Integration of 3rd party AI / ML models #121
miquelmassot
started this conversation in
General
Replies: 1 comment
-
Likewise with pretrained model weights. If two different users want to use the same model (e.g. Detectron2) and have different weights (e.g. different applications) they should use the same plugin and just provide different URL or local paths to the PTH file or weights file. |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Take for example detectron2 (https://github.com/facebookresearch/detectron2). I would expect a plugin-like folder in scivision where wrappers for 3rd party models are implemented. If a user wants to use scivision to perform inference / classification of a large dataset (using your intake drivers/plugins too!) they should be able to instance a plugin and that should try to use the desired AI model. If that model cannot be found because a package is missing, there could be some help on how to install it, or refer the users to the main model repo (e.g. detectron2 github page).
Beta Was this translation helpful? Give feedback.
All reactions