mirror/tesseract
1
0
Fork 0
mirror of https://github.com/tesseract-ocr/tesseract.git synced 2024-12-21 22:07:49 +08:00
Code Issues Actions 10 Packages Projects Releases Wiki Activity
5,772 Commits 10 Branches 66 Tags 222 MiB
2a66694754
Commit Graph

5772 Commits

This Branch
This Branch
All Branches
Author SHA1 Message Date
Stefan Weil
4c8799ac40 codeql-analysis: Disable analysis of Python code 
It should be enabled again (and also the analysis for Java)
as soon as it does not compile all C++ code, too.

Signed-off-by: Stefan Weil <sw@weilnetz.de>
 2021-08-03 11:53:51 +02:00
Stefan Weil
a73e7b97a4 Add float dotproduct implementation for NEON 
Signed-off-by: Stefan Weil <stefan.weil@bib.uni-mannheim.de>
 2021-08-03 10:35:22 +02:00
Stefan Weil
bb4a1219d7 Improve setting of dot product functions via environment variable 
Apply the settings which are selected by environment variable DOTPRODUCT
after the autodetection which detects the available SIMD hardware.

'accelerate', 'fma' and 'std::inner_product' now no longer change
the setting for intSimdMatrix to 'generic' because they don't provide
their own implementation for it.

Signed-off-by: Stefan Weil <sw@weilnetz.de>
 2021-08-03 10:34:33 +02:00
Stefan Weil
2786a887cd Update codeql-analysis.yml for Tesseract autotools build 
Signed-off-by: Stefan Weil <sw@weilnetz.de>
 2021-08-01 13:23:28 +02:00
Stefan Weil
ebae27435a Update codeql-analysis.yml for Tesseract autotools build 
Signed-off-by: Stefan Weil <sw@weilnetz.de>
 2021-08-01 13:17:45 +02:00
Stefan Weil
edcf4fcd3b Fix comment 
Signed-off-by: Stefan Weil <sw@weilnetz.de>
 2021-08-01 13:17:45 +02:00
Stefan Weil
40d12d0945
Create codeql-analysis.yml 
Signed-off-by: Stefan Weil <sw@weilnetz.de>
 2021-08-01 12:54:30 +02:00
Amit D
c8bb526afb
Merge pull request #3510 from stweil/enable-float32 
Add new configure option --enable-float32 for faster LSTM with float
 2021-07-29 18:01:21 +03:00
Stefan Weil
0d0f203509 Add new configure option --enable-float32 for faster LSTM with float 
Signed-off-by: Stefan Weil <sw@weilnetz.de>
 2021-07-29 06:49:08 +02:00
Stefan Weil
553ab64d8d Rename UnicityTable<T>::get_id to UnicityTable<T>::get_index 
This prepares replacing UnicityTable<FontInfo> by FontInfoTable.

Signed-off-by: Stefan Weil <sw@weilnetz.de>
 2021-07-26 07:59:58 +02:00
Stefan Weil
c9f42ce62b
Add unittest for static TessBaseAPI object (#3509) 
Signed-off-by: Stefan Weil <sw@weilnetz.de>
 2021-07-25 14:34:43 +03:00
Stefan Weil
df1295ea6b
Simplify *_VAR_H macros (#3508) 
This avoids duplicate (and potentially inconsistent) code.

Signed-off-by: Stefan Weil <sw@weilnetz.de>
 2021-07-25 12:09:07 +03:00
Amit D
e538cd7152
Merge pull request #3486 from stweil/tfloat 
Add TFloat data type for neural network
 2021-07-25 00:03:56 +03:00
Ger Hobbelt
27597883db Implement DotProductSSE() for FAST_FLOAT 
[sw] Formatted commit message
 2021-07-24 15:14:17 +02:00
Ger Hobbelt
79e8b4f344 bugfixing the AVX2 Extract8+16 codes 
There's lines like `__m256d scale01234567 = _mm256_loadu_ps(scales)`,
i.e. loading float vectors into double vector types.

[sw] Formatted commit message
 2021-07-24 15:14:17 +02:00
Ger Hobbelt
24a29b79e5 bugfix of FMA port to FAST_FLOAT 
8 float FPs fit in a single 256bit vector (8x32)
(contrasting 4 double FPs: 4*64)

[sw] Format commit message and use float instead of TFloat
 2021-07-24 15:14:17 +02:00
Stefan Weil
472f5d9020 Add TFloat data type for neural network 
Up to now Tesseract used double for training and recognition
with "best" models.

