Search results

I had to take a break from my project and it looks like some other interesting things have happened! I recently completed a project for a DARPA challenge and there was some knowledge that could be transferred to this project. I've dusted off my code and made a number of improvements. In order to...

Regrettably, work has interfered with any progress. I'm happy to send you info on LOS data parsing as I have done that and that should be all you need for your purposes.

I've solved a couple of the issues I was having. Here's a sample of where I am right now. This is a 256x256 pixel extract from a board that I'm training on. This is a visualization of the actual LOSData file for that piece (downscaled 4 times): This is the current prediction of the LOSData...

Yes, I'm making progress. I've begun testing a couple different ML architectures (both derived from uNet). I'm having an odd issue where, during learning, the network reports nearly 100% accuracy however, even when running one of the training map tiles through the network, it's returning junk...

My initial thought was to teach the ML system to learn LOS from the map but it turns out that is unnecessary. The LOS algorithm is already set in VASL and it needs the LOS data in the map file to determine it. My goal is to generate the LOS data (visually represented by the map above) so that...

The existing LOSData files and boards will serve as the training data. The goal is to generate LOS data for previously un-LOSed boards (such as HASLs).

At this point, it doesn't look as difficult as at first glance. The LOS rules are built into VASL; what VASL needs is an encoding of the map that it can use. That encoding can be visualized as the map above. So the AI needs to learn how to create such a visualization from a standard ASL map...

I have a pretty decent translator working for the LOSData file now. Please send me your email address or a note to scott@starsman.com and I'll send you the Python file.

Close. The above is a representation of the line of site file used by VASL to calculate LOS. My initial goal is to use Machine Learning to "look" at an ASL map and automatically create such a representation. At this point, it seems like a pretty doable project and I'll update this thread as I...

Thanks to Doug, making some excellent progress. Below is a color coded LOS map. It looks like I'm going to have to dive into my Java configuration so I can run the LOS. Looking forward to that!

Once I get a few more steps forward, I'm happy to share my code. I presume you're using Python. Do you have a deeper description of your AI project? I love seeing what other folks are doing.

I installed IntelliJ and I think I've configured it correctly. I downloaded the LOS_GUI from the Dropbox location you noted above. I'm getting the following error when I try to run it: java: cannot access VASL.LOS.Map.Map bad class file: /C:/Users/scott/OneDrive/Documents/VASSAL...

I figured it out (at least mostly). Python is reading two bytes, whereas Java is writing one. I still have to figure out a relatively minor alignment issue but my big question is resolved

The file is saved through several nested OutputStreams so the presence of some metadata (hopefully ignorable...) would explain that. The blank rows are concerning and I'm not sure how to explain them. I'll probably need to look more deeply into how Java is storing the data and Python is pulling...

The code is available and very straight forward. That's one of the reasons I'm surprised to have run into some extra bytes and missing lines. I'll check it in Java to see if that helps guide me to the issue.

So, despite the wedding, I was able to dive in a bit. I've managed to extract data from the LOSData file (see below). There are a couple issues with the file that don't align with the WriteLOSData method. There were 9 bytes in the header that don't seem to be accounted for as well as 6 or so at...

I'm in Kitty Hawk so a bit far. I am trying to get my two sons in and they both live in the Raleigh area. I'll let you know of in successful.

Thanks! I'm at a wedding this weekend and will download everything and start setting it up next week. I've used Eclipse in the past but I'm not currently. It's probably better to use IntelliJ since you are.

Doug, that's fantastic info and sounds like enough to get me started. I would appreciate a copy of the latest los gui app, if you don't mind sharing a link. Thanks!

I'm interested in developing a deep learning system to find LOS even on new maps/overlays. To do so, I would like to use the current LOS system to generate training and testing data. Is the format of the LOSData file (contained within each board file) documented anywhere. Thanks!