mirrors/opencv
2
0
Fork 0
mirror of https://github.com/opencv/opencv.git synced 2026-09-25 04:09:57 +03:00
Code Issues Packages Projects Releases Wiki Activity
37,608 Commits 5 Branches 132 Tags
7e9fc68d218fc2a8bc04382317e35f18b9574c07
Commit Graph
37608 Commits
This Branch
This Branch
All Branches
Author SHA1 Message Date
Alexander Smorkalov 7e9fc68d21 Merge pull request #29673 from abhishek-gola:lstm_optimization 
Optimize LSTM (batched input projection, weight pre-packing, parallel directions)
 2026-08-10 15:32:26 +03:00
Abhishek Gola 841e8a7984 Merge pull request #29628 from velonica0/dnn-rvv-hal-depthwise 
dnn(rvv): HAL kernel for depthwise convolution (blocked NCHWc)
 2026-08-09 21:31:43 +05:30
Abhishek Gola b13dff7b2c Merge pull request #29608 from dheerajsingh544/fix-22259-ueye-gain-5x 
videoio: fix CAP_PROP_GAIN not working with ueye backend
 2026-08-07 00:08:34 +05:30
dheeraj singh da9fec9c47 videoio: clamp ueye CAP_PROP_GAIN to SDK's documented [0,100] range 
is_SetHardwareGain's master-gain parameter is documented as 0-100;
out-of-range values are not guaranteed to be handled predictably by
the SDK, so clamp before passing through.
 2026-08-06 17:23:05 +05:30
Abhishek Gola 34600510e4 lstm optimizations 2026-08-06 16:55:09 +05:30
Abhishek Gola dcf26f8ede Merge pull request #29657 from varun-jaiswal17/fix/python-64bit-and-bfloat16-dtype 
Add dnn conformance python tests
 2026-08-05 18:06:47 +05:30
velonica0 515ad7725a activation is nullptr 2026-08-04 20:27:23 -07:00
velonica0 484dc43c22 ci: re-trigger against fixed opencv_contrib 2026-08-04 19:41:43 -07:00
velonica0 253642c1ef dnn:rvv: add register-fill fast path to depthwise convolution for wide VLEN 2026-08-04 19:41:43 -07:00
velonica0 baf21fa180 dnn:rvv: implement RVV HAL kernel for depthwise convolution 2026-08-04 19:41:43 -07:00
velonica0 7253ddaa9b dnn: add HAL replacement hook for depthwise convolution 2026-08-04 19:40:58 -07:00
Abhishek Gola 6998403472 Merge pull request #29609 from Teddy-Yangjiale/rvv-fisheye-undistortpoints 
geometry: Vectorize fisheye::undistortPoints
 2026-08-04 23:31:25 +05:30
vrooomy c964008158 remove equal_nan for older npy compatibility U20 2026-08-04 21:21:14 +05:30
Abhishek Gola 2c00686ec7 Merge pull request #29625 from velonica0/dnn-rvv-hal-pooling 
RVV HAL kernels for DNN max/average pooling
 2026-08-04 20:09:01 +05:30
vrooomy 851ee7046c adding the deny list to python similarly 2026-08-04 14:32:03 +05:30
vrooomy 6e338fb03b extract test list into a separate .py file 2026-08-04 13:46:33 +05:30
Abhishek Gola 0635bffb3d Merge pull request #29653 from lazerg/fix-29652-mser-mindiversity 
features: fix MSER discarding stable regions via minDiversity (#29652)
 2026-08-04 09:57:04 +05:30
Abhishek Gola 68a918b900 Merge pull request #29621 from YangGuanyuhan/christylinux/fix-onnx-gather-cast 
dnn: (bug fix )preserve Cast semantics after ONNX Gather
 2026-08-04 01:01:18 +05:30
vrooomy 5fcdb9b01e code cleanup 2026-08-03 14:41:43 +05:30
Lazizbek Ergashev 02466fac6b features: fix MSER discarding stable regions via minDiversity (#29652) 2026-08-03 13:13:36 +05:00
vrooomy 25ed5d4c0d map 64bit int anf bfloat16 2026-08-03 12:26:19 +05:30
velonica0 f04ff17028 dnn: fix pooling HAL doc-build warnings (unresolvable ConvState @ref and partial avgpool param docs) 2026-08-02 19:12:51 -07:00
Abhishek Gola d6eb13adb9 Merge pull request #29616 from kirtijindal14/ptcloud-viz3d 
ptcloud: OpenGL-based 3D visualization (cv::viz3d)
 2026-08-01 13:34:56 +05:30
Yang Guanyuhan 143b084be2 dnn: remove redundant ORT test skips 2026-07-31 23:42:48 +08:00
Abhishek Gola a48e427e78 Merge pull request #29629 from biconcavelens/fix/stylization-oob-single-column 
photo: fix OOB read in Domain_Filter::compute_NCfilter for degenerate masks 🤖🤖🤖
 2026-07-31 18:53:08 +05:30
Abhishek Gola 32dcb2ea6c Merge pull request #29535 from vrabaud/persistence 
Fix benign TSAN warning in TRUCO
 2026-07-31 16:43:12 +05:30
Yang Guanyuhan 7669897910 dnn: preserve Cast semantics after ONNX Mul 2026-07-31 00:21:42 +08:00
vrooomy 4ed880f045 Added dnn conformance test python scripts 2026-07-30 20:07:29 +05:30
Yang Guanyuhan 0e36cafcf4 dnn: preserve Cast semantics after ONNX Gather 2026-07-30 22:22:31 +08:00
Abhishek Gola 9548c7ea4e Merge pull request #29632 from ArneshBanerjee/fix-29568-extractchannel-inplace 
core: restore in-place support for extractChannel (#29568)
 2026-07-30 16:01:32 +05:30
Abhishek Gola 0f83d516bf Merge pull request #29630 from varun-jaiswal17/heavy_test_skip 
skip test DNNTestNetwork.AlexNet/0 on 32-bit target
 2026-07-30 15:02:28 +05:30
kirtijindal14 f3bdb6cf01 highgui: enable depth buffer for GTK3 GtkGLArea so GL_DEPTH_TEST works 2026-07-30 14:14:31 +05:30
velonica0 b907960e0b dnn:rvv: rename pooling HAL hooks to *pool3d and correct the NCDHWc layout comment 2026-07-29 23:17:19 -07:00
velonica0 55fdef8581 dnn:rvv: implement RVV HAL kernels for max and average pooling 2026-07-29 23:17:19 -07:00
velonica0 4d976ffb19 dnn: add HAL replacement hooks for max and average pooling 2026-07-29 23:17:19 -07:00
Abhishek Gola ccfa3a733b Merge pull request #29286 from abhishek-gola/cuda_layer_dispatch 
CUDA support and Layer Split + per-op executors
 2026-07-30 10:45:17 +05:30
Arnesh BanerjeeandClaude Opus 4.8 ba5b40b1a2 core: restore in-place support for extractChannel (#29568) 
extractChannel() stopped supporting in-place operation (dst aliasing
src) after commit 416bf3253 (PR #23473), which replaced