This commit replaces double by a new data type TFloat which
is double by default, but float if FAST_FLOAT is defined.

Ideally this should allow faster training.

Signed-off-by: Stefan Weil <sw@weilnetz.de>
 2021-07-24 15:14:17 +02:00
Stefan Weil
66b77e6639 Prepare using float instead of double for LSTM calculations 
The new header file ccutils/tesstypes.h also prepares support
for larger images by introducing a new data type for image
size and coordinates (still unused).

FloatToDouble is now a local function.

Signed-off-by: Stefan Weil <sw@weilnetz.de>
 2021-07-24 13:59:37 +02:00
Stefan Weil
c3fb050daa Remove TODO comment which is no longer open 
Signed-off-by: Stefan Weil <sw@weilnetz.de>
 2021-07-24 11:20:29 +02:00
Stefan Weil
4df822a3fc
Revert "Merge pull request #3330 from Sintun/master" (#3505) 
This reverts commit 122daf1d64, reversing
changes made to 4cd56dc5f5.

Those changes caused two regressions which resulted in an assertion
or a segmentation fault.

Signed-off-by: Stefan Weil <sw@weilnetz.de>
 2021-07-22 09:04:23 +03:00
Stefan Weil
e176169a90 Remove stray spaces at line endings 
Signed-off-by: Stefan Weil <sw@weilnetz.de>
 2021-07-20 20:59:15 +02:00
Ger Hobbelt
444fe14273 Fix a couple of 'shadowed local variables' compiler warnings 
These fixes got through while I manually extracted the template work
from my mainline (warnings due to running MSVC at Level 4)

[sw]: Format commit message and use different fix for blamer.cpp

Signed-off-by: Stefan Weil <sw@weilnetz.de>
 2021-07-20 20:49:03 +02:00
Stefan Weil
0fc6d8d7f0 Add missing hint for dotproduct parameter value "fma" 
Signed-off-by: Stefan Weil <sw@weilnetz.de>
 2021-07-20 20:44:29 +02:00
Ger Hobbelt
f72d4b1fe7 NEON arch: dead ref cycle fix 
When neon_available_ is ON, the DotProduct was set to point to DotProduct,
which should have been DotProductNative, as dotProduct is the *target* global itself:
see simddetect.h --> effectively making that part of the SetDotProduct() call
identical to this (no-op) statement: `DotProduct = DotProduct;`

Also added the Neon check in the Update() API, where it exists together
with the other checks (for AVX/SSE/etc.)

[sw: formatted commit message and merged into main branch]

Signed-off-by: Stefan Weil <sw@weilnetz.de>
 2021-07-20 20:40:16 +02:00
Stefan Weil
dff7312aed Modernize code in SIMDDetect::Update 
Signed-off-by: Stefan Weil <sw@weilnetz.de>
 2021-07-20 20:16:49 +02:00
Stefan Weil
3ab8dcbf72 Use Apple Accelerate framework for training and best models 
Signed-off-by: Stefan Weil <sw@weilnetz.de>
 2021-07-20 19:27:54 +02:00
Johannes Künsebeck
3be11f12a9 Removed unused parameters declarations and definitions 2021-07-20 15:08:10 +02:00
zdenop
8dd7936475
Solve clang reporting unused variable in ExtractMicros function (#3501) 
* mark attribute as unused for compiler
* try c++17 standard https://en.cppreference.com/w/cpp/language/attributes/maybe_unused
 2021-07-18 01:59:49 +02:00
SpaceIm
b2fea77a27 fix cross-build to iOS/tvOS/watchOS 
On these OS, executables are bundle, and they need a BUNDLE DESTINATION, otherwise CMake configuration fails.
See https://cmake.org/cmake/help/latest/policy/CMP0006.html
 2021-07-17 09:29:34 +02:00
nagadomi
7fe0624838
Fix spec string of convolution layer (#3499) 2021-07-16 18:21:52 +03:00
Stefan Weil
88d4028a5a Enable pragma for SIMD also when _OPENMP is defined 
Signed-off-by: Stefan Weil <sw@weilnetz.de>
 2021-07-15 16:03:43 +02:00
Stefan Weil
f0fb6809e3 Use SIMD instructions for DotProductNative 
Signed-off-by: Stefan Weil <sw@weilnetz.de>
 2021-07-14 19:13:01 +02:00
Tadahito Yao
12e0fb4e01
Fix deadlock in lstmtraing. (#3488) 2021-07-10 10:59:10 +03:00
Stefan Weil
767fb5a177 Fix LSTMTrainerTest.BidiTest 
Signed-off-by: Stefan Weil <sw@weilnetz.de>
 2021-07-04 18:41:19 +02:00
Stefan Weil
915c29e3c8 Fix IntSimdMatrixTest.AVX2 
Fixes: 872816897a
Signed-off-by: Stefan Weil <sw@weilnetz.de>
 2021-07-04 09:07:35 +02:00
Stefan Weil
e0af8d12e6 Fix check for NEON on 32 bit ARM 
Signed-off-by: Stefan Weil <sw@weilnetz.de>
 2021-07-03 15:43:10 +02:00
Stefan Weil
158c845228 Catch another FP division by 0 (fixes issue #3483) 
Rewriting the code avoids FP operations (so makes it potentially faster)
and fixes the division by 0.

