Open miguelriemoliveira opened 1 month ago
clear && \
export CSF="lambda x: int(x) not in []" \
&& rosrun atom_calibration calibrate \
-json $ATOM_DATASETS/zau/inesc_day2_5_full/dataset_corrected_with_odometry_and_depth.json \
-uic -v \
-csf "$CSF" \
-jsf "lambda x: False" \
-jpsf "lambda x: x in ['origin_yaw']"
Results using the depth camera initial estimate corrected dataset
+------------+----------------+--------------------+---------------------+---------------------+
| Collection | lidar_body [m] | rgb_body_left [px] | rgb_body_right [px] | rgbd_hand_depth [m] |
+------------+----------------+--------------------+---------------------+---------------------+
| Averages | 0.0092 | 11.1405 | 5.7684 | 0.0073 |
+------------+----------------+--------------------+---------------------+---------------------+
I guess this means we have good results.
The color sensor in the rgbd_hand camera must also be corrected. I will try to do this also.
Early results with the calibration of the zau rgbd hand camera showed that the camera was badly positioned (#970).
I then worked on #971 to be able to correct the initial guess so that it is more or less plausible, as can be seen at the end of #970.
Now the idea is to calibrate the system with depth cameras to see if it works better (before it was over 100 pixels of error).