Hi,
I am using pyvips (Libvips) to create tiles (small jpegs) from larger pyramid tifs (.svs-files).
It takes very long time to execute in Datalore. The below code block ran for more than 1 hour before I interupted the execution. By then the notebook/Datalore had created 385 jpegs (of 5450).
For comparison:
- Using Pycharm on my laptop produced 5450 jpegs in 79 sec.
- Using cloud notebook service from another service provider produced 5450 jpegs in 6 min 47 sec. (using their least expensive CPU)
The svs-file (47854.svs) is 512 MB
Code used in Datalore notebook:
import pyvips
import time
exec_start = time.perf_counter()
image = pyvips.Image.new_from_file('WSI/Liver/47854.svs')
image.dzsave('47854_TS_512', overlap=0, tile_size=512, depth="one")
exec_end = time.perf_counter()
print(f"Code execution in {exec_end - exec_start:0.4f} seconds")
Code used on laptop:
import pyvips
import time
exec_start = time.perf_counter()
image = pyvips.Image.new_from_file('47854.svs')
image.dzsave('C:/Localdata/47854_TS_512', overlap=0, tile_size=512, depth="one")
exec_end = time.perf_counter()
print(f"Code execution in {exec_end - exec_start:0.4f} seconds")