Kwang Son [Wed, 18 Aug 2021 08:15:32 +0000 (04:15 -0400)]
test: Add ObjectDetection test
Add inference dog object detection test.
Move common shared code to helper test class.
Change-Id: I7c02d62fa1c45422ecd5550853adaf39f7fc6a5a
Signed-off-by: Kwang Son <k.son@samsung.com>
Tae-Young Chung [Wed, 18 Aug 2021 04:45:02 +0000 (04:45 +0000)]
Merge "Change inputmetadata's shape_type, data_type, color_space to string" into tizen_devel
Kwang Son [Wed, 18 Aug 2021 03:12:11 +0000 (23:12 -0400)]
test: Reduce memcpy in loadImageToSource
loadImageToBuffer and mv_source_fill_by_buffer has own memcpy functions
which are duplicated.
- Before patch
[ PASSED ] 117 tests.
real 0m24.457s
user 0m23.040s
sys 0m6.565s
- After patch
[ PASSED ] 117 tests.
real 0m23.857s
user 0m22.802s
sys 0m6.085s
Change-Id: Ibc6dbcc01a1b4a768ea6a8b9311f3484008e4ff0
Signed-off-by: Kwang Son <k.son@samsung.com>
Tae-Young Chung [Tue, 17 Aug 2021 09:35:02 +0000 (18:35 +0900)]
Change inputmetadata's shape_type, data_type, color_space to string
Those metadata are user unfriendly and difficult to use because of a lack of decscription.
They are changed to string and user can understand what they mean.
The string values are parsed and converted to proper enumerations.
Change-Id: Ia98f29790d68fe2bf2fb38e2e922cc0b9b4d80e4
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
kwang son [Wed, 18 Aug 2021 02:16:46 +0000 (02:16 +0000)]
Merge "test: Add Image classicifation models" into tizen_devel
Kwang Son [Tue, 17 Aug 2021 09:26:32 +0000 (05:26 -0400)]
test: Add Image classicifation models
Change-Id: Ie394dedaf72ffbfc76b2f69d41ddf6ee4c83cfaf
Signed-off-by: Kwang Son <k.son@samsung.com>
Tae-Young Chung [Tue, 17 Aug 2021 06:33:04 +0000 (15:33 +0900)]
Fix unsafe mix of types 'gboolean' and 'bool'
Change-Id: I547f2e907f1be1d3774026e9fe36b299f4b4cca3
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Tae-Young Chung [Tue, 17 Aug 2021 02:10:02 +0000 (11:10 +0900)]
Fix printf type of size_t from i to zu
Change-Id: I97a0b981a599ef971016eca85860fcb51df49d1f
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
kwang son [Tue, 17 Aug 2021 00:46:27 +0000 (00:46 +0000)]
Merge "test: Update assessment test instruction" into tizen_devel
Kwang Son [Fri, 13 Aug 2021 08:24:35 +0000 (04:24 -0400)]
test: Update assessment test instruction
Change-Id: Ic7782efdb199aab123dd2f0488a7b5cbc4952c91
Signed-off-by: Kwang Son <k.son@samsung.com>
Kwang Son [Fri, 13 Aug 2021 08:10:33 +0000 (04:10 -0400)]
test: Replace barcode libcheck to gtest
Change-Id: I22339de3c880f3958ee8e049a29d902846fc0acf
Signed-off-by: Kwang Son <k.son@samsung.com>
Kwang Son [Wed, 11 Aug 2021 02:02:26 +0000 (22:02 -0400)]
test: Move load image function to helper
ImageHelper already has similar function.
Change-Id: Ibeecaf6ba387a1504811c70423f9d7b216cec154
Signed-off-by: Kwang Son <k.son@samsung.com>
Kwang Son [Mon, 9 Aug 2021 03:24:45 +0000 (23:24 -0400)]
test: Add mv_test
media vision unit test based on Gtest
media vision already has testsuites and assessment but each has some limitations.
See test/README.md more details.
