search

kevinlin311tw / caffe-cvprw15

:heart::coffee: Deep Learning of Binary Hash Codes for Fast Image Retrieval (CVPRW15)
Other
526 stars 218 forks source link

trainning myself datasets #35

Closed IvyYZ closed 6 years ago

IvyYZ commented 6 years ago

My data set contains 10,200 images, and the next four images are classified as a single category . The number of my category is 2500. With the deephash network training, loss dropped from 7.1 to 3.79. Then iterate 50000 times, loss keeps stable. I want to know if this is a normal phenomenon. How can we let loss continue to decline?

kevinlin311tw commented 6 years ago

It depends on datasets. If you switch to other dataset, it is very possible that you have to customize the learning rates, and the iteration numbers (or epochs).

koalaofpoint commented 6 years ago

can you share the related dataset that you had got with me?,please @aaaaxiaodui

IvyYZ commented 6 years ago

I have some common datasets, I can send to you if you need. @koalaofpoint