46 nodes on those splines, which is still a lot. I...

Thanks for pointing this out. If disclosing the location of the tiles is not a problem please notify me about this kind of errors and warnings (providing the tile name), because this can mean that there is something new, or at least something that I didn't seen yet in those tiles.

So here’s what I found after comparing debug info provided by verygreen with the tiles data. DEBUGgpsLat and DEBUGgpsLon are displayed with barely three digits after the decimal point. That’s not enough to determine the road, because it’s approximating position within like 100m range. Much more...

You are digging into this too much. All names, terms and concepts in the viewer are put there by me, and I'm not a map/GPS expert. This is the first time I'm working on such topic. To this point there are no actual names defined by Tesla engineers in the viewer, because tile files does not...

Holy... that's a lot of text. Thanks for that! This will take some time to digest.

Ah, now I understand. Yes, it is exactly as you said - there is a byte in the spline structure, that is equal to 1 when the road does NOT have a central divider. At first this line of text was just "F22: Yes/No", later I changed "F22" to "Not Divided" and this is the result. Maybe I should have...

You mean the other way around - that it is a two way street? Not really, because it is possible to find both two way and one way streets (with a single line) with that flag active. But don't pay too big attention to this name. It's just something that I came up with. It may mean something...

I believe you might confuse the roads on the maps, so maybe lets try to figure it out. At your first screenshot the highlighted spline (Id 0198BCC3) is not Road 83, its a part of Main St. Highlighted spline on your second screenshot (Id 01987D57) is not a highway, its a part of Maple Ave. Both...

If the features from the map can be seen in the debug view as variables, than yeah, that's a good idea. It just requires to find appropriate parts of the road on the map and actually drive there. One of the things that I am afraid of right now is that I will send somebody to the wild hunt...

No, I didn't seen anything that would look to be related to slope or elevation.

I'm under impression, that you should not really look at those files as "maps" in classic meaning of this word, where map is telling us what is where and is an aid in navigation. I would rather call those files a road information database. The road curvature in real world coordinates is forming...

Yes, I did consider that. It is theoretically possible to match those maps with any other maps, not just because there are boundaries of tiles, but because everything is using geographical coordinates as its coordinate system, so the Lat/Lon coordinates are know for every point, and every...

I think I need to clarify a bit. First, we still don't know how exactly those maps are created, so we can't say for sure that all Tesla cars are generating maps at the moment. We just don't know. Second, this 1cm accuracy I mentioned is the accuracy of the mechanism of storing geographical...

I promised to provide some kind of viewer for map tiles, so here is a very basic tool. First, and most importantly: Please DO NOT take the visualizations and information provided by this tool too seriously. I figured out the meaning of most of the data using "common sense", just by looking at...

I'm not able to help here, sorry. I don't have any experience with radars, never worked with any radar device, and I didn't really analyzed the radar "code" in Autopilot, just tried to figure out the meaning of some files in the snapshots. I haven't seen anything in the snapshots data that could...

Yes it does. It is still not clear exactly which field does what, but things simply make sense. One of the missing things from your firs post that bothered me was the speed limits. I think I finally found them in the maps. It looks to be specified as one per spline in meters per second. Each...

I didn't saved those so I started searching for them again. Because of this I realized that there is one more common reason for those places to show up - power cables. Low hanging cables above the road seem to be another source of radar echo. That explained some situations. This time I assumed...

Yes. Thanks to the help from verygreen I know, that one of the fields in the spline path entry is related to "radar". This field looks to be a list of sections along the spline length. That is, it is defined based on the start point and the end point along the spline (where 0.0 is the beginning...

No, I'm not able to give any definitive answer right now. A single path record can have at least 27 fields, I figured out the meaning of 6. There is still a lot of work to do. I tried to look however at the data of record defining this loop closest to the entrance, an at first glance it does...

Right now it looks like this: I'm finding Geohash of the place using an online decoder. I'm downloading the tile files using a web browser. Then processing this tile using a piece of code in C# that was created as a result of my experiments with the file format, and that right now is a horrible...

That is possible but not necessary. There are more parking lots mapped that way. I looked at some tiles to the north of Tesla HQ, and this is how Stanford Shopping Center looks like (sorry for the poor quality this time): So regardless how the creation of this maps is done, this kind of...

