Train using original image size on two GPUs

HPI-DeepLearning / SegMed

An online platform based on adversarial training for pixel wise classification on medical images.

7 stars 4 forks source link

Closed Highlife1911 closed 7 years ago

Highlife1911 commented 7 years ago

train with batch size of 32 consumes to much memory. Training with a batch size of 24. Each batch takes ~12 Seconds.