search

QVPR / Patch-NetVLAD

Code for the CVPR2021 paper "Patch-NetVLAD: Multi-Scale Fusion of Locally-Global Descriptors for Place Recognition"
MIT License
517 stars 72 forks source link

Training Time #23

Closed michaelschleiss closed 2 years ago

michaelschleiss commented 2 years ago

I was wondering what hardware you used for training and how much time it took to train the model on the full dataset.

StephenHausler commented 2 years ago

We used a single Nvidia T4 GPU on a high performance computing system. Iirc, it takes about 60 hours per epoch. We stopped training after 22 epochs based on the progression of the validation recall, which in total took ~55 days of continuous training.