Replies: 3 comments
-
just to elaborate: benifit:
Additionally, by loading multiple results with gt results, we can create a compare panel that computes all the "error or difference" for each result against "gt". WIth the help of ploting view, we then can quickly visualize the error peaks and jump between them with timeline synced. |
Beta Was this translation helpful? Give feedback.
-
Thanks to @amacneil for mentioning this issue. Linking my comment here for reference: #109
|
Beta Was this translation helpful? Give feedback.
-
hope the feature can be implemented soon, it would be better to have "showdiff" feature (same like what webviz implemented), so that we could compare the results generated from two models. |
Beta Was this translation helpful? Give feedback.
-
Use case: An engineering team records data from their robot. Later, they rerun some of the recorded data against an updated codebase (new algorithms / models), and record the results. In some cases, they may replay the same recorded data against several new alternative versions of the code, and record the results.
Then they would like to compare the output from the original recording with the new recording, in the same studio instance. For example, overlaying the new behavior visualization on top of the old behavior.
This feature was partially supported in original Webviz code, but was removed due to being hard coded to Cruise conventions in many places and not flexible.
Workaround: It is possible to achieve this today by rewriting the newer bag using a topic prefix. I'm not sure if existing CLI tools can do so, but a very short python script using the
rosbag
library would achieve it.Beta Was this translation helpful? Give feedback.
All reactions