    Mat src = _src.getMat();
    _dst.create(src.dims, &src.size[0], depth);

with a single up-front

    _dst.createSameSize(_src, depth);
    ...
    Mat src = _src.getMat();

When _src and _dst reference the same array, the up-front reallocation
reshapes the shared buffer to a single channel before the multi-channel
source header is fetched. mixChannels() then sees a 1-channel source and
throws for any coi >= 1.

Move createSameSize() back to after the source Mat/UMat is obtained, so
the source header keeps the original multi-channel data alive across the
destination reallocation. This preserves the 0D/1D handling introduced
by createSameSize while restoring the pre-existing in-place contract.

Adds regression test Core_Mat.extractChannel_inplace_29568.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
 2026-07-30 03:09:43 +05:30
Abhishek Gola 3ef693c48a warning fix 2026-07-29 21:18:11 +05:30
Varun Jaiswal e0e52c14b1 restrict skip to windows 32 only 2026-07-29 20:38:18 +05:30
Abhishek Gola 34c478016d Adapt merged 5.x Scan helpers to the LayerInfo split 
# Ptr<LayerInfo> over body->prog().
 # (matches sibling parseLoop/parseIf).
 # sliceScanAxis/stackScanAxis helpers (definition order only).
 2026-07-29 20:32:42 +05:30
Abhishek Gola 58c28e1e82 wrapper-free GpuMatND forward path 2026-07-29 20:23:02 +05:30
Abhishek Gola 059a93339c code refactoring 2026-07-29 20:23:02 +05:30
Abhishek Gola 12fb9a6c24 using gpuMat instead of backend wrappers 2026-07-29 20:23:02 +05:30
Abhishek Gola 7f63fede4c added unsupported tests to denylist 2026-07-29 20:23:02 +05:30
Abhishek Gola e7dc3a9a9b added support check 2026-07-29 20:23:02 +05:30
Abhishek Gola 0843a0bb17 fusion fix 2026-07-29 20:23:02 +05:30
Abhishek Gola 2089d7fdb1 fixed test 2026-07-29 20:19:40 +05:30
Abhishek Gola dd3f3f6d9d build issue fixed 2026-07-29 20:19:40 +05:30
Abhishek Gola c48e3f7f3e changed OpData to LayerInfo 2026-07-29 20:19:40 +05:30
Abhishek Gola bc5c1b4da8 bug fix 2026-07-29 20:17:42 +05:30