nagadomi
2f24c079fc
Add pretrained cunet/upcunet models for art
2018-11-14 17:33:05 +09:00
nagadomi
d5c2277e0e
misc
2018-11-14 17:21:34 +09:00
nagadomi
581c27a2cb
Change the default crop_size from 128 to 256
2018-11-09 17:47:17 +09:00
nagadomi
6efd7f890e
Add cunet for 1x; Remove unused code
2018-10-29 04:07:05 +09:00
nagadomi
b28b6172ca
clean; Add upcunet_v3
2018-10-28 16:03:52 +09:00
nagadomi
929c7f85a9
Support Linear module
2018-06-05 03:57:31 +09:00
nagadomi
3c46906cb7
Merge branch 'dev'
2018-04-28 11:08:05 +09:00
nagadomi
e5e73c0ba7
Fix zero division error in jaccard
2018-04-28 11:04:53 +09:00
nagadomi
0b1a13d9c0
Add to json formt model; Fix json format;
2018-04-28 11:00:13 +09:00
nagadomi
f65e723ef9
Add a tool for visualizing layer outputs
2018-01-26 08:20:26 +09:00
nagadomi
fa6ee00624
merge
2018-01-26 08:09:40 +09:00
nagadomi
4ec0b4530d
merge
2018-01-26 08:08:52 +09:00
nagadomi
eb3926b61c
Fix to work on Lua52
...
web.lua does not work on Lua52 because turbo does not support Lua52 FFI.
2017-03-02 14:47:49 +09:00
nagadomi
d8b7df4505
add support for jaccard in benchmark
2017-02-21 21:39:56 +09:00
nagadomi
cf2493a1a4
Add support for resnet_14l in tools/export.lua; Add resnet_14l.prototxt
2017-01-26 05:43:55 +09:00
nagadomi
02cf265d48
Reduce memory usage in benchmark
2017-01-04 19:53:17 +09:00
nagadomi
16731fb634
Add support for scale4 in benchmark
2016-10-10 02:20:57 +09:00
nagadomi
64952bad1f
Add upconv_7l models. only 2x.
2016-10-08 19:36:09 +09:00
nagadomi
72862f182e
benchmark: add border option; use %.3f;
2016-10-05 15:57:19 +09:00
nagadomi
d4bab939ab
Fix basename in diff method
2016-09-27 19:05:53 +09:00
nagadomi
1ce5a8d038
more detailed log in benchmark
2016-09-27 18:23:10 +09:00
nagadomi
51e189b6c0
Add count of won in benchmark
2016-09-27 11:43:39 +09:00
nagadomi
3a70d09921
create new directory when the specified directory is not found
2016-08-30 17:13:25 +09:00
nagadomi
83a8cbe3bd
Add diff method to tools/benchmark
2016-08-26 06:34:23 +09:00
nagadomi
8d77d6e2e9
Remove duplicated option
2016-08-23 18:57:50 +09:00
nagadomi
8af1cb7b94
Remove debug print
2016-08-23 18:57:11 +09:00
nagadomi
bfb67e61f4
Add support for the test filelist in benchmark
2016-08-21 06:54:23 +09:00
nagadomi
f72d756172
Add helper script for user method
2016-08-21 04:07:29 +09:00
nagadomi
7040dff6f8
Fix noise0 model (backend: cudnn -> cunn)
2016-08-07 01:41:42 +09:00
nagadomi
c759f040bc
Add support for individual channel measure in benchmark
2016-08-01 04:37:39 +09:00
nagadomi
3add726da3
Fix a bug that the time benchmark does not works when method=user and scale=1
2016-07-22 19:56:33 +09:00
nagadomi
962bdcf300
Add support for user method in benchmark
2016-07-22 02:15:00 +09:00
nagadomi
5786099bc8
Fix a issue of time comparison in benchmark when using cuDNN
2016-07-21 18:04:01 +09:00
nagadomi
9563d84302
No bias at FullConvolution
2016-07-12 11:15:12 +09:00
nagadomi
83e5d885f5
Update scripts
2016-07-09 13:41:51 +09:00
nagadomi
0300c714eb
raise load error
2016-07-09 13:20:30 +09:00
nagadomi
4379d4d0fd
Fix a inf in benchmark.lua
2016-06-19 22:28:42 +09:00
nagadomi
b89833daa3
Add -yuv420 option to tools/benchmark.lua
2016-06-18 20:25:47 +09:00
nagadomi
6cca2f8488
Add -force_cudnn option to tools/benchmark.lua
2016-06-12 22:03:19 +09:00
nagadomi
67d36a1220
benchmark time
2016-06-12 21:56:49 +09:00
nagadomi
83188c5ab7
Add support for method=noise_scale to tools/benchmark.lua
2016-06-12 05:12:31 +09:00
nagadomi
c16d0a07a2
Add -crop_size and -batch_size option to tools/benchmark.lua. Fix a bug in tta mode.
2016-06-10 09:20:43 +09:00
nagadomi
b8ff8c6787
Remove -gamma_correction option
2016-06-10 07:37:39 +09:00
nagadomi
5e222a3981
Add -tta and -resize_blur option to benchmark
2016-06-02 10:15:54 +09:00
nagadomi
a210090033
Convert model files; Add new pretrained model
...
- Add new pretrained model to ./models/upconv_7
- Move old models to ./models/vgg_7
- Use nn.LeakyReLU instead of w2nn.LeakyReLU
- Add useful attribute to .json
New JSON attribute:
The first layer has `model_config` attribute.
It contains:
model_arch: architecture name of model. see `lib/srcnn.lua`
scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor.
channels: input/output channels. if channels == 3, model is RGB model.
offset: pixel size that is to be removed from output.
for example:
(scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7)
(scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12)
And each layer has `class_name` attribute.
2016-05-15 03:04:08 +09:00
nagadomi
7af5c9443d
Add model option and 12 layers net
2016-04-23 09:19:03 +09:00
nagadomi
8e9e4433d3
Add thread option to benchmark
2016-04-15 09:15:19 +09:00
nagadomi
d31fbe9bb1
Improve benchmark script
2016-04-12 03:48:44 +09:00
nagadomi
7710a30225
Add useful option to benchmark
2016-04-11 23:19:47 +09:00
nagadomi
d9474702f6
Fix NaN in benchmark script
2016-04-11 21:48:00 +09:00
