Open jamesMo84 opened 1 year ago
Hi @jamesMo84, There's currently not a way baked into the notebook to rely on a local installation of Mesmer/Cellpose, mostly because it would significantly increase the complexity of the dependencies. However, if you already have a local segmentation solution set up to process large images, you'd just need to name them appropriately so that the rest of the notebook could read them in and process them. You can then skip the cell that formats and uploads for the deepcell server, and do the process yourself.
There's also a dedicated repo with a containerized version: https://github.com/vanvalenlab/deepcell-applications
Hi, @jamesMo84 Have you solved your problem? I encounter the same problem,too. I will appreciate if you can share some experience. Thanks a lot !
Hi, Great package, I've started applying the pipeline to TMA cores but am running into the issue with Deepcell Kiosk max tile size of 2048x2048. Is there scope to specify a local segmentation server (DeepCell or Cellpose) into ark.utils.deepcell_service_utils to overcome this tile size limit? how does one find and call the host name from alocal Deepcell server installed as per https://github.com/vanvalenlab/deepcell-tf