Updated 2021 (markdown)

2021.md

@@ -30,6 +30,47 @@
 <><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>
 </pre>
 
+## 2021-11-10
+
+### *__Minutes__*
+
+* Vadim:
+  * Ficus bindings for OpenCV is in progress; core & imgproc module are supported now, dnn is in progress.
+  * OpenCV China continues working on NPU support in DNN, based on Tim-VX; 8-bit convolution is now working correctly and several issues with memory management have been caught and resolved.
+* Shiqi:
+  * The first patch from the group of students from Sustech supervised by Shiqi has been merged to OpenCV: https://github.com/opencv/opencv/pull/20784
+  * 2 groups of students are working on OpenCV: the first one is working on plane detection and other point cloud processing algorithms. The second group is working on car detection in point clouds.
+  * Last week had meeting with CSDN. Jia will be working on OpenCV knowledge tree. CSDN will help. There is also mirror of OpenCV at CSDN (more friendly for Chinese audience). There is also negotiation about organization of OpenCV conference in China in Spring-summer 2022.
+* Alexander:
+  * Alexander A has created 4.x and 5.x branches. Q: What to do with existing PRs?
+  * Upgraded 3rd-party libprotobuf was upgraded to v3.19.1.
+  * Found and fixed issue in trackbar::setRange().
+  * cvat team submitted issue to OpenCV bug tracker: added MIL tracker. Deep learning-based trackers are not feasible because of the huge model size (~1Gb). **Would be nice to find and add to opencv_zoo more compact object tracking model**
+* Julia:
+  * Done:
+    * Andrey: Fixed OpenCV packaging for Python 3.9. #572
+    * Andrey: Fixed pylint no-member warnings for OpenCV python packages and autocomplete in IDE. #570
+    * Alexander P: review radon PR #3090 user fixes (after first review) was checked PR was merged
+    * Egor: reviewed #20904, it was merged.
+    * Julia: investigated ability to import TFLite networks:
+      * run simple sample to read model from flatbuffer
+      * prepared draft branch w/ flatbuffer integrated to OpenCV CMake
+    * Julia: reviewed: Uniqueness ratio support for cuda::StereoBM #3081 and #3091
+  * In progress:
+    * Andrey: Preparing source packages for old opencv-python releases.
+    * Andrey: Updating 3.4 branch after new changes in master in opencv-python repository.
+    * Alexander P: issue “gen_pattern.py doesn't work” #21026 fixed, prepared PR "fix markers parse in doc/pattern_tools/gen_pattern.py" #21028
+    * Alexander P: issue “Python binding with cv.aruco.refineDetectedMarkers #14014“ the function returns matrices of different dimensions (for similar cases); error was found, the solution is being prepared. C++ reproducer was created
+    * Egor: Figured out how onnx tests are being generated and why sionnx auto-generated tests require rewrite of onnx test infrastructure(tests are collected into a single list and we run out of RAM). Ran 717 ONNX tests on OpenCV and 133 of them seem to pass. In the process of figuring out whether code coverage increased or not.
+    * Julia: add Erf layer: added perf and accuracy tests with basic trivial implementation using std::erf to compare; speed up implementation of cv::erf from 17 ms to 9 ms (try about 5 different approximations); but it's still not faster than trivial loop with std::erf (#14357)
+* Vincent:
+  * Couple of issues have been found by the sanitiser:
+     * assert() is used instead of CV_Assert()
+     * Rect::area() overflows.
+* Gary:
+  * negotiations with Futureway are at the final stage; some funding is hopefully coming.
+  * need to design a robotics toolkit
+
 ## 2021-11-03
 
 ### *__Minutes__*