Change-Id: I1d62adaeac92efe5ddc9cc6fa7a350e1aa3c846a
Signed-off-by: Kwang Son <k.son@samsung.com>
Kwang Son [Thu, 5 Aug 2021 02:28:39 +0000 (22:28 -0400)]
test: Cleanup image, object detection
Base on commit
0a7ab1f50e641c4937d2ca87ab0950067688860b
alloactate mv_source, mv_engine, mv_infer in a function
so makes hard to deallocate resource and error handling.
Change-Id: Ibd9586fe5e999e3013ab989643f47b6a4547f3fa
Signed-off-by: Kwang Son <k.son@samsung.com>
Kwang Son [Wed, 4 Aug 2021 07:06:26 +0000 (03:06 -0400)]
test: Cleanup perform_face_detection
Base on commit
0a7ab1f50e641c4937d2ca87ab0950067688860b
alloactate mv_source, mv_engine, mv_infer in a function
so makes hard to deallocate resource and error handling.
Change-Id: Iaf1124821e39c7785955bc9b06a49b5997bb0430
Signed-off-by: Kwang Son <k.son@samsung.com>
Kwang Son [Wed, 4 Aug 2021 01:42:19 +0000 (21:42 -0400)]
test: Cleanup perform_facial_landmark_detection
Base on commit
0a7ab1f50e641c4937d2ca87ab0950067688860b
alloactate mv_source, mv_engine, mv_infer in a function
so makes hard to deallocate resource and error handling.
Change-Id: Ib242154ec33602a8a9788203cabc34e37e5f878d
Signed-off-by: Kwang Son <k.son@samsung.com>
Kwang Son [Mon, 2 Aug 2021 07:37:46 +0000 (03:37 -0400)]
test: Seperate resource allocation
perform_pose_landmark_detection alloactate mv_source, mv_engine,
mv_infer in a function so makes hard to deallocate resource and
error handling.
Change-Id: I80df947730b734e9cb4f95bc238d66cccbe0ca7d
Signed-off-by: Kwang Son <k.son@samsung.com>
Kwang Son [Mon, 2 Aug 2021 01:28:41 +0000 (21:28 -0400)]
test: Remove test while loop
While loop keep previous test context which contains allocated resource.
This resource check only increase code not usability.
Change-Id: I1f2bfdd17fd6305d313d98362c4549b82f214d62
Signed-off-by: Kwang Son <k.son@samsung.com>
Kwang Son [Fri, 23 Jul 2021 02:19:11 +0000 (22:19 -0400)]
test: Add movenet for pose detection
Change-Id: Iba6c9fe101b0d3f0976aac41fc94ebd332d218e6
Signed-off-by: Kwang Son <k.son@samsung.com>
Kwang Son [Thu, 15 Jul 2021 08:39:20 +0000 (17:39 +0900)]
test: Add show_menu helper functions
show_menu -> show_menu_linear: clean manual numbering
show_menu_yes_or_no: clean name_last
Change-Id: Ib2b1579cf16351a2d1299a6285b7f6e9a0aed540
Signed-off-by: Kwang Son <k.son@samsung.com>
Tae-Young Chung [Fri, 16 Jul 2021 03:38:03 +0000 (12:38 +0900)]
Remove unnecessary variables copying outputData, instead use outputData itself
Change-Id: I03f67afaaa83042812de7323f9b90714ba37f80a
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Tae-Young Chung [Wed, 14 Jul 2021 07:04:28 +0000 (16:04 +0900)]
Update testsuite to run facial landmark models with meta files
Change-Id: Id3b8c985a7dcdc55f3e034107f7e4e9f5cd3760b
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Tae-Young Chung [Wed, 14 Jul 2021 07:01:34 +0000 (16:01 +0900)]
Add facial landmark detection inference with outputmetadata
Change-Id: I03deac554f22ec9fe079b38d9562fd667b854495
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Tae-Young Chung [Wed, 14 Jul 2021 02:52:12 +0000 (11:52 +0900)]
Add coordinate to Landmark outputmetadata
Coordinate outputmetadata of Landmark indicates that
an model's landmark coordinate is based on ratio between 0 ~ 1 or
pixel. According to it, internal processes are applied to get proper
landmark results.
