What's new in OCRopus 0.3.1
Apr 30, 2009
- image understanding related code is now in a separate project (www.iulib.org)
- move to autoconf/automake build system
- support for beam search decoding (OpenFST is now optional)
- fixed memory leaks in Voronoi page segmentation code
- ported Voronoi page segmentation code to 64-bit architecture
- word segmentation
- refactored text/image segmentation code
- optional access to text/image segmentation from Leptonica
- inclusion of images and other non-text zones in the hOCR output
- optional inclusion of character boxes in the hOCR output
- improved edit distance with block movement for OCR evaluation
- scripts for evaluation of Layout and OCR performance
- bug fixes in bpnet (normalization of input data, right loading of a bpnet)
- feature selection, feature size selection, fix bug in skeleton feature, added connected component segmenter for train and test
- scripts for training/testing bpnet, documentation
- better loggers for training and testing (confusion matrix, lineinfo, segmenter, fst, etc)
- more accurate text line analysis