I think if we want to make use of the decoder model @sunu produced with https://github.com/vietanhdev/samexporter then we need this logic in javascript. I lean towards DS-Annotate supporting this even if it takes some more figuring out @willemarcel @Rub21
For JOSM, I think it sounds like we would need a separate endpoint deployed so we don't need to code this in Java. Is that right @Rub21 @willemarcel ? Is this worth doing @geohacker given that we can use DS-Annotate for the segment-anything functionality?
I can do this, it would mean that I'd edit the torchserve handler to accept geotiffs or regular images and handle them separately in the same endpoint. I don't think adding multiple non-model endpoints are supported, but I asked here: https://github.com/pytorch/serve/issues/2352
rbavery
commented
1 year ago
Where should this logic live?
I think if we want to make use of the decoder model @sunu produced with https://github.com/vietanhdev/samexporter then we need this logic in javascript. I lean towards DS-Annotate supporting this even if it takes some more figuring out @willemarcel @Rub21
For JOSM, I think it sounds like we would need a separate endpoint deployed so we don't need to code this in Java. Is that right @Rub21 @willemarcel ? Is this worth doing @geohacker given that we can use DS-Annotate for the segment-anything functionality?
I can do this, it would mean that I'd edit the torchserve handler to accept geotiffs or regular images and handle them separately in the same endpoint. I don't think adding multiple non-model endpoints are supported, but I asked here: https://github.com/pytorch/serve/issues/2352