Change-Id: I5e24b34cf6ca2c9e9404f082382fb7fc3d662e7f
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Kwang Son [Mon, 12 Jul 2021 08:05:42 +0000 (17:05 +0900)]
[MVQA] Change mv_infer_test order
Change-Id: I5a156f2192ec59fdf276413843af06852d47c2e4
Signed-off-by: Kwang Son <k.son@samsung.com>
Tae-Young Chung [Thu, 20 May 2021 01:30:02 +0000 (10:30 +0900)]
Update testsuite to run pose landmark models with meta files
Change-Id: I62307b534309ecbbf0c4e97ff2029e62fbdaea44
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Tae-Young Chung [Thu, 20 May 2021 01:12:54 +0000 (10:12 +0900)]
Add pose landmark detection inference with outputmetadata
Support models which require heatmap decoding with additional refinement
as well as models which just provide landmark results.
Change-Id: Ic993510d2655d488ea8a43e08a56c13d2f9bc94f
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Kwang Son [Thu, 24 Jun 2021 07:20:12 +0000 (16:20 +0900)]
Clean unused package dependency
Change-Id: I0308800190cf895893f1592ae5c102799f909fee
Signed-off-by: Kwang Son <k.son@samsung.com>
Kwang Son [Thu, 24 Jun 2021 05:23:56 +0000 (14:23 +0900)]
Fix printf format
Change-Id: I010e68687d67482a05ecd77e1bff4a6b72e27c21
Signed-off-by: Kwang Son <k.son@samsung.com>
kwang son [Thu, 24 Jun 2021 03:11:48 +0000 (03:11 +0000)]
Merge "Add packaging ml_only flag" into tizen_devel
Tae-Young Chung [Thu, 20 May 2021 00:57:02 +0000 (09:57 +0900)]
Add PoseDecoder and Landmark to decode various type of pose output tensor
Change-Id: I8be806ff3522aec1f7026912b8c317055e9e16db
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
kwang son [Tue, 22 Jun 2021 03:47:45 +0000 (03:47 +0000)]
Merge "Fix gcc warning" into tizen_devel
Kwang Son [Wed, 16 Jun 2021 07:55:52 +0000 (16:55 +0900)]
Fix gcc warning
Change-Id: I2c98fb88ee4247c8976f316baf5c6592b03f5b65
Signed-off-by: Kwang Son <k.son@samsung.com>
Kwang Son [Mon, 21 Jun 2021 03:09:14 +0000 (12:09 +0900)]
Add object detection script
Change-Id: I50d3708cdfac0547064ac3bae55238a321d27a09
Signed-off-by: Kwang Son <k.son@samsung.com>
Kwang Son [Mon, 21 Jun 2021 06:11:25 +0000 (15:11 +0900)]
Add packaging ml_only flag
Change-Id: I64a51a7a3b6fda3b3411e7ddee662b415be7a478
Signed-off-by: Kwang Son <k.son@samsung.com>
Tae-Young Chung [Fri, 18 Jun 2021 03:00:39 +0000 (03:00 +0000)]
Merge "Add new Landmark, OffsetVec, DispVec, Edge outputmetadata for pose landmark detection" into tizen_devel
Tae-Young Chung [Tue, 18 May 2021 10:18:35 +0000 (19:18 +0900)]
Add new Landmark, OffsetVec, DispVec, Edge outputmetadata for pose landmark detection
Change-Id: I7a2799005ebb887cf273c672132b3d9eb5a7ce40
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Kwang Son [Wed, 16 Jun 2021 02:24:27 +0000 (11:24 +0900)]
Cmake mediavision build refactoring
- build type to RelWithDebInfo mode
- fix wrong naming(machine_learning -> inference)
- public to private include
- remove unused project
- remove unused libraries
Change-Id: Iab2866be23830c0a8d3e3296e921572fa548c108
Signed-off-by: Kwang Son <k.son@samsung.com>
Kwang Son [Tue, 15 Jun 2021 06:43:23 +0000 (15:43 +0900)]
