summary |
shortlog | log |
commit |
commitdiff |
tree
first ⋅ prev ⋅ next
Tae-Young Chung [Thu, 16 Jun 2022 05:19:11 +0000 (14:19 +0900)]
Change output type from 32float to unsigned 16bit integer
The output is now real depth range, distance, not disparity.
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Tae-Young Chung [Fri, 6 May 2022 07:43:01 +0000 (16:43 +0900)]
Set pointCloudData and pointCloudSize to nullptr and zero
[Version] 1.0.4-0
[Issue type] bug fix
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Tae-Young Chung [Mon, 2 May 2022 02:33:09 +0000 (11:33 +0900)]
Change mP1 and mP2 values
[Version] 1.0.3-0
[Issue type] update
dfs now handles gray level color space. In case of that, the values of
mP1 and P2 should be adjusted.
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Tae-Young Chung [Tue, 26 Apr 2022 07:35:55 +0000 (16:35 +0900)]
Include and install camera calibration data
[Version] 1.0.2-2
[Issue type] update
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Tae-Young Chung [Thu, 27 Jan 2022 05:54:28 +0000 (14:54 +0900)]
Change constant 16 and 1.0/16.0 values to constexpr
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Tae-Young Chung [Mon, 24 Jan 2022 05:30:03 +0000 (14:30 +0900)]
Add stereoConfigPath to get stereoConfig information for point cloud generation
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Tae-Young Chung [Fri, 14 Jan 2022 06:47:48 +0000 (15:47 +0900)]
Change depth data type to float
[Version] 1.0.1
[Issue type] update
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Tae-Young Chung [Fri, 3 Dec 2021 05:04:45 +0000 (14:04 +0900)]
Create cv::StereoSGBM instance with initial values and change parameters values for post-processing
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Tae-Young Chung [Thu, 18 Nov 2021 01:07:49 +0000 (10:07 +0900)]
Fix resize bug
When resized images are back to their original sizes,
cv::Size() should be used, not an integer scale value
to handle the case of odd number size.
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Tae-Young Chung [Wed, 10 Nov 2021 05:55:25 +0000 (14:55 +0900)]
Release 1.0.0
[Version]: 1.0.0
[Issue type]: new feature
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Tae-Young Chung [Wed, 10 Nov 2021 05:54:08 +0000 (14:54 +0900)]
Disable calibration and exclude camera parameters files from packaging
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Tae-Young Chung [Wed, 10 Nov 2021 05:38:57 +0000 (14:38 +0900)]
Fix erratum
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Tae-Young Chung [Tue, 2 Nov 2021 09:01:08 +0000 (18:01 +0900)]
Apply changed DFS_DATA_TYPE_* enumeration and a case which a given source is single typed of side-by-side format
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Tae-Young Chung [Tue, 2 Nov 2021 05:04:10 +0000 (14:04 +0900)]
Add threads for computeL and computeR
In case of post-processing, LR check is required
which a left and a right disparities are mandatory.
To reduce excution time, add two threads which compute
the both disparities at the same time.
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Tae-Young Chung [Tue, 26 Oct 2021 07:57:51 +0000 (16:57 +0900)]
Add parameters of minimum and maximum disparities to Initialize()
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Tae-Young Chung [Mon, 25 Oct 2021 06:58:24 +0000 (15:58 +0900)]
Add downscaling to reduce execution time
Before running compute() for disparity estimation,
resize the input images' size to half in both width and height.
After compute(), resize the size to the original size.
The, apply WLS filter as post-process.
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Tae-Young Chung [Fri, 22 Oct 2021 07:00:00 +0000 (16:00 +0900)]
Fix to use mMinDisparity in StereoSGBM::create()
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Tae-Young Chung [Thu, 21 Oct 2021 08:19:35 +0000 (17:19 +0900)]
Add WLS filter
WLS filter provides two approaches. One is to use a left color.
Another is to use a left and a right images after getting disparity
at the right.
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Tae-Young Chung [Thu, 21 Oct 2021 04:25:06 +0000 (13:25 +0900)]
Update calibration files according to resolutions and enable remap
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Tae-Young Chung [Thu, 21 Oct 2021 03:53:29 +0000 (12:53 +0900)]
Fix build error
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Tae-Young Chung [Thu, 21 Oct 2021 01:52:59 +0000 (10:52 +0900)]
Fix base and extra types to CV_8UC3 and disable remap
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Tae-Young Chung [Wed, 20 Oct 2021 04:22:44 +0000 (13:22 +0900)]
Convert disparity to CV_8UC1 and scale down to 1/16
SGBM outs disparity map as CV_16SC type after scaling up 16 times.
To dump and check disparity results, convert them CV_8UC1 and
scale down to 1/16.
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Tae-Young Chung [Tue, 19 Oct 2021 04:21:09 +0000 (13:21 +0900)]
Add width and height parameters to Initialized()'s input parameters
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Tae-Young Chung [Tue, 19 Oct 2021 01:04:08 +0000 (10:04 +0900)]
Add InitRectifyMap() and change distortionCoeff order
mapping table, which is used to rectify distorted images,
is pre-computed in InitRectifyMap().
Before actual depth computation by compute() function of opencv,
rectification with mapping table is done by remap() function
of opencv.
The order of distortion coefficients is k1, k2, p1, p2, and k3,
not k1, k2, k3, p1, and p2.
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Tae-Young Chung [Mon, 18 Oct 2021 06:51:18 +0000 (15:51 +0900)]
Add stereoCalibZed.yaml for stereo calibration and add InitializeStereoCalibration() method and variables for calibration
The stereoCalibZed.yaml file is for zed camera only. It is added to test dfs-opencv using zed camera.
You should change the file if you try to use other camera.
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Tae-Young Chung [Thu, 14 Oct 2021 04:03:14 +0000 (13:03 +0900)]
Fix bug
base.type should be extra.type
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Tae-Young Chung [Wed, 13 Oct 2021 00:26:41 +0000 (09:26 +0900)]
implement Run() and GetDepthData()
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Tae-Young Chung [Fri, 8 Oct 2021 02:41:44 +0000 (11:41 +0900)]
Change parameters of Intialize() to DepthParameter
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Tae-Young Chung [Thu, 7 Oct 2021 09:18:14 +0000 (18:18 +0900)]
Fix errata and add LOG to check enter/leave api
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Tae-Young Chung [Thu, 7 Oct 2021 05:27:48 +0000 (14:27 +0900)]
Add Initialize()
Initialize() creates cv::StereoSGBM as a smart pointer and
initializes some parameters for DFS.
It only supports MODE_SGBM_3WAY.
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Tae-Young Chung [Tue, 5 Oct 2021 06:32:16 +0000 (15:32 +0900)]
Add dfs-opencv
This commit is to provide dummy library to test with
dfs-adaptation and mediavision's depth module.
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
Tae-Young Chung [Wed, 29 Sep 2021 02:04:35 +0000 (11:04 +0900)]
Initial commit
Signed-off-by: Tae-Young Chung <ty83.chung@samsung.com>
projects / platform / core / multimedia / dfs-opencv.git / log