search

sacmehta / ESPNet

ESPNet: Efficient Spatial Pyramid of Dilated Convolutions for Semantic Segmentation
https://sacmehta.github.io/ESPNet/
MIT License
541 stars 112 forks source link

Training time #6

Closed KengChiLiu closed 6 years ago

KengChiLiu commented 6 years ago

Hi @sacmehta , How long did it take you to train the model ? Training with different scales and 300 epoch (default setting) seems to take a long long time......

sacmehta commented 6 years ago

Training should be done in 2-3 days on TItanX PASCAL Single GPU.

Note that the Cityscape dataset has high resolution images (2048 x 1024). These images are large in comparison to other standard datasets such as the ImageNet and the PASCAL VOC.