Updated 2021 (markdown)

2021.md

@@ -40,6 +40,7 @@
     * fixed problem incorrect results/crashes when processing small images with Tim-VX backend.
     * most of the layer tests now pass, with a notable exception of Leaky-ReLU. Tim-VX implementation is fixed-point, so it produces a LUT slightly different from the reference C implementation.
     * problem with different hash sums of .zip archives is still unsolved and would probably be difficult to solve.
+    * experiments on per-channel vs per-tensor quantization are in progress.
 * Alexander V:
   * The team is working on pre-release package validation.
   * 4.5.5 Changelog placeholder at Wiki: https://github.com/opencv/opencv/wiki/ChangeLog#version454
@@ -47,6 +48,33 @@
   * Lapack patch has been merged: https://github.com/opencv/opencv/pull/21114
   * C++ and Python spectrogram samples has been merged: https://github.com/opencv/opencv/pull/20934
   * Integration Audio I/O functionality into GSoC speech recognition sample is going to be merged soon: https://github.com/opencv/opencv/pull/21197
+* Alexander S:
+  * Done:
+    * Egor: Merged:
+       * const/x fix for Div PR.
+       * Flatten fix for ONNX PR.
+       * Add celu, hardsigmoid, selu, thresholdedrelu layers PR.
+       * Add acos, acosh, asin, asinh, atan, atanh, cos, cosh, erf, hardswish, sin, sinh, softplus, softsign, tan layers PR.
+    * Egor: Reviewed:
+       * add argmax and argmin parsing for tensorflow #21268.
+       * dnn(test): update ONNX conformance filters (4.x) #21265.
+       * Fixed several issues found by static analysis #21267.
+    * Alexander P: Update ArUco tutorial #3126:
+       * Fixed and merged. In total:
+       * fixed ArUco, ChArUco, diamonds, calibration chapters;
+       * added read dictionary and read params API;
+       * added calibration photos;
+       * added tests of tutorial to avoid regression;
+       * added aruco_samples_utility.hpp to avoid code duplication.
+       * on sick leave currently.
+    * Andrey: Tested OpenCV python packages building after merging the PR with changes for python limited API support on all platforms and different python versions.
+  * In progress:
+    * Egor: Benchmarked attempts #1 and #2 of general broadcasting implementations, they turned out to be pretty slow. WIP on dims*log(MAX_DIM) solution - it is working and is faster than alternatives, but I think I need to get rid of allocations.
+* Shiqi:
+  * 2 PRs have been merged from the students:
+    * Square operator from TF
+    * batch Mat mul
+  * argmax, argmin for TF is in progress.
 
 <pre>
 <><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>