Cmake mediavision build refactoring
- include -> target_include
- clean pkg_check_modules usage
- remove manual compile option
- remove unused libraries
- remove unused command
- remove arm neon flag
- remove manual prefix
Change-Id: Ie30dfe0e04321fef8ed42652ec06b4f0d3cc0bff
Signed-off-by: Kwang Son <k.son@samsung.com>
Kwang Son [Mon, 14 Jun 2021 09:00:23 +0000 (18:00 +0900)]
Cmake test build refactoring
- include -> target_include
- clean pkg_check_modules usage
- remove manual compile option
- remove unused libraries
- remove unused command
Change-Id: I1e3a389ac41ed18637ff15fcd8898b2652319fbf
Signed-off-by: Kwang Son <k.son@samsung.com>
Kwang Son [Mon, 14 Jun 2021 00:46:21 +0000 (09:46 +0900)]
Cmake to lower case consistency
Change-Id: I3f66547d57d01aceb87278843898b146b2fb7c14
Signed-off-by: Kwang Son <k.son@samsung.com>
kwang son [Mon, 14 Jun 2021 00:14:57 +0000 (00:14 +0000)]
Merge "Deprecate inference lic" into tizen_devel
kwang son [Mon, 14 Jun 2021 00:14:52 +0000 (00:14 +0000)]
Merge "Deprecate surveillance lic" into tizen_devel
kwang son [Mon, 14 Jun 2021 00:14:39 +0000 (00:14 +0000)]
Merge "Deprecate face lic" into tizen_devel
Kwang Son [Fri, 11 Jun 2021 07:01:59 +0000 (16:01 +0900)]
Deprecate inference lic
Inference licensed port code is support to change licensed version library
from opensource version. However practically no one used anymore.
Change-Id: I086f91800eb917ce89b6be9a86d765f8e1c643c4
Signed-off-by: Kwang Son <k.son@samsung.com>
Kwang Son [Fri, 11 Jun 2021 06:53:46 +0000 (15:53 +0900)]
Deprecate surveillance lic
Surveillance licensed port code is support to change licensed version library
from opensource version. However practically no one used anymore.
Change-Id: If77d4174ebbe55bc639c6a18eed30833ecee0fe7
Signed-off-by: Kwang Son <k.son@samsung.com>
Kwang Son [Fri, 11 Jun 2021 03:22:25 +0000 (12:22 +0900)]
Apply meta json data type
Change-Id: Ibd754d1b9aa78f594bd70fe1040c54758f477aff
Signed-off-by: Kwang Son <k.son@samsung.com>
Kwang Son [Fri, 11 Jun 2021 04:39:23 +0000 (13:39 +0900)]
Deprecate face lic
Face licensed port code is support to change licensed version library
from opensource version. However practically no one used anymore.
Change-Id: Iacefbd9de9307836442f66504fdaafeddc8bf612
Signed-off-by: Kwang Son <k.son@samsung.com>
Tae-Young Chung [Fri, 11 Jun 2021 02:07:27 +0000 (02:07 +0000)]
Merge "Update testsuite to run google face detection models with meta files" into tizen_devel
kwang son [Fri, 11 Jun 2021 00:34:48 +0000 (00:34 +0000)]
Merge "Deprecate image lic" into tizen_devel
Kwang Son [Thu, 10 Jun 2021 05:42:07 +0000 (14:42 +0900)]
AddressSanitizer: attempting double-free
mBackend is not allocated by Load function.
return error code is enough.
Change-Id: I3ece30b743d08892c97c2d325a14bd53685bdd43
Signed-off-by: Kwang Son <k.son@samsung.com>
Kwang Son [Wed, 9 Jun 2021 09:49:56 +0000 (18:49 +0900)]
Deprecate image lic
Image licensed port code is support to change licensed version library
from opensource version. However practically no one used anymore.
