search

NVIDIA / caffe

Caffe: a fast open framework for deep learning.
http://caffe.berkeleyvision.org/
Other
672 stars 263 forks source link

Questions about prototxts and lmdb for SSD training #581

Closed kadirbeytorun closed 4 years ago

kadirbeytorun commented 4 years ago

Hello, I am trying to grasp the details about each prototxt files and lmdb files. 1- Is train.prototxt only for training or does it include validation? Because the lmdb you called in it is named trainval.lmdb, but there is only one data layer, and its phase is TRAIN 2-Is test prototxt actually for validation or testing? its phase is TEST, but lmdb file is named test.lmdb, so its quite confusing. 3- Is deploy prototxt actually testing network? whats the difference between this and test prototxt? Regards, Kadir

drnikolaev commented 4 years ago

Hi @kadirbeytorun please have a look here: http://caffe.berkeleyvision.org/