One more tile fragment visualization, from Tesla HQ neighborhood (forgive my poor alignment with Google maps): This one seems to be mapped to some smaller details, including their parking lot (again, forgive my obviously wrong curve smoothing):

This is what more-less the content of that tile file looks like (at least partially): Please DO NOT take this visualization too seriously, as I pretty much don't know what I'm doing, things are figured out just by looking at it, visualization is not very precise, and the method of smoothing...

Ah, you mean more blur when compared to the main camera. Yes, angular velocity do matter, but it is also possible, that fisheye and main camera were working at the time with different exposure times (those are controlled individually per camera). You can check that by comparing "Integration...

At night time Autopilot will increase the exposure time in the sensors, so naturally motion blur and HDR ghosting become more visible, in all cameras. When combined with the color calculation HDR ghosting will often result in blobs of opposite colors around bright places (because red and gray...

@stopcrazypp, for very low light condition you may consider using only a grayscale channel images ("-og" switch). Red channel has lower sensitivity and is amplified before color computation. This my inject some noise into the image. I don't expect the difference to be big, but just mentioning this.

You mean for converting replay videos? That is quite a hackish multi-step process that I used. From the beginning, step by step: As I said before I'm using ffmpeg with rawvideo output. First the replay frame files need to be concatenated (concatenation at files level, not video stream level)...

As I said earlier - below is a new version of image conversion tool with a function for converting images from the backup camera. It's accessible with "-back" switch (with no additional options, this is a straightforward conversion to RGB BMP). I was also experimenting a bit with this tool...

Well, I've used a proprietary vision library from my work and a couple lines of C++, so I can't really recommend any publicly available tool. Sorry. I could add a function to the tool I previously posted, when I have some free time.

Yes, its YUV422, but with channels stored as separate planes.

Combined video of main camera replays from all rain snapshot (Remember that the quality of compressed replay videos is bad, significantly worse than the uncompressed still images from the same camera, this is just more interesting to watch.)

The reason they are using RCCC cameras instead of RGB is not the price. RCCC sensors are certainly not cheaper than the full color ones. Those cameras have the same hardware structure, same electronics, and pretty much the same sensor design. The only difference is the different color filter...

Raindrops as seen by the fisheye camera when stationary, you were wondering about earlier: And the effect of wipers swipe, also seen by the fisheye camera:

@stopcrazypp already explained it very well and I have not much new to add. There is no Green, no Blue, not even Yellow. This is the whole palette used in my color visualizations: When I was seeing the first results in my experiments I could swore I can see blue sky and green information sign...

Here it is: videoFCW - Streamable The road is a bit dark, because I didn't wanted the sky to overexposure.

Finally took some time to extract what I was working on into a stand-alone command line tool. Putting it as an attachment. I also improved color computation since last time (aka corrected one of my mistakes). Now lots of images look almost "normal". If anyone would want to generate some good...

Below are samples from the last sunset snapshot. Main (like previously, almost full range, and dark range portion): Side cameras (only cropped to the dark range):

Use "RGB" setting to convert Bayer-like image to RGB image. So RGB is a destination format here. That suggests, that the software needs to know the format of the source image on its own. I think this applies to the use case, when you are streaming the video on-line from the camera and the...

Good point. I hadn't thought about it, and this actually might be it. Well... not really. Or at least I can't force it to cooperate. The package you linked directly doesn't contain much (just some pieces of documentation and registering tool that is missing Qt runtime to work). But readme...

This will be a bit lengthy and technical update of what I found out in the snapshot images, that looks more like a blog post. I’m posting it because it partially shows some info about the cameras, partially straightens some incorrect things I said or shown earlier. Hope it will not bore you too...

I used ffmpeg to convert it into "rawvideo" format. Ended up with stream of raw 16bit/px images, grayscale and quarter size red images interleaved. I didn't thought abut that for long and just modified the tool that I was creating for color computation to convert them frame by frame into color...

Looks like Autopilot in your car has calibrated the rest of the cameras, because now I'm observing pretty much the same brightness range on all of them. High dynamic range is also more noticeable, and I'm cutting some significant part of dark regions to make visually nice image out of it.