Change-Id: Ia5e478816c319ec0808cd3e21bd34fd6b807815f
Signed-off-by: Kwang Son <k.son@samsung.com>
kwang son [Wed, 9 Jun 2021 08:51:12 +0000 (08:51 +0000)]
Merge "Error handling" into tizen_devel
Kwang Son [Wed, 9 Jun 2021 07:34:25 +0000 (16:34 +0900)]
Deprecate barcode lic
Barcode licensed port code is support to change licensed version library
from opensource version. However practically no one used anymore.
Change-Id: I7619d78d200bee0d87de0ee015b2a2d2a179059a
Signed-off-by: Kwang Son <k.son@samsung.com>
Inki Dae [Wed, 9 Jun 2021 06:56:59 +0000 (15:56 +0900)]
packaging: Change machine learning package name
Changed machine learning package name from capi-media-vision-inference*
tp capi-media-vision-machine_learning*.
Machine learning package includes not only inference library
also learning library so change existing inference specific package name
to generic one.
Change-Id: I6e5354f802ab836a1d4f8558d149d13b341ef323
Signed-off-by: Inki Dae <inki.dae@samsung.com>
Kwang Son [Wed, 9 Jun 2021 05:24:16 +0000 (14:24 +0900)]
Error handling
Change-Id: Ifb5cbb49afd26b564203a10577f48a0004322bf7
Signed-off-by: Kwang Son <k.son@samsung.com>
Kwang Son [Tue, 8 Jun 2021 05:20:12 +0000 (14:20 +0900)]
Add EfficientDet object detection model
Change-Id: I1cae546a2decf1b20169c4b78adcf063a31190ea
Signed-off-by: Kwang Son <k.son@samsung.com>
Tae-Young Chung [Wed, 28 Apr 2021 04:49:51 +0000 (13:49 +0900)]
Update testsuite to run google face detection models with meta files
Change-Id: I95ac44eaed2d69bf11af8ad063ff06f1dcddf6df
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Inki Dae [Fri, 4 Jun 2021 03:46:30 +0000 (12:46 +0900)]
Move mv_inference into mv_machine_learning
Moved mv_inference directory in mv_machine_learning directory.
Mediavision will support not only inference API also learning API.
So it makes a mv_machine_learning directory on top of Mediavision,
and moves existing mv_inference into mv_machine_learning directory.
This is a first step for adding learning API support for Mediavision.
Change-Id: I2d8597064141b2392bfdad0854e08dadd8d98258
Signed-off-by: Inki Dae <inki.dae@samsung.com>
Tae-Young Chung [Wed, 28 Apr 2021 04:45:24 +0000 (13:45 +0900)]
Add face detection inference with new BoxInfo, Label, Number outputmetadata
Support a model without Label and Number as well.
As an example, meta files of google hosted models for face detection
are also added.
Note that face detection is basically identical with object detection except
label is one (face). So, most codes are similar to the inference codes of
object detection.
Change-Id: I58b6e4b9fb4664e5741e9adb05a0d253ce12c5da
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
sangho park [Mon, 31 May 2021 03:44:53 +0000 (12:44 +0900)]
remove unused lines
Change-Id: If41ae77001d5caedefbdabbdc5f512b2bc950ff1
Signed-off-by: sangho park <sangho.g.park@samsung.com>
Tae-Young Chung [Tue, 27 Apr 2021 02:16:57 +0000 (11:16 +0900)]
Update testsuite to run google object detection models with meta files
Change-Id: I538039aa055379e7ba308e21d32f97329860441a
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
sangho park [Wed, 26 May 2021 22:12:24 +0000 (07:12 +0900)]
remove warnings
Change-Id: I025afa35747a9caaaba855eae15eb48243043b1b
Signed-off-by: sangho park <sangho.g.park@samsung.com>
Tae-Young Chung [Tue, 27 Apr 2021 02:03:32 +0000 (11:03 +0900)]
Apply object detection inference With new BoxInfo, Label, Number outputmetadata
Support a model without Label and Number as well.
A model may not support post-op and it should be decoded
with anchor boxes manually and post-processed such as non-maximum suppression.
In this patch, those are also added.
