Open openseauser opened 5 years ago
Since this CAPTCHA always has the same number of characters, you could play with the exact_characters
setting. That'll repeatedly split the largest extracted character in half until you have exactly that many characters.
var cbl = new CBL({
preprocess: function(img) {
img.debugImage("debugPreprocessed");
img.binarize(32);
img.debugImage("debugPreprocessed");
img.colorRegions(50, true, 1);
img.debugImage("debugPreprocessed");
},
character_set: "0123456789abcdefghijklmnopqrstuvwxyz",
exact_characters: 6,
pattern_width: 24,
pattern_height: 24,
blob_min_pixels: 1,
blob_max_pixels: 10000,
allow_console_log: true,
blob_console_debug: true,
blob_debug: "debugSegmented"
});
This seems to get fairly good results. With a little work on the preprocessing steps you should be able to get near 100% accuracy on this.
Works for me, thank you!
Just fixed bug #33 which will improve the segmentation for this one.
I'm trying to train this thing to solve an easy captcha, but getting stuck a little. If the captcha has two characters that are touching then there's no combination I can try to get it to pick up both characters.
Do you know how this could be done? Maybe I'm missing a setting that should be changed? Some examples are attached.
The first one won't recognize the 5 and a, second one won't recognize the m and the d, and third one won't recognize the p or the 4.