Signed-off-by: Stefan Weil <sw@weilnetz.de>
 2021-07-03 15:37:24 +02:00
Stefan Weil
4333b2cea3
Use CMAKE_SYSTEM_PROCESSOR to check for SIMD support options (#3484) 
Signed-off-by: Stefan Weil <sw@weilnetz.de>
 2021-07-02 16:23:58 +03:00
Stefan Weil
4b630a8813 Catch FP division by 0 (fixes issue #3483) 
Signed-off-by: Stefan Weil <sw@weilnetz.de>
 2021-07-02 15:04:31 +02:00
Łukasz Nocuń
38f0fdcd88
Fix CMake Linux build (#3478) 2021-06-30 00:16:51 +03:00
OgreTransporter
4d0f027f58
Bugfix OpenMP with Visual Studio (#3475) 
* Bugfix OpenMP with Visual Studio

* Test for working VS2019 update instead of first version of VS2019
 2021-06-29 21:06:38 +03:00
Stefan Weil
a701454ae5
Fix vector resize with init for all elements (issue #3473) (#3474) 
Fixes: c8b8d266d6
Fixes: 9710bc0465
Signed-off-by: Stefan Weil <sw@weilnetz.de>
 2021-06-29 21:05:29 +03:00
nagadomi
ff1062d39d
Add --reset_learning_rate option to lstmtraining (#3470) 
When the --reset_learning_rate option is specified,
it resets the learning rate stored in each layer of the network
loaded with --continue_from to the value specified by the --learning_rate option.
If checkpoint is available, it does nothing.
 2021-06-28 11:48:07 +03:00
nagadomi
d8bd78f8e2
Fix missing reset of best_error_history_ in LSTMTrainer::InitIterations() (#3469) 2021-06-27 09:26:32 +03:00
Stefan Weil
29e842df46 CI: Replace g++-8 by g++-11 for MacOS 
g++-8 is no longer installed, therefore CI fails for that compiler.

Signed-off-by: Stefan Weil <sw@weilnetz.de>
 2021-06-26 14:56:07 +02:00
nagadomi
b2fa77f8f0
Show layer specified learning rates with combine_tessdata -l (#3468) 2021-06-26 08:08:54 +03:00
Łukasz Nocuń
c583ecef29
Fix permanently disabled optimizations in CMake (#3467) 2021-06-24 21:16:23 +03:00
MonkeybreadSoftware
75e6c3ea4c
Null check for GetSourceYResolution (#3457) 
* Null check for GetSourceYResolution

Added missing NULL check to avoid crash when we read property in our tesseract wrapper.

* Added missing return value.

added -1 to return if undefined.
 2021-06-16 16:35:24 +03:00
Egor Pugin
7a308edcb1
Merge pull request #3439 from amitdo/remove-var 
Remove unused variable
 2021-05-21 23:07:26 +03:00
Amit Dovev
bf979c801a Remove unused variable 2021-05-21 20:34:09 +03:00