Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Modify OCR for inverted text #3141

Merged
merged 2 commits into from
Oct 27, 2020
Merged

Modify OCR for inverted text #3141

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
cfb1fb2
Try OCR on inverted line only if mean confidence is below 50 %
stweil Oct 26, 2020
eaf72ac
Prefer result from inverted image if the mean confidence is better
stweil Oct 26, 2020
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 2 additions & 3 deletions src/lstm/lstmrecognizer.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,7 +2,6 @@
// File: lstmrecognizer.cpp
// Description: Top-level line recognizer class for LSTM-based networks.
// Author: Ray Smith
// Created: Thu May 02 10:59:06 PST 2013
//
// (C) Copyright 2013, Google Inc.
// Licensed under the Apache License, Version 2.0 (the "License");
Expand Down Expand Up @@ -291,7 +290,7 @@ bool LSTMRecognizer::RecognizeLine(const ImageData& image_data, bool invert,
// Check for auto inversion.
float pos_min, pos_mean, pos_sd;
OutputStats(*outputs, &pos_min, &pos_mean, &pos_sd);
if (invert && pos_min < 0.5) {
if (invert && pos_mean < 0.5) {
// Run again inverted and see if it is any better.
NetworkIO inv_inputs, inv_outputs;
inv_inputs.set_int_mode(IsIntMode());
Expand All @@ -303,7 +302,7 @@ bool LSTMRecognizer::RecognizeLine(const ImageData& image_data, bool invert,
&inv_outputs);
float inv_min, inv_mean, inv_sd;
OutputStats(inv_outputs, &inv_min, &inv_mean, &inv_sd);
if (inv_min > pos_min && inv_mean > pos_mean && inv_sd < pos_sd) {
if (inv_mean > pos_mean) {
// Inverted did better. Use inverted data.
if (debug) {
tprintf("Inverting image: old min=%g, mean=%g, sd=%g, inv %g,%g,%g\n",
Expand Down