rolisz's site

Character segmentation overfitting

I'm doing a project about doing OCR on receipts and today, while trying to do character seg­men­ta­tion, I made a pretty stupid mistake that led to my model over­fit­ting almost perfectly pretty neatly (in some cases I got 100% correct clas­si­fi­ca­tion accuracy).

I already had my own data about letters (with the help of my parents, I labeled 7000 letters, with their bounding boxes in about 25 receipts) and my classifier (a simple linear SVM) on individual letters did pretty good: between 90-94% accuracy. For something obtained with almost 0 fiddling, it's pretty good, and good enough for my purposes. Also it's pretty much impossible to tell apart 0 continue.

Google Maps

[

](http://maps.google.ro/?ie=UTF8&ll=47.341644,22.332137&spn=0.004711,0.01134&z=17)

Să nu vă bazați pe Google Maps dacă veniți prin Marghita :P

Windows Visual Styles

Azi am vrut sa imi schimb Visual Styleul la Windows 7 si dupa multe cautari pe DeviantArt am gasit cateva care themeuri care imi placeau. M-am hotarat sa incep cu Appows2010 de neiio. Dupa jumate de ora de incercari de a schimba 4 fisiere din system32 fara succes, trec pe Linux si le schimb acolo. Dupa restart totul e ok, dar themeul ales de mine e prea inchisa la culoare si are contrast prea slab, abia vedeam butoanele pentru inchis si scroll­barul (am inceput cu varianta Peek) asa ca incep sa schimb fisierele pentru Lite. Linux, copy, restart, dar un anumit fisier am uitat sa copiez, dar imi dau seama ca continue.