Change-Id: I9d0dc121910c05add2c0083532d7ff42b573268e
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
sangho park [Mon, 24 May 2021 07:47:05 +0000 (16:47 +0900)]
remove warnings
Change-Id: Ib444593a0bf641bb4b0f2a3bd86fb7ea4041c304
Signed-off-by: sangho park <sangho.g.park@samsung.com>
Kwang Son [Mon, 24 May 2021 23:23:45 +0000 (08:23 +0900)]
Delete unused code comment
Change-Id: Ic01e34d9d0cd506e3bfcddd3440fe5d7c81af9a9
Signed-off-by: Kwang Son <k.son@samsung.com>
sangho park [Mon, 24 May 2021 05:08:00 +0000 (14:08 +0900)]
remove warnings
Change-Id: Ie93b8373b224b65a2c091ffc35931712b2f5b753
Signed-off-by: sangho park <sangho.g.park@samsung.com>
Tae-Young Chung [Tue, 27 Apr 2021 01:57:38 +0000 (10:57 +0900)]
Add BoxInfo, Label, Number classes to OutputMetadata for object detection
BoxInfo, Label, Number classes are required for object detection.
As an example, meta files of goolge hosted models for object detection
are also added.
Change-Id: I968267c1108a5e79ce9fbadbb5ec0d258fc38f8b
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
sangho park [Tue, 4 May 2021 02:39:54 +0000 (11:39 +0900)]
add Surveillance assessment
Usage: mv_surveillance_assessment [OPTION...] SOURCE THRESHOLD
[VQA-surveillance]
-?, --help Give this help list
--usage Give a short usage message
mediavision surveillance assessment
e.g) mv_surveillance_assessment surveillance.avi 3
Threshold range is 0 to 10
Change-Id: Ib69e260d8a1a2610c3909093fd3bc085ce35be8d
Signed-off-by: sangho park <sangho.g.park@samsung.com>
Kwang Son [Mon, 10 May 2021 06:06:57 +0000 (15:06 +0900)]
[MVQA] Add image classification verify
Change-Id: Ie46ccd79ae547de48a71b30278b6e51fca770b8e
Signed-off-by: Kwang Son <k.son@samsung.com>
Kwang Son [Fri, 23 Apr 2021 05:52:50 +0000 (14:52 +0900)]
[MVQA] Add image classification
Change-Id: Ie1c0f8bf96c0f0749c50978b8bfd9a9f298061ab
Signed-off-by: Kwang Son <k.son@samsung.com>
Hyunsoo Park [Wed, 2 Jun 2021 00:23:17 +0000 (09:23 +0900)]
Change path of configuration file
Change-Id: I957ea08224f9fb7654f0e79520f975cc2a293f08
Signed-off-by: Hyunsoo Park <hance.park@samsung.com>
Hyunsoo Park [Tue, 11 May 2021 06:07:53 +0000 (15:07 +0900)]
[Face/Barcode] Handle Error return
Change-Id: Ia1fcefc4b567fb67924113b64c9ac22ccc6094d9
Signed-off-by: Hyunsoo Park <hance.park@samsung.com>
Tae-Young Chung [Thu, 6 May 2021 07:37:53 +0000 (16:37 +0900)]
Remove MEDIA_VISION_COLORSPACE_INVALID case in convertSourceMV2GrayCV()
It is one of colorspace which is not supported.
We don't need to separate it from other not supported error cases.
Change-Id: I2a757ed7338897f5638527aac7c2daec01661386
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Tae-Young Chung [Thu, 6 May 2021 07:23:14 +0000 (16:23 +0900)]
Add new attributes to media-vision-config.json
New attributes as rotation and image enhancement features in barcode detection,
- MV_BARCODE_DETECT_ATTR_ROTATION_DEGREES
- MV_BARCODE_DETECT_ATTR_ROTATION_COUNT
- MV_BARCODE_DETECT_ATTR_ROTATION_DIRECTION
- MV_BARCODE_DETECT_ATTR_USE_ENHANCEMENT,
should be added to media-vision-config.json
Change-Id: I2d6dcbd215d50e32dd9a43ac31f7e11309aa38a0
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Tae-Young Chung [Mon, 19 Apr 2021 08:21:56 +0000 (17:21 +0900)]
Add rotation functionality and CLAHE preprocessing in Barcode module
Change-Id: I806ad1e20a2b3a436b14b26d41218dbfa3c7fe32
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
(cherry picked from commit
a9bcad59c6c72b617a836e0e4687f0a9ef52a7e6)
Tae-Young Chung [Thu, 29 Apr 2021 03:50:55 +0000 (12:50 +0900)]
Fix failure of SetInputLayerProperty
If inputMeta isn't used, property is empty resulting to failure of SetInputLayerProperty().
