projects
/
pysvrt.git
/ commitdiff
commit
grep
author
committer
pickaxe
?
search:
re
summary
|
shortlog
|
log
|
commit
| commitdiff |
tree
raw
|
patch
|
inline
| side by side (from parent 1:
77153ad
)
Update.
author
Francois Fleuret
<francois@fleuret.org>
Mon, 19 Jun 2017 12:41:57 +0000
(14:41 +0200)
committer
Francois Fleuret
<francois@fleuret.org>
Mon, 19 Jun 2017 12:41:57 +0000
(14:41 +0200)
README.md
patch
|
blob
|
history
diff --git
a/README.md
b/README.md
index
d0ec215
..
f4e07af
100644
(file)
--- a/
README.md
+++ b/
README.md
@@
-36,6
+36,9
@@
and the following method to return one batch
(torch.FloatTensor, torch.LongTensor) get_batch(b)
```
(torch.FloatTensor, torch.LongTensor) get_batch(b)
```
+as a pair composed of a 4d 'input' Tensor (i.e. single channel 128x128
+images), and a 1d 'target' Tensor (i.e. Boolean labels).
+
## Low-level functions ##
The main function for genering vignettes is
## Low-level functions ##
The main function for genering vignettes is
@@
-80,15
+83,7
@@
See vignette_set.py for a class CompressedVignetteSet using it.
# Testing convolution networks #
# Testing convolution networks #
-The file
-
-```
-cnn-svrt.py
-```
-
-provides the implementation of two deep networks, and use the
-compressed vignette code to allow the training with several millions
-vignettes on a PC with 16Gb and a GPU with 8Gb.
-
-The networks were designed by Afroze Baqapuri during an internship at
-Idiap.
+The file `cnn-svrt.py` provides the implementation of two deep
+networks designed by Afroze Baqapuri during an internship at Idiap,
+and allows to train them with several millions vignettes on a PC with
+16Gb and a GPU with 8Gb.