Jaeyun [Tue, 4 May 2021 10:24:06 +0000 (19:24 +0900)]
[Conf/Subplugin] get registered subplugins
Util function to get registrable subplugins list.
If conf file is not loaded (or does not exists), nnstreamer cannot get the subplugin list from configuration.
In this case (e.g., Android), conveter cannot set the pad template from subplugins, and will make negotiation failure.
To prevent this error, get subplugins from table and conf file.
Signed-off-by: Jaeyun <jy1210.jung@samsung.com>
Gichan Jang [Tue, 4 May 2021 02:19:26 +0000 (11:19 +0900)]
[TEST] Seperate postive and negative tests
Seperate positive and negative tests of the unittest_armnn
Signed-off-by: Gichan Jang <gichan2.jang@samsung.com>
Jaeyun [Fri, 30 Apr 2021 04:53:55 +0000 (13:53 +0900)]
[Test] condition to check file size
In test-util function, set condition to check file size while running a pipeline.
Signed-off-by: Jaeyun <jy1210.jung@samsung.com>
Jaeyun [Fri, 30 Apr 2021 03:06:29 +0000 (12:06 +0900)]
[Filter/Prop] parse combination option
1. parse combi option as unsigned int and check the index (less than 16).
2. fix possible mem leak (release combi option list).
Signed-off-by: Jaeyun <jy1210.jung@samsung.com>
Jaeyun [Thu, 29 Apr 2021 10:04:15 +0000 (19:04 +0900)]
[Filter] input combi option
While getting tensor info in caps negotiation, check input combi option and compare info with combined.
Signed-off-by: Jaeyun <jy1210.jung@samsung.com>
Jaeyun [Mon, 3 May 2021 08:06:43 +0000 (17:06 +0900)]
[If] invalid map info
memory map info is valid until gst_memory_unmap() is called.
Signed-off-by: Jaeyun <jy1210.jung@samsung.com>
Wook Song [Fri, 23 Apr 2021 09:42:10 +0000 (18:42 +0900)]
[Gst/MQTT/Sink] Support dynamic buffer re-allocation
This patch changes the message buffer allocation mechanism. When the
size of the incoming GStreamer buffer exceeds that of the current
message buffer allocated, the existing buffer would be freed and
re-allocated. Note that in the case the size of the message buffer
is fixed by the max-buffer-size property, this feature is disabled.
Signed-off-by: Wook Song <wook16.song@samsung.com>
Wook Song [Fri, 23 Apr 2021 06:16:41 +0000 (15:16 +0900)]
[Gst/MQTT/Sink] Introduce a new property, max-buffer-size
This patch introduces a new property, max-buffer-size, which specifies
the maximum size of a message buffer.
Signed-off-by: Wook Song <wook16.song@samsung.com>
Wook Song [Thu, 22 Apr 2021 06:21:05 +0000 (15:21 +0900)]
[Gst/MQTT] Support basic timestamp synchronization
This patch adds a feature for the basic timestamp synchronization
support. Note that this feature works under the assumption that all the
device's system date and time values have been synchronized.
Signed-off-by: Wook Song <wook16.song@samsung.com>
shortcipher3 [Fri, 30 Apr 2021 23:05:59 +0000 (17:05 -0600)]
with GST_PLUGIN_PATH set `gst-inspect-1.0 nnstreamer` works, also renamed Dockerfile to use standard naming convention
Signed-off-by: shortcipher3 <chris.jon.hall@gmail.com>
Jaeyun [Fri, 30 Apr 2021 04:55:19 +0000 (13:55 +0900)]
[Format] clang format for typecast
add space when using c-style typecast.
Signed-off-by: Jaeyun <jy1210.jung@samsung.com>
Hyoung Joo Ahn [Mon, 26 Apr 2021 02:45:03 +0000 (11:45 +0900)]
[Decoders] fix the mem leaks of the external decoders
Since the ref count is raised when `gst_buffer_get_all_memory ()` is used, `gst_memory_unref ()` has to be called after using the memory.
Signed-off-by: Hyoung Joo Ahn <hello.ahn@samsung.com>
Jaeyun [Thu, 29 Apr 2021 07:03:10 +0000 (16:03 +0900)]
[Test/NNFW] init info in testcase
Initialize tensor-info struct before calling filter callback.