property, which is set to inference engine backends through inference engine interface,
should be set by mConfig, regardless of inputMeta usage.
Change-Id: I4424d3c935288ee66ac3fab966e0e06f13a22e7f
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Inki Dae [Wed, 28 Apr 2021 07:52:16 +0000 (16:52 +0900)]
mv_inference: fix coverity issue
This patch returns an error if mPreProc.Run call failed correctly.
Change-Id: Ie8b75481e773115a1253ac448175da6cd0efa99d
Signed-off-by: Inki Dae <inki.dae@samsung.com>
Tae-Young Chung [Wed, 24 Mar 2021 07:07:06 +0000 (16:07 +0900)]
Fix uninitialized Bvh* bvh_ member variable
Coverity issue(WGID 1192835): Uninitialized pointer field(CWE-457)
Change-Id: I28ee00646f3a70196dcfec08b819657148447a2d
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Tae-Young Chung [Wed, 10 Mar 2021 04:10:15 +0000 (13:10 +0900)]
Remove vulnerable strcpy() and strcat(), use memcpy() instead.
Fix svace issue PROC_USE.VULERABLE
WGID: 456309
Change-Id: Ic547e9faa06aef0fd70def6ad1e829c15b795f69
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Kwang Son [Wed, 10 Mar 2021 02:38:24 +0000 (11:38 +0900)]
Add manifest
manifest is used for apply smack rule in RPM packages.
Change-Id: I385cb7ffd3ddce4a552f6c65d9e29685dd862809
Signed-off-by: Kwang Son <k.son@samsung.com>
Kwang Son [Wed, 21 Apr 2021 04:29:11 +0000 (13:29 +0900)]
Add readme file for MVQA
Change-Id: I21a60641bac220bbb9ca3e34a4134fbaba85aa25
Signed-off-by: Kwang Son <k.son@samsung.com>
Kwang Son [Tue, 20 Apr 2021 04:27:39 +0000 (13:27 +0900)]
Add Face detection using TFlite
Change-Id: I62f9de68a9d101525e2fc0df0f5791d2ed1d4f4a
Signed-off-by: Kwang Son <k.son@samsung.com>
kwang son [Mon, 19 Apr 2021 05:22:03 +0000 (05:22 +0000)]
Merge "Change load data method" into tizen_devel
Kwang Son [Tue, 13 Apr 2021 07:36:45 +0000 (16:36 +0900)]
Change load data method
- Change method to download single data of dataset from NAS server
Change-Id: I69c46a9ab1761bdf80e3ff4913dad93de8afcaef
Signed-off-by: Kwang Son <k.son@samsung.com>
Tae-Young Chung [Fri, 9 Apr 2021 08:17:37 +0000 (17:17 +0900)]
Add meta files for google hosted models and update testsuite
Added meta files which support google hosted models for
image classification, and update testsuite to run the models with meta files.
