Detect chess pieces on chesstempo boards using opencv in python.
>python detect.py -h
usage: detect.py [-h] [-color {w,b}] [-t TIME] [-v] [--show_move]
[--hide_move]
[--castle {KQkq,Kkq,Qkq,kq,KQk,Kk,Qk,k,KQq,Kq,Qq,q,K,Q,KQ,-}]
file
positional arguments:
file png image filename to parse
optional arguments:
-h, --help show this help message and exit
-color {w,b} color to move, "w" or "b" (color is autodetected if
omitted)
-t TIME, --time TIME sunfish thinking time, default=5
-v, --verbose increase output verbosity and saving of status images
--show_move show best move window (default)
--hide_move hide best move window
--castle {KQkq,Kkq,Qkq,kq,KQk,Kk,Qk,k,KQq,Kq,Qq,q,K,Q,KQ,-}
castling possibilities (default: -)
Or use the provided screenshot.py script to grab a screenshot and immediately start analyzing it:
python screenshot.py
Take the screenshot 
After template matching the detected board is passed to sunfish, which calculates the optimal move and prints it in the original picture:
c:\dev\chesspy>python detect.py samples\stellung3.png
Parsing file samples\stellung3.png None
FEN 1rb1nrk1/2q2p1p/p1p3p1/2QNP3/P7/6P1/1PP2P1P/3RR1K1
Detected board: (b)
. k . r r . . .
p . p . . p p .
. p . . . . . .
. . . . . . . p
. . . p n q . .
. P . . . P . P
P . P . . Q . .
. K R N . B R .
Suggested move (score): c6d5 1181
c:\dev\chesspy>
Using the flask-server fserver.py, a webservice will listen at the configured port. Images can be transferred via POST request, which will be parsed, analyzed and the result will be printed back to the client.
Example POST call:
curl -F "file=@samples/stellung8-1.png" localhost:5000/api/test
Using http://opencv.org/ to detect the pieces and Sunfish python chess engine to analyse the board and predict the next move.
Python version is 2.7.
Flask is used to provide upload capabilities via HTTP.