Signed-off-by: Jaeyun <jy1210.jung@samsung.com>
Junhwan Kim [Fri, 16 Apr 2021 12:55:29 +0000 (21:55 +0900)]
[filter/TVM] Add tensor-filter subplugin for TVM
- `dlpack` and either `tvm` or `tvm_runtime` should be included
- Only supports tvm model compiled with relay (graph_executor)
Signed-off-by: Junhwan Kim <jejudo.kim@samsung.com>
Wook Song [Tue, 20 Apr 2021 01:45:38 +0000 (10:45 +0900)]
[Gst/MQTT] bugfix: Change default values for pub/sub-topic and client-id
This patch changes the default values which are used when the values for
the pub/sub-topic and client-id properties are not given.
Signed-off-by: Wook Song <wook16.song@samsung.com>
Wook Song [Fri, 2 Apr 2021 15:33:20 +0000 (00:33 +0900)]
[Gst/MQTT/Src] Eliminate unnecessary critical sections
This patch removes unnecessary critical sections. Mutual exclusion
controlling is only needed when trying to change the pipeline's state.
In addition, instead of locking and unlocking the element object itself,
a GMutex variable in the object is used.
Signed-off-by: Wook Song <wook16.song@samsung.com>
Wook Song [Fri, 2 Apr 2021 15:16:29 +0000 (00:16 +0900)]
[Gst/MQTT/Src] Remove unnecessary dynamic memory allocations
In order to reduce the burden from memory management, this patch removes
unnecessary dynamic memory allocations.
Signed-off-by: Wook Song <wook16.song@samsung.com>
Wook Song [Fri, 2 Apr 2021 14:40:27 +0000 (23:40 +0900)]
[Gst/MQTT/Sink] Revise mqttsink to use the MQTTAsync C client library
This patch revises the mqttsink element to use the Asynchronous MQTT
client library for C. Unnecessary dynamic allocation and lock operations
are also removed.
Signed-off-by: Wook Song <wook16.song@samsung.com>
Wook Song [Fri, 26 Mar 2021 13:18:16 +0000 (22:18 +0900)]
[Gst/MQTT/Src] Support asynchronous subscribe via paho.mqtt.c
This patch enables mqttsrc to support an asynchronous subscribe to the
given topic. In the mqttsrc implementation, the asynchronous MQTT client
library for C has been used instead of the MQTT Client library for C,
used for implementation of mqttsink.
Signed-off-by: Wook Song <wook16.song@samsung.com>
Wook Song [Thu, 18 Mar 2021 07:58:05 +0000 (16:58 +0900)]
[Gst/MQTT/Sink] Support synchronize publish via paho.mqtt.c
This patch updates mqttsink to support synchronize publish using the
MQTT protocol. Note that only basic features required to publish a
topic are currently implemented.
Signed-off-by: Wook Song <wook16.song@samsung.com>
Sangjung Woo [Thu, 29 Apr 2021 06:18:53 +0000 (15:18 +0900)]
[TCM] Rename TCM config file
To run the TCM on AHub infra, the default name of config file should be
'config.yaml'.
Signed-off-by: Sangjung Woo <sangjung.woo@samsung.com>
Jaeyun [Tue, 27 Apr 2021 07:40:03 +0000 (16:40 +0900)]
[Filter] reduce duplicated source
Code clean, add common functions for combi option and set-info callback.
Signed-off-by: Jaeyun <jy1210.jung@samsung.com>
Jaeyun [Thu, 29 Apr 2021 03:13:26 +0000 (12:13 +0900)]
[Format] fix format and code style
fix clang format error in taos-ci and code style.
Signed-off-by: Jaeyun <jy1210.jung@samsung.com>
Jaeyun [Wed, 28 Apr 2021 08:07:08 +0000 (17:07 +0900)]
[Format] fix clang format
fix clang format error in taos-ci and testcase name rule.
Signed-off-by: Jaeyun <jy1210.jung@samsung.com>
Jaeyun [Mon, 26 Apr 2021 04:59:56 +0000 (13:59 +0900)]
[Common] add init-config
Before updating tensor config, initialize config struct.
