search

boxuLibrary / drt-vio-init

The official repository of our CVPR2023 paper "A Rotation-Translation-Decoupled Solution for Robust and Efficient Visual-Inertial Initialization".
GNU General Public License v3.0
159 stars 20 forks source link

How to transplant drt init into VINS_MONO，replace the original vins initialization? #15

Closed wuerhan closed 8 months ago

boilcy commented 6 months ago

check my repo if you still need it. https://github.com/boilcy/VINS-Mono.git

INDEX108 commented 6 months ago

check my repo if you still need it. https://github.com/boilcy/VINS-Mono.git

Hi,I've tried this repo.But there's something wrong with it. NODES / feature_tracker (feature_tracker/feature_tracker) pose_graph (pose_graph/pose_graph) vins_estimator (vins_estimator/vins_estimator)

ROS_MASTER_URI=http://localhost:11311

process[feature_tracker-1]: started with pid [61405] process[vins_estimator-2]: started with pid [61406] process[pose_graph-3]: started with pid [61407] [ INFO] [1715834716.083318903]: init begins [ INFO] [1715834716.090406815]: Loaded config_file: /home/index108/VINS/DRT_VINS/src/VINS-Mono/feature_tracker/../config/euroc/euroc_config.yaml result path /home/index108/output//vins_result_no_loop.csv [ INFO] [1715834716.091029590]: ROW: 480.000000 COL: 752.000000 [ WARN] [1715834716.091046352]: fix extrinsic param [ INFO] [1715834716.091103482]: Extrinsic_R : 0.0148655 -0.999881 0.0041403 0.999557 0.0149672 0.0257155 -0.0257744 0.00375619 0.999661 [ INFO] [1715834716.091119755]: Extrinsic_T : -0.0216401 -0.064677 0.00981073 [ INFO] [1715834716.091131139]: Synchronized sensors, fix time offset: 0 [ INFO] [1715834716.091142733]: INIT: DRT Loosely [ WARN] [1715834716.091151672]: waiting for image and imu... vocabulary_file/home/index108/VINS/DRT_VINS/src/VINS-Mono/pose_graph/../support_files/brief_k10L6.bin loop start load vocabulary BRIEF_PATTERN_FILE/home/index108/VINS/DRT_VINS/src/VINS-Mono/pose_graph/../support_files/brief_pattern.yml no previous pose graph [ WARN] [1715834724.343304612]: gyroscope bias not converge -0.00245898 0.0153236 0.0799477 [ WARN] [1715834724.543914328]: gyroscope bias not converge -0.00159427 0.0169537 0.0830316 [ WARN] [1715834724.735796609]: gyroscope bias not converge -0.00197735 0.0266758 0.083536 [ WARN] [1715834724.947585679]: gyroscope bias not converge -0.00333356 0.0388138 0.0807412 [ WARN] [1715834725.036877546]: gyroscope bias not converge -0.00271899 0.0404334 0.0827656 [ WARN] [1715834725.242807072]: gyroscope bias not converge -0.0032043 0.0344875 0.0805693 [ WARN] [1715834725.437371956]: gyroscope bias not converge -0.00428633 0.0301807 0.077629 [ WARN] [1715834725.551780702]: gyroscope bias not converge -0.00462144 0.0296536 0.0767685 [ WARN] [1715834725.739333941]: gyroscope bias not converge -0.00649432 0.0287829 0.0705355 [ WARN] [1715834725.932372374]: gyroscope bias not converge -0.00660054 0.00589244 0.0695617 [ WARN] [1715834726.035257309]: gyroscope bias not converge -0.0064082 0.00627914 0.07041 [ WARN] [1715834726.240511960]: gyroscope bias not converge -0.00528582 0.0145278 0.0759611 [ WARN] [1715834726.439162558]: gyroscope bias initial calibration -0.00358731 0.011268 0.0821361 [vins_estimator-2] process has died [pid 61406, exit code -11, cmd /home/index108/VINS/DRT_VINS/devel/lib/vins_estimator/vins_estimator __name:=vins_estimator __log:=/home/index108/.ros/log/b5f4be1a-133c-11ef-bf30-6df1c94c400a/vins_estimator-2.log]. log file: /home/index108/.ros/log/b5f4be1a-133c-11ef-bf30-6df1c94c400a/vins_estimator-2*.log

boilcy commented 6 months ago

image it works on EuRoC MH_01_easy data. It is possible that the initialization error situation was handled incorrectly. could you show me the logfile?

