-
Notifications
You must be signed in to change notification settings - Fork 3
CLIP encoder #69
Copy link
Copy link
Open
Labels
P1Priority 1Priority 1contributor neededNeeds a contributor to take over the issue/featureNeeds a contributor to take over the issue/featuredependenciesPull requests that update a dependency filePull requests that update a dependency file
Description
Metadata
Metadata
Assignees
Labels
P1Priority 1Priority 1contributor neededNeeds a contributor to take over the issue/featureNeeds a contributor to take over the issue/featuredependenciesPull requests that update a dependency filePull requests that update a dependency file
Is your feature request related to a problem? Please describe.
Besides the typical encoder used here like
VLADorFisher Vector, I would like to implement a one-shot encoder which uses theCLIPbackbone. It needs no processing, just one-shot and the embedding is there :)Describe the solution you'd like
Implement a
CLIPEncoder(ImageEncoderBase)that extracts the CLIP feature as a vector. Probably the model backbone would come fromtransformerslibrary.Describe alternatives you've considered
A clear and concise description of any alternative solutions or features you've considered.
Additional context
Add any other context or screenshots about the feature request here.