Signed-off-by: Jaeyun <jy1210.jung@samsung.com>
Wook Song [Mon, 15 Mar 2021 07:06:11 +0000 (16:06 +0900)]
[Meson/MQTT] Add an option to support MQTT
This patch adds a Meson configuration option to support MQTT. In order
to avoid build break, this option is disabled for the Debian and Tizen
packages.
Signed-off-by: Wook Song <wook16.song@samsung.com>
Wook Song [Wed, 10 Mar 2021 09:15:01 +0000 (18:15 +0900)]
[Gst/MQTT/Sink] Make it work as a dummy sink
This patch makes MQTTSink work as a dummy sink that has a single sink
pad with "ANY" caps.
Signed-off-by: Wook Song <wook16.song@samsung.com>
Wook Song [Tue, 9 Mar 2021 02:01:28 +0000 (11:01 +0900)]
[Meson/MQTT] Generate libgstmqtt including MQTT GstElements
This patch modifies the meson build scripts to add the skeleton code of
MQTT plugins and generate libgstmqtt, the collection of them.
Signed-off-by: Wook Song <wook16.song@samsung.com>
Wook Song [Tue, 9 Mar 2021 01:52:29 +0000 (10:52 +0900)]
[Gst/MQTT] Add skeleton codes for MQTT plugins
This patch adds skeleton codes of sink and src elements for MQTT support.
Signed-off-by: Wook Song <wook16.song@samsung.com>
Gichan Jang [Tue, 27 Apr 2021 09:51:35 +0000 (18:51 +0900)]
[Filter] python3 helper funciton
nnstreamer_python3 package include python3 helper functiom.
Change it to use by converter and decoder.
Signed-off-by: Gichan Jang <gichan2.jang@samsung.com>
Jaeyun [Fri, 23 Apr 2021 10:58:17 +0000 (19:58 +0900)]
[Sink] accept flex tensor
add pad template to handle flexible tensor stream.
Signed-off-by: Jaeyun <jy1210.jung@samsung.com>
Gichan Jang [Wed, 14 Apr 2021 04:57:14 +0000 (13:57 +0900)]
[Converter] Change external converter param
Change external converter param
- Remove unnecessary param (frames_in, frame_size)
Signed-off-by: Gichan Jang <gichan2.jang@samsung.com>
Jaeyun [Tue, 20 Apr 2021 09:54:18 +0000 (18:54 +0900)]
[Common] predefined mime type
Code clean, compare predefined mime type - other/tensor(s).
Signed-off-by: Jaeyun <jy1210.jung@samsung.com>
Yongjoo Ahn [Fri, 9 Apr 2021 02:50:01 +0000 (11:50 +0900)]
[filter/tflite] Add enum for tflite delegates
- Add enum for delegates
- Refactor parsing custom props
Signed-off-by: Yongjoo Ahn <yongjoo1.ahn@samsung.com>
Hyoung Joo Ahn [Fri, 23 Apr 2021 02:31:51 +0000 (11:31 +0900)]
[Decoder/BoundingBox] add a testcase for the deprecated modes
To guarantee backward compatibility with the deprecated modes(`tflite-ssd`, `tf-ssd`), some test cases are required.
Signed-off-by: Hyoung Joo Ahn <hello.ahn@samsung.com>
MyungJoo Ham [Wed, 16 Dec 2020 08:26:00 +0000 (17:26 +0900)]
[Tools/Parser] GST Object Emulators
These aux routines emulate GST Objects so that
parsers wouldn't rely on actual GStreamer libraries
and pipeline instances.
This commit is WIP and will be continued by DJ.
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>
Jaeyun [Mon, 19 Apr 2021 11:04:31 +0000 (20:04 +0900)]
[FlexTensors] define mimetype for flexible tensors
add definition to handle flexible tensors stream.
Signed-off-by: Jaeyun <jy1210.jung@samsung.com>
Jaeyun [Mon, 19 Apr 2021 10:20:57 +0000 (19:20 +0900)]
[Converter] reduce complexity
code clean, reduce code complexity.
add internal function to parse incoming caps for custom mode.
Signed-off-by: Jaeyun <jy1210.jung@samsung.com>
Hyoung Joo Ahn [Wed, 21 Apr 2021 04:26:55 +0000 (13:26 +0900)]
[Decoder/BoundingBox] revive old modes for backward compatibility
For the old-pipeline, make the previous names are usable.
