 nagadomi
|
a72af8cfef
Use 'image' module because the deadlock problem has fixed
|
8 年之前 |
|
nagadomi
|
4998078d2a
use threads.safe to avoid deadlock problem
|
8 年之前 |
|
nagadomi
|
1e80e45a03
add experimental models
|
8 年之前 |
|
nagadomi
|
558527e268
merge
|
8 年之前 |
|
nagadomi
|
50fd999c38
Add -gpu option
|
8 年之前 |
|
nagadomi
|
16731fb634
Add support for scale4 in benchmark
|
8 年之前 |
|
nagadomi
|
ef8bb4b0b9
update Benchmark.md
|
8 年之前 |
|
nagadomi
|
83fc4684d8
update Benchmark.md
|
8 年之前 |
|
nagadomi
|
6f437e4014
Update benchmark
|
8 年之前 |
|
nagadomi
|
8d16fbaa9a
Update benchmark
|
8 年之前 |
|
nagadomi
|
64952bad1f
Add upconv_7l models. only 2x.
|
8 年之前 |
|
nagadomi
|
66bff68ef1
split first
|
8 年之前 |
|
nagadomi
|
a604aa3b7a
Add some experimental model
|
8 年之前 |
|
nagadomi
|
e71ff1fc6a
Fix a bug in random blur
|
8 年之前 |
|
nagadomi
|
72862f182e
benchmark: add border option; use %.3f;
|
8 年之前 |
|
nagadomi
|
3e77378983
Add SSIMCriterion (supports only forward())
|
8 年之前 |
|
nagadomi
|
d4bab939ab
Fix basename in diff method
|
8 年之前 |
|
nagadomi
|
1ce5a8d038
more detailed log in benchmark
|
8 年之前 |
|
nagadomi
|
51e189b6c0
Add count of won in benchmark
|
8 年之前 |
|
nagadomi
|
d7ab10581c
Use clamp
|
8 年之前 |
|
nagadomi
|
f1a45b19c5
test
|
8 年之前 |
|
nagadomi
|
06e08253e4
Fix some multhread bug
|
8 年之前 |
|
nagadomi
|
0517605d0d
Remove flip() from preprocess() because the data will flip() at flip_augmentation()
|
8 年之前 |
|
nagadomi
|
f65132dadb
Use conv2d instead of nn.SpatialConvolutionMM
|
8 年之前 |
|
nagadomi
|
a14e6acec3
mutex
|
8 年之前 |
|
nagadomi
|
5a3d012f4e
Add random blur method for data augmentation
|
8 年之前 |
|
nagadomi
|
cabeeed2a7
Add Clip(0,1) to last layer
|
8 年之前 |
|
nagadomi
|
f11108cba1
Add comment about RNG seed
|
8 年之前 |
|
nagadomi
|
451ee1407f
Stop calculate the instance loss when oracle_rate=0
|
8 年之前 |
|
nagadomi
|
33e6bc888e
Fix the number of threads
|
8 年之前 |
