Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Activity
    SteinscheisserKarl
    @SteinscheisserKarl
    I used an HR adapter system for quick attach/detach to the roof. These are often times used for smartphone holders in cars. This one is on top of the housing: https://www.hr-imotion.com/rasterplatte-590-111-21.html and this one is under the roof: https://www.ebay.de/itm/Kugelgelenk-Sockel-Auto-KFZ-Halter-Halterung-mit-4-Krallen-System-RICHTER-HR/153450976551?_trksid=p2485497.m4902.l9144
    SteinscheisserKarl
    @SteinscheisserKarl
    I everything works out as expected I will be able to setup this cam in our chessclub by next week and we will be able to record videos of OTB games quite rapidly and under conditions of real tournament games...
    Wolfgang Fahl
    @WolfgangFahl
    Great - i love this ...
    SteinscheisserKarl
    @SteinscheisserKarl
    Hi all, I made the first few test Videos with the above setup yesterday night. I just wanted to upload to Github but the max filesize is 100MB - even with lfs (large file support). Although I could certainly shrink yesterdays files I am very sure that we will exceed this filesize limit in the future...
    SteinscheisserKarl
    @SteinscheisserKarl
    This raises the question how we should handle the exchange of larger videos in the future... I could setup a nextcloud or a git repo on my own server... what do you think?
    SteinscheisserKarl
    @SteinscheisserKarl
    For the moment I uploaded 2 more videos to https://github.com/SteinscheisserKarl/Chess-Testmedia/
    These are SGVR_10_Jan_2020_1.avi and SGVR_10_Jan_2020_2.avi . These are videos with different resolutions. Both Videos are not recognized by pcwawc. I am not really sure why - it was almost midnight when I recorded the OTB Game yesterday, so I had no chance to try out all the different cam settings. Maybe I should have fixed the shutter/exposure. However, warping works for SGVR_10_Jan_2020_1.avi.
    The warppoints for SGVR_10_Jan_2020_2.avi are [[80,51],[234,44],[238,202],[83,204]]
    SteinscheisserKarl
    @SteinscheisserKarl
    Hi all - time for some update... we had the chance to record some testvideos yesterday in our local chessclub.
    SteinscheisserKarl
    @SteinscheisserKarl
    We used the new raspi-cam setup (see above). The cam was mounted directly under the roof and the videos were recorded from a distance of approx. 2m with a 8mm M8 lense (which is a short telephoto lense) and written to usb stick. Pcwawc is able to recognize some of the moves.
    But: I recorded the videos with raspivid which creates h264 files (without the avi container). Surprisingly opencv can read those files so there is no need for a conversion. However these files are quite large. One of the videos is ~500MB which means I can not upload into the github repo.
    SteinscheisserKarl
    @SteinscheisserKarl
    For now I put these files in a nextcloud: https://files.klaube.net/s/Chess-Testmedia but I have to admit that I am not really happy with filesharing through Nextcloud... However, for chessgame_16.01.2020_23:49.h264 pcwawc can "see" the first couple of moves: scripts/run --input testMedia/chessgame_16.01.2020_23\:49.h264 --autowarp --event 'Testing Chess-Cam' --site 'Stuttgart, Germany' --nomoves --debug --rotation 90 --startframe 30
    SteinscheisserKarl
    @SteinscheisserKarl
    I have now uploaded all the videos I have recorded so far into the nextcloud. There are 54 Videos with ~4.4 GB of data...
    SteinscheisserKarl
    @SteinscheisserKarl
    Juri and I recorded 3 longer videos yesterday night. They can be found here: https://files.klaube.net/s/Chess-Testmedia (chessgame_23.01*)
    There are .h264 and .avi files for each video. What I found out: although opencv can read h264 files, it kind of gets the colors wrong and there are other issues as well. So I suggest you use the avi files. The files are recorded with resolution 480x480 with 25 fps. The games are approx 10-15min each.
    SteinscheisserKarl
    @SteinscheisserKarl
    I have "hacked" an alternative approach to find the chessboard on the image/video. If somebody is interested, please have a look: https://github.com/SteinscheisserKarl/ChessBoardFinder
    It is more kind of my "first try" to get something done with opencv and python....
    Robert Grant
    @rhgrant10
    hey folks, just wanted to let you know berserk once again fully implements the available Lichess API endpoints: https://pypi.org/project/berserk/0.7.0/ (cc @WolfgangFahl)
    let me know if you run into any issues :)
    Wolfgang Fahl
    @WolfgangFahl
    @rhgrant10 and @SteinscheisserKarl thx for the input - please be patient with me for a few days ...
    SteinscheisserKarl
    @SteinscheisserKarl
    I just found out that opencv VideoCapture can not only read from a device (/dev/video0) or a file but also from URL...
    My apache log suggests that it even supports partial reads (http code 206), that means opencv will not fetch a 50mb Video if only the first couple of frames are relevant...
    SteinscheisserKarl
    @SteinscheisserKarl
    As I am unhappy with the way nextcloud provides file access I start thinking if it would'nt be the most straight forward way to setup a apache and just put all the videos we have in one folder and switch on directory index for this folder... This would allow easy access to all the recorded Videos without the need to sync all the files. It would also allow travis to access these files directly... (this would not be so easy with nextcloud I assume).
