From 97b9ef7851d92a8505e4507ac88c0c2ba9390830 Mon Sep 17 00:00:00 2001 From: Mateo de Mayo Date: Wed, 20 Jul 2022 17:54:44 -0300 Subject: [PATCH] Fix issues with Index docs and toml file --- data/monado/index.toml | 2 +- doc/monado/Vive.md | 24 ++++++++++++------------ 2 files changed, 13 insertions(+), 13 deletions(-) diff --git a/data/monado/index.toml b/data/monado/index.toml index c14dd14..a4124ae 100644 --- a/data/monado/index.toml +++ b/data/monado/index.toml @@ -4,7 +4,7 @@ show-gui=1 # Ground-truth camera calibration used for simulation. -cam-calib="/home/mateo/Documents/apps/bsltdeps/basalt/data/index_calib_oxrimu.json" +cam-calib="/home/mateo/Documents/apps/bsltdeps/basalt/data/index_calib.json" # Path to config file. config-path="/home/mateo/Documents/apps/bsltdeps/basalt/data/euroc_config.json" diff --git a/doc/monado/Vive.md b/doc/monado/Vive.md index 32c4019..ec204a5 100644 --- a/doc/monado/Vive.md +++ b/doc/monado/Vive.md @@ -15,20 +15,20 @@ work very well. The general calibration procedure is as follows: 1. Setup the calibration target: - 1. Download [the one from kalibr](https://drive.google.com/file/d/1DqKWgePodCpAKJCd_Bz-hfiEQOSnn_k0/view) - 2. Open the pdf file and measure with a ruler the black square side (e.g., 24 inch 1080p screen, 34% zoom in Okular viewer, gives 3cm) - 3. Update aprlgrid_6x6.json "tagSize" property in meters (e.g., 3cm would be "0.03") + 1. Download [the one from kalibr](https://drive.google.com/file/d/1DqKWgePodCpAKJCd_Bz-hfiEQOSnn_k0/view) + 2. Open the pdf file and measure with a ruler the black square side (e.g., 24 inch 1080p screen, 34% zoom in Okular viewer, gives 3cm) + 3. Update aprlgrid_6x6.json "tagSize" property in meters (e.g., 3cm would be "0.03") 2. Record camera calibration sequence with the euroc recorder in Monado: - - See calib-cam3 example from TUM-VI https://vision.in.tum.de/data/datasets/visual-inertial-dataset - - Or this example from ORB-SLAM3: https://www.youtube.com/watch?v=R_K9-O4ool8 - - Note that for stereo calibration you want to cover as much as possible of - both cameras images but always trying to keep >90% of the calibration target - visible in both views. This can be a little tricky so practice it beforehand. + - See calib-cam3 example from TUM-VI https://vision.in.tum.de/data/datasets/visual-inertial-dataset + - Or this example from ORB-SLAM3: https://www.youtube.com/watch?v=R_K9-O4ool8 + - Note that for stereo calibration you want to cover as much as possible of + both cameras images but always trying to keep >90% of the calibration target + visible in both views. This can be a little tricky so practice it beforehand. 3. Record camera-imu calibration sequence: (faster motions) - - Similar recommendations as in the previous sequence but this time we want - faster motions that excite all IMU axes, see these examples: - - calib-imu1 from TUM-VI https://vision.in.tum.de/data/datasets/visual-inertial-dataset - - Or this from ORB-SLAM3 https://www.youtube.com/watch?v=4XkivVLw5k4 + - Similar recommendations as in the previous sequence but this time we want + faster motions that excite all IMU axes, see these examples: + - calib-imu1 from TUM-VI https://vision.in.tum.de/data/datasets/visual-inertial-dataset + - Or this from ORB-SLAM3 https://www.youtube.com/watch?v=4XkivVLw5k4 4. Now run camera calibration as explained [here](https://gitlab.freedesktop.org/mateosss/basalt/-/blob/xrtslam/doc/Calibration.md#camera-calibration) 5. Then camera+imu (no mocap) calibration as explained [here](https://gitlab.freedesktop.org/mateosss/basalt/-/blob/xrtslam/doc/Calibration.md#camera-imu-mocap-calibration) 6. If the datasets are not good, the calibration will likely not be very good as