Signed-off-by: Hyoung Joo Ahn <hello.ahn@samsung.com>
Hyoung Joo Ahn [Tue, 13 Apr 2021 01:21:21 +0000 (10:21 +0900)]
[Decoder/BoundingBox] change the name of option
The mode of this decoder should be decided according to the model rather than the framework.
Currently, however, some of the modes are decided by the framework: `tf-ssd`, `tflite-ssd`
If we want to use this decoder to run a TFLite model that include the `postprocess`, we have to use `tf-ssd`.
It could occur some confusion.
For this reason, I want to suggest that the name should be changed.
Signed-off-by: Hyoung Joo Ahn <hello.ahn@samsung.com>
Gichan Jang [Tue, 20 Apr 2021 04:55:27 +0000 (13:55 +0900)]
[indent] Change indent style
Change coding style
- but no space after unary operators:
& * + - ~ ! sizeof typeof alignof __attribute__ defined
reference: https://www.kernel.org/doc/html/v4.10/process/coding-style.html
Signed-off-by: Gichan Jang <gichan2.jang@samsung.com>
Gichan Jang [Tue, 20 Apr 2021 05:39:35 +0000 (14:39 +0900)]
[Docs] Change nnstreamer CI server link
Change nnstreamer CI server link
Signed-off-by: Gichan Jang <gichan2.jang@samsung.com>
Jihoon Lee [Wed, 10 Mar 2021 12:13:20 +0000 (21:13 +0900)]
Enable detection for nntrainer singleshot
This patch maps *.ini as nntrainer to utilize auto detection for
nntrainer.
**Self evaluation:**
1. Build test: [X]Passed [ ]Failed [ ]Skipped
2. Run test: [X]Passed [ ]Failed [ ]Skipped
Signed-off-by: Jihoon Lee <jhoon.it.lee@samsung.com>
Jaeyun [Mon, 12 Apr 2021 03:24:40 +0000 (12:24 +0900)]
[Common/Collect] check null param
add condition to check null param (public functions only)
Signed-off-by: Jaeyun <jy1210.jung@samsung.com>
Jaeyun [Fri, 9 Apr 2021 08:22:09 +0000 (17:22 +0900)]
[Mux/Merge] clear collect pad data
Clear old buffer in collect pad data when received flush event.
Signed-off-by: Jaeyun <jy1210.jung@samsung.com>
Gichan Jang [Mon, 12 Apr 2021 02:00:11 +0000 (11:00 +0900)]
[Converter] Add caps filter
Add caps filter of tensor converter for _NNS_MEDIA_ANY type.
Example launchline:
```
...
tensor_filter silent=false framework=tensorflow-lite model=tflite_model/ssd_mobilenet_v2_coco.tflite ! \
tensor_decoder mode=flexbuf ! other/flexbuf ! tensor_converter silent=false ! \
other/tensors, num_tensors=2, framerate=30/1, dimensions=\"4:1:1917:1\,91:1917:1:1\", types=\"float32\,float32\" ! \
tensor_decoder mode=bounding_boxes option1=tflite-ssd option2=tflite_model/coco_labels_list.txt option3=tflite_model/box_priors.txt option4=640:480 option5=300:300 ! ... \
```
Signed-off-by: Gichan Jang <gichan2.jang@samsung.com>
Gichan Jang [Mon, 12 Apr 2021 07:50:52 +0000 (16:50 +0900)]
[Doc] Change menu tree name of contributing doc.
Change menu tree name of Documentation/contributing.md.
Related issue: #3100
Signed-off-by: Gichan Jang <gichan2.jang@samsung.com>
Hyoung Joo Ahn [Fri, 9 Apr 2021 03:38:02 +0000 (12:38 +0900)]
[Decoder/DirectVideo] add the video format RGBA
To handle various video formats, the available formats are required for the direct video decoder and add an option to handle it.