INDEX108 commented 6 months ago

image it works on EuRoC MH_01_easy data. It is possible that the initialization error situation was handled incorrectly. could you show me the logfile?

Thank you very much, I have resolved it. By the way, is the accuracy of this repo higher than Vinsmono? How did you evaluate the initialization effect of the DRT method?

boilcy commented 6 months ago

EuRoc dataset has Ground-Truth. .../datasets/EuRoC/MH_03_medium/mav0/state_groundtruth_estimate0 if it's in folder format, topic /leica/position 3099 msgs : geometry_msgs/PointStamped if it's rosbag format. I haven't evaluated it yet.

Yingsheng123 commented 5 months ago

hello，the Initialization effect is not well ，do you know the reason

Yingsheng123 commented 5 months ago

EuRoc dataset has Ground-Truth. .../datasets/EuRoC/MH_03_medium/mav0/state_groundtruth_estimate0 if it's in folder format, topic /leica/position 3099 msgs : geometry_msgs/PointStamped if it's rosbag format. I haven't evaluated it yet.

have you evaluated it?

jerryjifjsdbf commented 4 months ago

EuRoc dataset has Ground-Truth. .../datasets/EuRoC/MH_03_medium/mav0/state_groundtruth_estimate0 if it's in folder format, topic /leica/position 3099 msgs : geometry_msgs/PointStamped if it's rosbag format. I haven't evaluated it yet.

请问你已经把这个方法集成到vins-mono中吗，定位精度的效果咋样

Yingsheng123 commented 4 months ago

EuRoc dataset has Ground-Truth. .../datasets/EuRoC/MH_03_medium/mav0/state_groundtruth_estimate0 if it's in folder format, topic /leica/position 3099 msgs : geometry_msgs/PointStamped if it's rosbag format. I haven't evaluated it yet.

请问你已经把这个方法集成到vins-mono中吗，定位精度的效果咋样

我使用重力计算方法和vins-mono一样，其余部分使用了这个方法，效果不是很好。发现ros中初始化部分的效果不如直接运行这个源码，我也不太清楚问题在哪里

jerryjifjsdbf commented 4 months ago

EuRoc dataset has Ground-Truth. .../datasets/EuRoC/MH_03_medium/mav0/state_groundtruth_estimate0 if it's in folder format, topic /leica/position 3099 msgs : geometry_msgs/PointStamped if it's rosbag format. I haven't evaluated it yet.

请问你已经把这个方法集成到vins-mono中吗，定位精度的效果咋样

我使用重力计算方法和vins-mono一样，其余部分使用了这个方法，效果不是很好。发现ros中初始化部分的效果不如直接运行这个源码，我也不太清楚问题在哪里 请问你现在解决了吗

Yingsheng123 commented 4 months ago

EuRoc dataset has Ground-Truth. .../datasets/EuRoC/MH_03_medium/mav0/state_groundtruth_estimate0 if it's in folder format, topic /leica/position 3099 msgs : geometry_msgs/PointStamped if it's rosbag format. I haven't evaluated it yet.

请问你已经把这个方法集成到vins-mono中吗，定位精度的效果咋样

我使用重力计算方法和vins-mono一样，其余部分使用了这个方法，效果不是很好。发现ros中初始化部分的效果不如直接运行这个源码，我也不太清楚问题在哪里 请问你现在解决了吗

还没有，不过通过实验发现，这个方法和smf在存储要使用特征点有些许差别，需要制定一些策略