Change-Id: If171f82501cfa3b7246358341d30b547d5f1c701
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Tae-Young Chung [Tue, 13 Apr 2021 01:45:28 +0000 (01:45 +0000)]
Merge "Output layer names should be updated by a metafile if it exist" into tizen_devel
Tae-Young Chung [Tue, 13 Apr 2021 01:44:52 +0000 (01:44 +0000)]
Merge "Add return false to setTensorBuffer() if parameters are invalid" into tizen_devel
kwang son [Mon, 12 Apr 2021 02:23:38 +0000 (02:23 +0000)]
Merge "[MVQA] Add barcode and face detection" into tizen_devel
Kwang Son [Mon, 12 Apr 2021 01:46:38 +0000 (10:46 +0900)]
[MVQA] Add barcode and face detection
Change-Id: Icb8be5d0eed217b5b06861cd1c21d61030ad79a2
Signed-off-by: Kwang Son <k.son@samsung.com>
Tae-Young Chung [Fri, 9 Apr 2021 08:03:17 +0000 (17:03 +0900)]
Output layer names should be updated by a metafile if it exist
Change-Id: I06ae65a07528163cdff0a5fc211e129b6f9607ca
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Tae-Young Chung [Fri, 9 Apr 2021 07:58:19 +0000 (16:58 +0900)]
Add return false to setTensorBuffer() if parameters are invalid
Change-Id: I35c3778513437a84cb36a09d35de97a2ed127f35
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Tae-Young Chung [Wed, 7 Apr 2021 09:48:14 +0000 (18:48 +0900)]
Add TensorBuffer class
It wraps inference_engine_tensor_buffer as a member variable.
It provides methods setTesnorBuffer(), getTensorbuffer(),
and getValue without type casting, which is annoying but important
when handling the value on postprocessing.
Change-Id: I5d7f516ae9aa3113f8e296e24f08840be0c63108
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Tae-Young Chung [Mon, 5 Apr 2021 08:31:48 +0000 (17:31 +0900)]
Add dequantization to postprocess and fix bug when there is no preprocess
Change-Id: I8bb520a7977ab61e9a8c4ecfc3349197a134b04a
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Kwang Son [Mon, 5 Apr 2021 00:49:27 +0000 (09:49 +0900)]
Add include map
Change-Id: I33441b80767fae60c738bcc83ec8b41699ccf2ae
Signed-off-by: Kwang Son <k.son@samsung.com>
Tae-Young Chung [Fri, 2 Apr 2021 00:45:09 +0000 (00:45 +0000)]
Merge "Add Metadata, PreProcess, PostProcess class" into tizen_devel
Tae-Young Chung [Wed, 17 Mar 2021 07:07:36 +0000 (16:07 +0900)]
Add Metadata, PreProcess, PostProcess class
Introduce metadata based pre-/post-process.
User can set the input/output information of a model to a metadata file(json file)
Based on information, input tensor and output tensor are preprocessed and postprocessed,
respectively.
Metadata class with InputMetadata and OutputMetadata are added.
PreProcess and PostProcess class are added.
A meta file for classification is added as an example.
It can be used as an referece for classification case.
Change-Id: Ie86b50ae32bf2f7f5d336fe4709d6ea938930c3d
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
sangho park [Tue, 30 Mar 2021 02:44:53 +0000 (11:44 +0900)]
add Face assessment
Usage: mv_face_assessment [OPTION...] SOURCE MODEL-TYPE
[VQA-face]
-?, --help Give this help list
--usage Give a short usage message
mediavision face assessment
e.g) mv_face_assessment face.png 0
ModelTypes are below:
0:haarcascade_frontalface_alt.xml
1:haarcascade_frontalface_alt2.xml
2:haarcascade_frontalface_alt_tree.xml
Change-Id: Idf3998911aab5ba4f9871a639df595634ece37b7
Signed-off-by: sangho park <sangho.g.park@samsung.com>
sangho park [Mon, 29 Mar 2021 05:16:24 +0000 (14:16 +0900)]
change VQA input/output format
e.g) mv_barcode_assessment barcode.png
2 # number of barcode
123456789 # 1'st barcode
987654321 # 2nd barcode
Change-Id: Id3017af1cf076a3338f66a0fe1c4497bc75c3166
Signed-off-by: sangho park <sangho.g.park@samsung.com>
Kwang Son [Fri, 26 Mar 2021 06:20:00 +0000 (15:20 +0900)]
[MVQA] Add benchmark database schema
Change-Id: I6d4ba715283f8960c9434e90207d736d659d22eb
Signed-off-by: Kwang Son <k.son@samsung.com>
projects / platform / core / api / mediavision.git / log