add all-at-once edition of FCN-8s for PASCAL VOC
This net is fine-tuned from VGG-16 all-at-once instead of in stages. All-at-once learning is faster and less tedious but gives an ever-so-slightly less accurate model.
Showing
voc-fcn8s-atonce/caffemodel-url
0 → 100644
voc-fcn8s-atonce/net.py
0 → 100644
voc-fcn8s-atonce/solve.py
0 → 100644
voc-fcn8s-atonce/solver.prototxt
0 → 100644
voc-fcn8s-atonce/train.prototxt
0 → 100644
voc-fcn8s-atonce/val.prototxt
0 → 100644
想要评论请 注册 或 登录