Signed-off-by: Hyoung Joo Ahn <hello.ahn@samsung.com>
Gichan Jang [Thu, 8 Apr 2021 02:21:45 +0000 (11:21 +0900)]
[TEST] Add converter and decoder subplugins test
Add tensor converter and decoder subplugin test
Signed-off-by: Gichan Jang <gichan2.jang@samsung.com>
Gichan Jang [Thu, 8 Apr 2021 06:38:51 +0000 (15:38 +0900)]
[Test] Add tensor transform mode option test
Add tensor transform mode option test
Signed-off-by: Gichan Jang <gichan2.jang@samsung.com>
Gichan Jang [Wed, 7 Apr 2021 05:18:14 +0000 (14:18 +0900)]
[TEST] Add tensor common unittests
Add unittest for tensor common functions.
Signed-off-by: Gichan Jang <gichan2.jang@samsung.com>
Gichan Jang [Tue, 6 Apr 2021 07:43:59 +0000 (16:43 +0900)]
[TCM] Modify tca config files
- Move android and tizen config file of the TCM to nnstreamer/api repo
- Modify nnstreamer config file of the TCM
Signed-off-by: Gichan Jang <gichan2.jang@samsung.com>
Gichan Jang [Tue, 6 Apr 2021 07:40:06 +0000 (16:40 +0900)]
[Converter/Decoder] Add error handling code and testcases
Add error handling code for tensor converter and decoder subplugins
Add testcases for tensor converter and decoder subplugins
Signed-off-by: Gichan Jang <gichan2.jang@samsung.com>
Gichan Jang [Tue, 6 Apr 2021 04:41:39 +0000 (13:41 +0900)]
[Converter] Define function to find subplugins
Define function to find converter subplugins.
Signed-off-by: Gichan Jang <gichan2.jang@samsung.com>
Yongjoo Ahn [Fri, 2 Apr 2021 09:45:41 +0000 (18:45 +0900)]
[aggregator/converter] Fix when GST_EVENT_FLUSH_*
- Do not reset tensor config when flush event occurs, just clear the buffers
Signed-off-by: Yongjoo Ahn <yongjoo1.ahn@samsung.com>
Jaeyun [Tue, 6 Apr 2021 03:22:17 +0000 (12:22 +0900)]
[Mux/Merge] change property name
change property name (sync mode and option), remove underbar in property name.
Signed-off-by: Jaeyun <jy1210.jung@samsung.com>
Dongju Chae [Mon, 5 Apr 2021 11:17:47 +0000 (20:17 +0900)]
[SRNPU/FILTER] Add auto detection for SRNPU subplugin
This patch adds auto detection for SRNPU subplugin.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Jaeyun [Thu, 1 Apr 2021 02:51:16 +0000 (11:51 +0900)]
[Android/AMC] clear looper
Clear looper when media steam is finished.
Signed-off-by: Jaeyun <jy1210.jung@samsung.com>
Gichan Jang [Mon, 5 Apr 2021 04:28:18 +0000 (13:28 +0900)]
Change gboolean comparison
Never directly compare the contents of a gboolean variable with the values TRUE or FALSE.
Use `if (condition)` to check a gboolean is "true", instead of `if (condition == TRUE)`. Likewise use `if (!condition)` to check a gboolean is "false".
There is no validation when assigning to a gboolean variable and so it could contain any value represented by a gint. This is why the use of if (condition) is recommended.
All non-zero values in C evaluate to "true".
Refer: https://developer.gnome.org/glib/stable/glib-Basic-Types.html#gboolean
Signed-off-by: Gichan Jang <gichan2.jang@samsung.com>
Dongju Chae [Mon, 5 Apr 2021 02:24:49 +0000 (11:24 +0900)]
[Fix/Python3] Fix install link name of python3 module
This patch fixes install link name of python3 module.
As python subplugin scripts import 'nnstreamer_python',
its name should be also matched.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Gichan Jang [Mon, 5 Apr 2021 02:19:18 +0000 (11:19 +0900)]
[Converter] Remove duplicated error handling code
Remove duplicated error handling code.
Signed-off-by: Gichan Jang <gichan2.jang@samsung.com>
Gichan Jang [Thu, 1 Apr 2021 10:08:18 +0000 (19:08 +0900)]
[Converter] Fix mem leak of the converter
A memory leak occurred because the name of the tensor was not released.
