search

biigle / maia

:m: BIIGLE module for the Machine Learning Assisted Image Annotation method
GNU General Public License v3.0
2 stars 3 forks source link

Optimize for scarce resources #4

Closed mzur closed 4 years ago

mzur commented 5 years ago

Optimize novelty detection and instance segmentation for systems with scarce resources (like the Jetson). This might be an interesting (but rather unrelated) topic for the planned MAIA/sync paper.

mzur commented 5 years ago

There is a working setup for BIIGLE with MAIA on the Jetson in the biigle/jetson gpus branch. Novelty detection works somewhat if the filesystem cache is flushed from RAM beforehand (sync; echo 3 > /proc/sys/vm/drop_caches). But instance segmentation (training) needs more memory. Inference might work but I could not test it.

This is also related to #32.

mzur commented 4 years ago

This is no longer relevant, as mobile computing platforms with adequate hardware are readily available on a ship (e.g.SHiPCC).