labkit-ui
labkit-ui copied to clipboard
Batch speed optimization
The way Labkit uses resources to batch process images can still be improved:
- process small images in parallel
- set optimal TaskExecutors for CPU and GPU processing
- match the number of parallel GPU accesses to the available GPU memory
- allow to increase the block size
The PR adds a BatchBenchmark that allows to measure Labkits performance when operating on multiple images, and also has some WIP code for setting TaskExecutors.