Free the name of the tensor and set default name as "" instead of
"Anonymous"
Signed-off-by: Gichan Jang <gichan2.jang@samsung.com>
Jaeyun [Thu, 1 Apr 2021 03:01:41 +0000 (12:01 +0900)]
[Transform] add regex for stand mode
1. add regex for stand mode, filter option string.
2. ignore upper/lowercase.
Signed-off-by: Jaeyun <jy1210.jung@samsung.com>
Jaeyun [Thu, 1 Apr 2021 08:00:37 +0000 (17:00 +0900)]
[Filter/EdgeTPU] tf-lite ver for EdgeTPU
When tensorflow2-lite is enabled, set dependency tf2-lite to build EdgeTPU sub-plugin.
Signed-off-by: Jaeyun <jy1210.jung@samsung.com>
Yongjoo Ahn [Mon, 29 Mar 2021 11:16:44 +0000 (20:16 +0900)]
[test/tensor_transform] Add testcases for stand
- Add SSAT testcases for stand: default and dc-average
Signed-off-by: Yongjoo Ahn <yongjoo1.ahn@samsung.com>
Yongjoo Ahn [Thu, 1 Apr 2021 05:55:53 +0000 (14:55 +0900)]
[tensor_data] Make tensor_data_raw_* return allocated memory
- `gst_tensor_data_raw_*` get double pointer for result as input. And allocate memory for it.
- Correspond change in tensor_transform and gsttensorif
- Divide with N, not N - 1 when calculate standard deviation
Signed-off-by: Yongjoo Ahn <yongjoo1.ahn@samsung.com>
Yongjoo Ahn [Mon, 29 Mar 2021 07:47:17 +0000 (16:47 +0900)]
[tensor_transform] Add "per-channel" option to stand
- "channel" is the first dimension.
- Support channel-wise default and dc-average transform
Signed-off-by: Yongjoo Ahn <yongjoo1.ahn@samsung.com>
Gichan Jang [Mon, 29 Mar 2021 05:18:01 +0000 (14:18 +0900)]
[TEST] Add apply option testcases for tensor transform
Add apply option testcases for tensor transform
Signed-off-by: Gichan Jang <gichan2.jang@samsung.com>
Gichan Jang [Mon, 29 Mar 2021 00:46:18 +0000 (09:46 +0900)]
[TRANSFORM] Add apply option
User can select tensors to apply transformation.
Signed-off-by: Gichan Jang <gichan2.jang@samsung.com>
Sangjung Woo [Tue, 30 Mar 2021 08:57:40 +0000 (17:57 +0900)]
[Test] Fix the Svace issue
ret variable is used without initialization. This patch fixes the bug.
Signed-off-by: Sangjung Woo <sangjung.woo@samsung.com>
Jaeyun [Tue, 30 Mar 2021 09:12:09 +0000 (18:12 +0900)]
[TizenSensor] init sensor type
set init value of sensor type. 'all' means no specific sensor and if user does not set type in pipeline, it will be failed.
Signed-off-by: Jaeyun <jy1210.jung@samsung.com>
Hyoung Joo Ahn [Tue, 30 Mar 2021 08:26:15 +0000 (17:26 +0900)]
[CodeQuality] add return error value
without error return, there are a lot of unreachable code blocks will be existed.
Signed-off-by: Hyoung Joo Ahn <hello.ahn@samsung.com>
Hyoung Joo Ahn [Tue, 30 Mar 2021 08:25:58 +0000 (17:25 +0900)]
[CodeQuality] check the variables before using it
check the allocated or passed variables has proper state.
Signed-off-by: Hyoung Joo Ahn <hello.ahn@samsung.com>
Hyoung Joo Ahn [Tue, 30 Mar 2021 08:25:35 +0000 (17:25 +0900)]
[CodeQuality] release memory when test is failed
when a test case is finalized with ASSERT, the allocated memory should be freed properly.
Signed-off-by: Hyoung Joo Ahn <hello.ahn@samsung.com>
Hyoung Joo Ahn [Tue, 30 Mar 2021 02:47:36 +0000 (11:47 +0900)]
[TensorIf] add a NULL checker
To improve code quality, I want to suggest a null checker.
