Updated 2021 (markdown)

2021.md

@@ -30,6 +30,66 @@
 <><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>
 </pre>
 
+## 2021-09-06
+
+### *__Minutes__*
+* Working on more internal tests, almost done
+* OpenCV customizable w different types and convert internally. Relieves user of having to convert.
+* Shiqi
+   * One week holiday in China
+   * Will have a talk on line (ad is in Chinese, but at end in English, talk is in English):
+      * https://mp.weixin.qq.com/s/cGFsWuxBNomaAU8wYy5e0g
+* Intel team on release activities
+   * Placeholder added to wiki 3.4, 4.5 end of this week
+   * Cleanup of incomplete 18 issues 
+   * Intern Anna testing DNN module. OpenVino accuracy checker from open model zoo
+      * 2/3s models checked for accuracy (all fine so far)
+      * Are discussing compatibility issues on the other 1/3 (size of input Tensor). Find and resolve issue next
+* Would be nice to have tuning models PyTorch and TF with the model zoo
+* Julia
+   * 3.10 python version works except for issues on MacOS
+   * Housekeeping issues -- 25
+   * Cuda backend being updated
+
+**Done:**
+* Andrey: Changed environment for building opencv-python packages with Python 3.10 and successfully built it on Windows and Linux.
+* Alexander: Reviewed, tested, helped to resolve issues with PRs: ## 20521, 20569, 20058, 20105, 20471, 20813
+* Alexander: Recreated PRs with fixes for abandoned and incomplete ## 20770 (extra # 916), 20810.
+* Alexander P: issue “Incorrect chessboard corners returned by findChessboardCornersSB #15712”:
+   * feature maps has been attached to issue;
+   * the ways of solving the is
+   * the ways of solving the issue are suggested;
+   * issue can be closed.
+* Julia: reviewed blog post by GSoC student
+
+**In progress:**
+* Andrey: Trying to include LAPACK into opencv-python packages on Windows.
+* Andrey: Fixing the build opencv-python packages on MacOS with Python 3.10.
+* Egor: Fixed a bug in contributor’s implementation of SoftNMS, added a test, exposed the function to the public API, refactored the code and sent a PR.
+* Egor: Backported CUDA implementation of YOLOV4x-mish new_coords, and sent a PR. 
+* Julia Bareeva7:37 AM
+* Alexander P: issue ORB https://github.com/opencv/opencv-python/issues/537
+   * the problem was found: the integer overflow has occurred;
+   * the problem was fixed;
+   * the test case was added;
+   **TO DO:** set tags test, rechecked performance test and create PR.
+* Julia: fixed issue w/ data layout corruption for LSTM layer on CUDA backend
+* Julia: addressed issues found on review to GSoC PR (module to save and load point cloud #20471); also added tests for Mesh' methods and refactored save\load functions, encoder finder etc.
+
+**TF use comments**
+* Keras stand alone under Google relying on stable TF deep learning, but high level pythonic API. But different APIs for embedded
+* Brand new run time executor so that all the code will converted to small inner core. 
+* Domain adaptation may be possible to embedded w/in separate domains such as robotic navigation, arms, etc
+* PyTorch LLVM to work on common stack between network definition and low level
+
+### *__To Dos__*
+* Gary
+  - [ ] GSoC Mentor payments
+
+<pre>
+<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>
+</pre>
+
 ## 2021-09-29
 
 
@@ -71,7 +131,7 @@
 ### *__To Dos__*
 * Gary
   - [ ] GSoC Mentor payments
-  - [ ] Video prep
+  - [x] Video prep
 
 <pre>
 <><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>