* init README of deepdoc, add picture processer. * add resume parsing
* rename vision, add layour and tsr recognizer * trivial fixing