Signed-off-by: Hyoung Joo Ahn <hello.ahn@samsung.com>
Gichan Jang [Tue, 30 Mar 2021 04:42:04 +0000 (13:42 +0900)]
[Doxygen] Add missing doxygen tag
Add missing doxygen tag for tensorrt
Signed-off-by: Gichan Jang <gichan2.jang@samsung.com>
Gichan Jang [Tue, 30 Mar 2021 01:10:29 +0000 (10:10 +0900)]
[Subplugins] Add extern "C"
Add extern "C" in converter, decoder and filter subplugins.
Signed-off-by: Gichan Jang <gichan2.jang@samsung.com>
Gichan Jang [Mon, 29 Mar 2021 05:48:45 +0000 (14:48 +0900)]
[Flatbuffers] Change function name
Change the name of functions to be consistent.
Signed-off-by: Gichan Jang <gichan2.jang@samsung.com>
Junhwan Kim [Tue, 30 Mar 2021 06:28:34 +0000 (15:28 +0900)]
[Doc] Fix typo in md files
Typo correction in md files.
Signed-off-by: Junhwan Kim <jejudo.kim@samsung.com>
Sangjung Woo [Fri, 26 Mar 2021 03:47:37 +0000 (12:47 +0900)]
[Spec] Remove archiving script
This patch removes an archiving script, which is not necessary.
Signed-off-by: Sangjung Woo <sangjung.woo@samsung.com>
Jihoon Lee [Tue, 23 Mar 2021 04:27:13 +0000 (13:27 +0900)]
[RPM] Add virtual conf package
Add virtual conf package to resolve build dependency issue as suggested
in https://github.com/nnstreamer/api/pull/21#issuecomment-
804599118
With this patch, it is now possible to install devel package with
`/etc/nntrainer.ini` or the packager choose to use `test-devel` when
building.
See also nnstreamer/nntrainer#1034
**Evaluation Method**
Cross building with nntrainer package
**Self evaluation:**
1. Build test: [X]Passed [ ]Failed [ ]Skipped
2. Run test: [X]Passed [ ]Failed [ ]Skipped
Cc: Myungjoo Ham <myungjoo.ham@samsung.com>
Signed-off-by: Jihoon Lee <jhoon.it.lee@samsung.com>
Gichan Jang [Fri, 26 Mar 2021 02:35:21 +0000 (11:35 +0900)]
[CONVERTER] Change custom mode name
Changed custom mode name to custom-code fot callback type.
Signed-off-by: Gichan Jang <gichan2.jang@samsung.com>
Jaeyun [Mon, 29 Mar 2021 08:29:15 +0000 (17:29 +0900)]
[Transform/Orc] orc acceleration for tensors
condition to support orc acceleration for each tensor.
Signed-off-by: Jaeyun <jy1210.jung@samsung.com>
Gichan Jang [Fri, 26 Mar 2021 05:44:17 +0000 (14:44 +0900)]
[Flatbuffer] Remove memcpy in converter
Remove memory copy in converter.
Memory allocation occurs two times on g_memdup and gst_memory_new_wrapped.
Changed to use gst_memory_share.
No memory copy is performed and the memory region is simply shared.
Signed-off-by: Gichan Jang <gichan2.jang@samsung.com>
Gichan Jang [Mon, 29 Mar 2021 01:49:11 +0000 (10:49 +0900)]
[Doc] Modify tensor_if README.md
- Remove description of unsupported properties.
- Elaborate the description.
Signed-off-by: Gichan Jang <gichan2.jang@samsung.com>
Xavier Roumegue [Thu, 25 Mar 2021 15:02:34 +0000 (15:02 +0000)]
tests/nnstreamer_repo_dynamicity: Fix unittest_repo binary path handling
In case the unit tests are run from their installed directory,
unittest_repo binary can't be found in this case because install test
directory path and unittest_repo binary location are invalid.
This fixes regression introduced by
bfb6630ff and
0c02456ba
Signed-off-by: Xavier Roumegue <xavier.roumegue@nxp.com>
MyungJoo Ham [Mon, 29 Mar 2021 08:16:45 +0000 (17:16 +0900)]
CODEOWNERS: Add members
We had been accepting new members.
