perf: skip frames, roi, fixed aspect ratio #225
Conversation
copy calibration tools from private repository
merge main for galactic release
…g elements for eval mode
…and calibration parameters into a file, change ui elements in eval mode.
…luation_mode Signed-off-by: Kenzo Lobos-Tsunekawa <kenzo.lobos@tier4.jp>
…rinsics_evaluation_mode Signed-off-by: Kenzo Lobos-Tsunekawa <kenzo.lobos@tier4.jp>
Signed-off-by: Kenzo Lobos-Tsunekawa <kenzo.lobos@tier4.jp>
…e' into feat/camera_profiles
...brator/intrinsic_camera_calibrator/intrinsic_camera_calibrator/camera_models/camera_model.py
Show resolved
Hide resolved
…rator/intrinsic_camera_calibrator/camera_calibrator.py Co-authored-by: Kenzo Lobos Tsunekawa <kenzo.lobos@tier4.jp>
…rator/intrinsic_camera_calibrator/camera_models/camera_model.py Co-authored-by: Kenzo Lobos Tsunekawa <kenzo.lobos@tier4.jp>
I was checking a book that has a good explanation of the rectification process, and the equations used on the undistortion map seem to be good, https://amroamroamro.github.io/mexopencv/matlab/cv.initUndistortRectifyMap.html https://books.google.ch/books?id=seAgiOfu2EIC&lpg=PA373&pg=PA373#v=onepage&q&f=false |
| # if chessboard was found, keep track of the region to try to detect easily in the next frame | ||
| if detected: | ||
| self.roi = get_roi(corners, img.shape[:2]) | ||
| self.lost_frames = 0 |
There was a problem hiding this comment.
self.lost_frames is reset in a protected way now, but it is not here
There was a problem hiding this comment.
Maybe I'm wrong, but we are protecting this value because it might change the max value on the UI side. So, is it necessary to add a lock here?
There was a problem hiding this comment.
In:
self.lost_frames = self.max_lost_frames.value
self.max_lost_frames is shared with the parameter edit widget, so it needs to be protected.
restart_lost_frames_counter modifies self.lost_frames in another thread, and as such it needs to be protected.
To avoid a flurry of locks, you can load one time into a local variable, and write (using a lock) just once after you finish whatever processing was being done.
Another choice is to use async signals with Qt instead of using a method (this is the best approach according to this framework).
If you want to understand how things go under the hood you can print the thread ids
...intrinsic_camera_calibrator/intrinsic_camera_calibrator/intrinsic_camera_calibrator/utils.py
Outdated
Show resolved
Hide resolved
..._camera_calibrator/intrinsic_camera_calibrator/intrinsic_camera_calibrator/data_collector.py
Outdated
Show resolved
Hide resolved
...mera_calibrator/intrinsic_camera_calibrator/intrinsic_camera_calibrator/camera_calibrator.py
Outdated
Show resolved
Hide resolved
…rator/intrinsic_camera_calibrator/data_collector.py Co-authored-by: Kenzo Lobos Tsunekawa <kenzo.lobos@tier4.jp>
…rator/intrinsic_camera_calibrator/utils.py Co-authored-by: Kenzo Lobos Tsunekawa <kenzo.lobos@tier4.jp>
|
This PR awaits a experimental confirmation of the rectification pipeline using camera-lidar calibration and a brief proof of how it works @SergioReyesSan @drwnz |
Description
This PR includes:
-ROI to keep track of the calibration board, making future detections faster (performance improvement).
-Will skip board detection on some future frames, if the board was not detected on the current frame (avoid long queues).
-Increase the decimal digits when saving intrinsic parameters, to avoid the bubble effect when rectifying the image.
-Fix the issue that suddenly was pausing the operation of the calibration tool.
-Implements a new method to fix the aspect ratio on the rectified image.
-Changes the distortion model according to the number of distortion coefficients.
Related links
Tests performed
Tested mainly on an x86 platform and using rosbags, but also non-extensive tests on anvil were performed.
Notes for reviewers
It might have some spelling errors, please let me know.
Pre-review checklist for the PR author
The PR author must check the checkboxes below when creating the PR.
In-review checklist for the PR reviewers
The PR reviewers must check the checkboxes below before approval.
Post-review checklist for the PR author
The PR author must check the checkboxes below before merging.
After all checkboxes are checked, anyone who has write access can merge the PR.