    SteinscheisserKarl
    @SteinscheisserKarl
    I have added a very simplistic way to detect if the Board in the image needs rotation in my version of ChessBoardFinder.
    it is possible now to access Videos by url like: python ChessBoardFinder.py https://files.klaube.net/TK_scholarsmate30.avi --debug --startframe 5
    Robert Grant
    @rhgrant10
    Ddd
    Wolfgang Fahl
    @WolfgangFahl
    @SteinscheisserKarl very good - that makes testing a lot easier
    SteinscheisserKarl
    @SteinscheisserKarl
    Hi all, we have recorded another 7 chessvideos yesterday in our chessclub here in Stuttgart. Files can be found at https://files.klaube.net/
    I changed the naming convention for my future files a bit to "chessgame_dd.mm.yyyyThhmm.avi" so there is no colon in the filename to prevent any URL encoding issues. Furthermore there is a subdir called "h264" with the original captures from raspivid, maybe this could be of interest (try different sizes/framerates etc)
    SteinscheisserKarl
    @SteinscheisserKarl
    all recent videos have been recorded with 480x480 resolution. This seems to be a good tradeoff between filesize and quality. As we are focusing on the chessboard the "square-like" resolution also makes sense...
    SteinscheisserKarl
    @SteinscheisserKarl
    Kind of interesting is this video: https://files.klaube.net/chessgame_30.01.2020T2026.avi
    SteinscheisserKarl
    @SteinscheisserKarl
    In this video I play against Marc (he has black, I have white pieces) and in the beginning of the game we are analyzing a bit and taking back some moves. But this is not the important thing. Marc is a tall guy and he is sitting close to the board and sometimes you see that his head is in the video for extended periods especially if he goes "into a think" for some time. Then his head is over the board for 20-30 seconds occluding several squares.
    I wonder if this could be a "showstopper"...
    SteinscheisserKarl
    @SteinscheisserKarl
    you can see a good example in this video at time index 18:00 - 22:00 min. Here Marc is thinking for a long time and his head is occluding ~15-20% of the board. But just before he makes the move the board is fully visible again for ~1sec - then he makes the move.
    I tend to say that this is quite typical for some chess players while thinking. I often see players leaning over the board, but just before they move, they sit up and then they make their move.
    Wolfgang Fahl
    @WolfgangFahl
    If you'd like to have a small debug output for your raspberry you might want to try http://wiki.bitplan.com/index.php/SSD1306
    SteinscheisserKarl
    @SteinscheisserKarl
    Hi all, we recorded a total of 7 Chessgames yesterday in our chessclub. Especially Juri (gratuxri) was very busy. I just uploaded all the games to https://files.klaube.net
    Please note that up until now these are all videos of casual games or even "mockups" (all the early scholarsmate videos etc.). We just did not have the chance to record a tournament game in the recent weeks.
    I mention this because tournament games will look a bit different, no matter what tournament (blitz games, rapid or standard games)
    Wolfgang Fahl
    @WolfgangFahl
    Thx for supplying the material - your comments on real world scenarios are very important and there is quite a bit of work todo to solve all these issues. At this point i did not even get to implement the analyzer i intended to create yet. Let's see when I'll be able to put more effort on to this. Adam is currently on vacation and I might meet him again when he's back. Whom else could we "recruit" for this project? My attempts to get people from the projects mentioned in the papers in the wiki involved were not too successful. Cedric is the only positive exception.
    SteinscheisserKarl
    @SteinscheisserKarl
    Hi all. We had a Blitz tournament on our chessclub yesterday and we were able to record a total of 12 games of this tournament.
    As it was a 5-minute Blitz tournament none of the videos is longer than 10min. You can find the files at https://files.klaube.net/
    I also updated the Readme file with some content information. We now have a total of 74 videos - which is quite a number I think...
    SteinscheisserKarl
    @SteinscheisserKarl
    If you look into the Blitz tournament games you will recognize some differences to the casual games which we recorded earlier this year: there are no take-backs, players strictly play by the "touch move" rule (you touch a piece, you must move this piece if there is any legal move possible with this piece), finally you will see each player pushing the clock after the move...
    SteinscheisserKarl
    @SteinscheisserKarl
    Today I was able to record the first standard tournament chessgame. You find the game at https://files.klaube.net/chessgame_26.02.2020T1915.avi
    The file is 1.4GB in size and the game starts ~ at frame 15200. Unfortunatly the game is not very long - this is because I lost (with black pieces) after some 10 moves... This was probably my most horrible game ever...
    Wolfgang Fahl
    @WolfgangFahl
    @SteinscheisserKarl - nice. I downloaded the video and hope I'll be able to work with it soon. The pillow distortion of the image is interesting - we might have to apply the camera matrix. The lighting looks good but might be too intense. The camera seems to work with 60 Hz lighting adjustment instead of 50 Hz. To reduce video size we might want to concentrate on general move detection to filter out the relevant frames and get rid of frames that are thinking time. That will make it also easier to run tests.
    SteinscheisserKarl
    @SteinscheisserKarl
    The cam is set to 50HZ but it is running from a battery (Powerbank)... I assume it is unavoidable to see the 50HZ flickering of the surrounding lights in this DC setup (if the cam would run from a AC power source there would probably no flickering)
    SteinscheisserKarl
    @SteinscheisserKarl
    yesterday we could record the second standard tournament game: https://files.klaube.net/chessgame_27.02.2020T1930.avi
    The games lasts for ~3 hours and the file is 2.6GB large. I had only a brief look into this game but it seems of very good quality...