8000 Commits · eveningdong/DeepLabV3-Tensorflow · GitHub

Commits

Commits on Feb 12, 2018

Update new val result
NanqingD
committed

Commits on Feb 11, 2018

fix restore variable bug
NanqingD
committed
fix val_summary
NanqingD
committed
remove duplicate
NanqingD
committed
restore poly learning policy
NanqingD
committed
reorder tf summary for sync display
NanqingD
committed

Commits on Feb 7, 2018

update setup.sh
NanqingD
committed

Commits on Feb 6, 2018

fix all branches concatenation
NanqingD
committed
implement imagenet pretrained
NanqingD
committed
delete branch tf_1.2 since code already compatible
NanqingD
committed
fix validation data
NanqingD
committed

Commits on Feb 5, 2018

fix multi-grid
NanqingD
committed

Commits on Jan 11, 2018

new val mIOU
NanqingD
committed

Commits on Jan 9, 2018

missed last validation sample
NanqingD
committed
better structure representation
NanqingD
committed

Commits on Dec 25, 2017

fix typo
NanqingD
committed

Commits on Dec 24, 2017

fix typo
NanqingD
committed
add training tricks
NanqingD
committed
finish VOC2012 training
NanqingD
committed

Commits on Dec 22, 2017

small inconsistency, no influence to performance
NanqingD
committed

Commits on Dec 21, 2017

upsample logits
NanqingD
committed
fix architecture
NanqingD
committed
Switch to train_val_voc12.py
NanqingD
committed
fix tf.train.string_input_producer() bug
NanqingD
committed
update README
NanqingD
committed
train and val in one script
NanqingD
committed

Commits on Dec 20, 2017

fix summary bug
NanqingD
committed
fix softmax layer bug and add train_val
NanqingD
committed
fix val_voc12.py bug
NanqingD
committed

Commits on Dec 18, 2017

fix args bug
NanqingD
committed
fix bug for block4
NanqingD
committed
better implementation of weight decay
NanqingD
committed

Commits on Dec 17, 2017

update README
NanqingD
committed
make multi-gpu version loss converge
NanqingD
committed

Commits on Dec 16, 2017

fix batch_norm update bug
NanqingD
committed

0