mirror of
https://github.com/tesseract-ocr/tesseract.git
synced 2024-11-30 23:49:05 +08:00
2b7df59187
They were provided by Jeff Breidenbach <jbreiden@google.com>. Signed-off-by: Stefan Weil <sw@weilnetz.de>
26 lines
922 B
C++
26 lines
922 B
C++
|
|
#include "tesseract/unittest/lstm_test.h"
|
|
|
|
namespace tesseract {
|
|
|
|
// Tests that training with unicharset recoding learns faster than without,
|
|
// for Korean. This test is split in two, so it can be run sharded.
|
|
TEST_F(LSTMTrainerTest, RecodeTestKorBase) {
|
|
// A basic single-layer, bi-di 1d LSTM on Korean.
|
|
SetupTrainer("[1,1,0,32 Lbx96 O1c1]", "kor-full", "kor.unicharset",
|
|
"arialuni.kor.lstmf", false, true, 5e-4, false);
|
|
double kor_full_err = TrainIterations(kTrainerIterations);
|
|
EXPECT_LT(kor_full_err, 88);
|
|
EXPECT_GT(kor_full_err, 85);
|
|
}
|
|
|
|
TEST_F(LSTMTrainerTest, RecodeTestKor) {
|
|
// A basic single-layer, bi-di 1d LSTM on Korean.
|
|
SetupTrainer("[1,1,0,32 Lbx96 O1c1]", "kor-recode", "kor.unicharset",
|
|
"arialuni.kor.lstmf", true, true, 5e-4, false);
|
|
double kor_recode_err = TrainIterations(kTrainerIterations);
|
|
EXPECT_LT(kor_recode_err, 60);
|
|
}
|
|
|
|
} // namespace tesseract.
|