Is your feature request related to a problem? Please describe.
The VISTA3D is a very brilliant model for everything segmentation in a medical image. However it suffers a slowdown issue when running on a GPU platform, which takes minutes to run a single inference. Therefore, a lower latency version of the VISTA3D model should be a user-friendly model and gain more usage.
Describe the solution you'd like
Update the VISTA3D bundle parameters to achieve a better latency.
Update the VISTA3D network to solve some latency bottlenecks.
Try to accelerate the bundle with acceleration tools like TensorRT.
Additional context
Some analyzation has already done in this PR.
The corresponding MONAI core code change is included here
Is your feature request related to a problem? Please describe. The VISTA3D is a very brilliant model for everything segmentation in a medical image. However it suffers a slowdown issue when running on a GPU platform, which takes minutes to run a single inference. Therefore, a lower latency version of the VISTA3D model should be a user-friendly model and gain more usage.
Describe the solution you'd like
Additional context Some analyzation has already done in this PR. The corresponding MONAI core code change is included here