enxio27 wrote:
Can you give a brief set of instructions?
Sure - haven't got around to a help file, as it is still very much an alpha Work In Progress
First thing is to get the grid image into the left hand window.
1) "Load" will pop up a file dialog to load an image file
2) "Paste" or Ctrl-V will paste an image from the clipboard
3) Drag and drop an image file
4) Click "Grab" - this will minimise the window and show a transparent selector window. Move and Resize this to contain an on-screen image and click the little tick mark (or press Enter).
With an image in the left window you can move it around with Left-Mouse and drag, and you can rotate left/right with the little arrow buttons. If the image is more than just a single grid, you can select the grid area using Right-Mouse and drag which will draw out a rectangular lasso (it will scroll the image if needed and you can start at any corner of the image). With a region highlighted, click "Crop" to get the area selected, zoomed and centred. If you got it wrong, click "Reset" to revert to the original image.
If it's a nice image, you can just go ahead and click "Interpret" and you should get the puzzle string in bottom right text box.
The options on the right panel control whether you're trying to read a Futoshiki or a standard Sudoku grid. Mainly they're there for me to play around with to try to optimise the selection process where the program tries to convert the image into numbers. It does this by converting to Black and White, doing some Edge Detection and then trying to work out where the cell borders are.
After pressing "Interpret" - you can take a look to see where it thinks the cells are by clicking "Cells" and you can take a look at how it has converted to black and white by clicking "Show". Again - this is mainly for me to play around with to see what's going on in the background and will probably be removed in the final version.
The program will then take each found cell and try to do a character match on it. If the character interpretation is wrong - assuming the grid ha been correctly identified - you can "Train" the program on the incorrect characters. This pops up a window showing on the left all the "Reference" images I use for character matching and on the right the "Detected" images. You can then persuade the program what the actual "Detected" images should have been interpreted as. Save the Training and you can go back and try to "Interpret" again.
Rgds
Richard