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>
Gichan Jang [Fri, 26 Mar 2021 01:34:04 +0000 (10:34 +0900)]
[TEST] Add tensor transform testcases for tensrs stream
Add tensor transform testcases for tensrs stream
Signed-off-by: Gichan Jang <gichan2.jang@samsung.com>
Gichan Jang [Wed, 24 Mar 2021 01:52:02 +0000 (10:52 +0900)]
[TRANSFORM] Changed to support tensors
Changd tensor transform to support tensors
Signed-off-by: Gichan Jang <gichan2.jang@samsung.com>
Gichan Jang [Mon, 22 Mar 2021 06:39:56 +0000 (15:39 +0900)]
[TEST] Add tensor decdoer cusotom mode unittest
Add tensor decoder custom mode unittest
Signed-off-by: Gichan Jang <gichan2.jang@samsung.com>
Gichan Jang [Fri, 19 Mar 2021 04:56:25 +0000 (13:56 +0900)]
[Decoder] Add decoder cusotm operation
It is similar to cusom-easy of the tensor filter.
Get an input tensors and make output media stream.
It can be used to create cusom media stream from tensors such as flexbuffer.
Output capability is application/octet-stream.
Signed-off-by: Gichan Jang <gichan2.jang@samsung.com>
Dongju Chae [Fri, 12 Mar 2021 03:54:31 +0000 (12:54 +0900)]
[grpc/debian] Disable grpc packages in debian distro
This patch disables grpc packages in debian distro.
For now, gRPC is only supported in Tizen env.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Thu, 24 Dec 2020 06:23:47 +0000 (15:23 +0900)]
[gRPC/Test] Add gRPC unittests using asynchronous APIs
This patch adds gRPC unittests using asynchronous APIs.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Jaeyun [Thu, 25 Mar 2021 10:17:36 +0000 (19:17 +0900)]
[Build/Android] prepare subplugins for flatbuf
Prepare Android API build, add definition for flatbuf/flexbuf subplugins.
Signed-off-by: Jaeyun <jy1210.jung@samsung.com>
Gichan Jang [Thu, 18 Mar 2021 10:45:22 +0000 (19:45 +0900)]
[TEST] Add tensor converter custom mode unittest
Add tensor converter custom mode unittest.
Signed-off-by: Gichan Jang <gichan2.jang@samsung.com>
Gichan Jang [Wed, 17 Mar 2021 07:31:44 +0000 (16:31 +0900)]
[Converter] Add converter cusotm operation
Add converter custom operation.
It is similar to cusom-easy of the tensor filter.
Get an input buffer and make output tensor.
It can be used to create tensors from not supported capabilities or
flexible data structure such as flexbuffer.
Signed-off-by: Gichan Jang <gichan2.jang@samsung.com>
Gichan Jang [Fri, 12 Mar 2021 01:13:12 +0000 (10:13 +0900)]
[TEST] Add flexbuf test for converter and decoder
Add flexbuf test cases for converter and decoder
Signed-off-by: Gichan Jang <gichan2.jang@samsung.com>
Gichan Jang [Thu, 11 Mar 2021 08:45:48 +0000 (17:45 +0900)]
[Converter] Add flex buffer converter
Add tensor converter subplugin of flexbuffer.
Convert flexbuffer to tensor(s)
Example launch line:
gst-launch-1.0 videotestsrc ! video/x-raw,format=RGB,width=320,height=240 ! tensor_converter ! tensor_decoder mode=flexbuf ! other/flexbuf ! tensor_converter ! tensor_decoder mode=direct_video ! videoconvert ! videoscale ! video/x-raw,width=320,height=240 ! xvimagesink
Signed-off-by: Gichan Jang <gichan2.jang@samsung.com>
Gichan Jang [Thu, 11 Mar 2021 08:00:46 +0000 (17:00 +0900)]
[Decoder] Add flex buffer decoder
Add tensor decoder subplugin offlexbuffer.
Decode tensor(s) to flexbuffer.
Signed-off-by: Gichan Jang <gichan2.jang@samsung.com>
Sangjung Woo [Wed, 24 Mar 2021 05:16:20 +0000 (14:16 +0900)]
[Spec] Add configuration for VD build
* Disable NNFW, ArmCL, Viviante, pytorch, caffe2 & tensorflow.
Signed-off-by: Sangjung Woo <sangjung.woo@samsung.com>