pc6tevj3f
/
style_transfer
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Delete 'styleTransfer_projectReport.txt'

Browse Source
master
pc6tevj3f 2 years ago
parent 18e60294e7
commit 4d760387b2
1 changed files with 0 additions and 303 deletions
Show Stats Download Patch File Download Diff File

303
styleTransfer_projectReport.txt
Unescape View File

@ -1,303 +0,0 @@
<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ǩ<EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ŀ<EFBFBD><EFBFBD><EFBFBD><EFBFBD>
10205101496 <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD>
ժҪ<EFBFBD><EFBFBD>
<20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ͼ<EFBFBD><CDBC><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ȡ<EFBFBD><C8A1><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ľ<EFBFBD>ϲ<EFBFBD><CFB2><EFBFBD><EFBFBD>Ǽ<EFBFBD><C7BC><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ĵ<EFBFBD><C4B4>£<EFBFBD><C2A3><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ա<EFBFBD>ĸ߶ȹ<DFB6>ע<EFBFBD><D7A2><EFBFBD>ʶ<EFBFBD><CAB6><EFBFBD>Щ<EFBFBD><D0A9><EFBFBD>һ<EFBFBD><D2BB><EFBFBD>pytorch<63>Է<EFBFBD><D4B7>Ǩ<EFBFBD>ƽ<EFBFBD><C6BD><EFBFBD><EFBFBD><EFBFBD>һ<EFBFBD><D2BB><EFBFBD>̶ȵ<CCB6>ѧϰ<D1A7><CFB0><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ھ<EFBFBD><DABE><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ʵ<EFBFBD><CAB5>ͼ<EFBFBD><CDBC><EFBFBD><EFBFBD>Ǩ<EFBFBD>ƣ<EFBFBD>Style Transfer<65><72><EFBFBD><EFBFBD>ͨ<EFBFBD><CDA8><EFBFBD>Կγ<D4BF><CEB3><EFBFBD><EFBFBD>ĵ<EFBFBD><C4B5>Ķ<EFBFBD><C4B6><EFBFBD><EFBFBD><EFBFBD>CSDN<44>ϲ鿴һЩ<D2BB><D0A9>صIJ<D8B5><C4B2>ͣ<EFBFBD><CDA3>Լ<EFBFBD>д<EFBFBD><D0B4>һ<EFBFBD><D2BB><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ʵ<EFBFBD><CAB5><EFBFBD>˷<EFBFBD><CBB7>Ǩ<EFBFBD>ƹ<EFBFBD><C6B9>ܡ<EFBFBD>
<EFBFBD><EFBFBD><EFBFBD>ģ<EFBFBD>
1 <20>˽<EFBFBD><CBBD><EFBFBD>Ǩ<EFBFBD><C7A8>
<EFBFBD><EFBFBD>ʵͼ<EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ǩ<EFBFBD><EFBFBD><EFBFBD>㷨<EFBFBD><EFBFBD>ʵ<EFBFBD><EFBFBD><EFBFBD>߼<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ӣ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ѡȡһ<EFBFBD><EFBFBD>ͼ<EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ϊ<EFBFBD><EFBFBD>׼ͼ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ͼ<EFBFBD>񣩣<EFBFBD>Ȼ<EFBFBD><EFBFBD>ѡȡ<EFBFBD><EFBFBD>һ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>߶ั<EFBFBD><EFBFBD>Ϊ<EFBFBD><EFBFBD><EFBFBD><EFBFBD>ϣ<EFBFBD><EFBFBD><EFBFBD><EFBFBD>ȡ<EFBFBD><EFBFBD>Ӧ<EFBFBD><EFBFBD><EFBFBD><EFBFBD>ͼ<EFBFBD>񣨷<EFBFBD><EFBFBD>ͼ<EFBFBD>񣩣<EFBFBD><EFBFBD><EFBFBD><EFBFBD>ڱ<EFBFBD>֤<EFBFBD><EFBFBD><EFBFBD><EFBFBD>ͼ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ե<EFBFBD>ǰ<EFBFBD><EFBFBD><EFBFBD>£<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ͼ<EFBFBD><EFBFBD>ķ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ͼ<EFBFBD><EFBFBD><EFBFBD>С<EFBFBD>
ͼ<EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ǩ<EFBFBD><EFBFBD>ʵ<EFBFBD>ֵ<EFBFBD><EFBFBD>ѵ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ч<EFBFBD><EFBFBD>ȡһ<EFBFBD><EFBFBD>ͼ<EFBFBD><EFBFBD>ķ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ͨ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ľ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ȡͼ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ҫ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ȼ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ȡ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>
<EFBFBD><EFBFBD>ʵ<EFBFBD><EFBFBD>ͼ<EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ǩ<EFBFBD>Ƴɹ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ڲ<EFBFBD>ͬ<EFBFBD><EFBFBD><EFBFBD>˶<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>б<EFBFBD>׼Ҳ<EFBFBD><EFBFBD><EFBFBD>ںܴ<EFBFBD>IJ<EFBFBD><EFBFBD>죬<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ѧ<EFBFBD><EFBFBD>Ҳ<EFBFBD><EFBFBD>û<EFBFBD>ж<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ͼ<EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ǩ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ϸ<EFBFBD>Ķ<EFBFBD><EFBFBD>塣ͼ<EFBFBD><EFBFBD>ķ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>˷ḻ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ݣ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ͼ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ɫ<EFBFBD><EFBFBD>ͼ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ͼ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ͼ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ҫ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ں<EFBFBD><EFBFBD>壬<EFBFBD>ȵȡ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ͨ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ǿ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ͼ<EFBFBD><EFBFBD><EFBFBD><EFBFBD>ijЩ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ͽ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ƣ<EFBFBD><EFBFBD>ͻ<EFBFBD><EFBFBD><EFBFBD>Ϊ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ͬһ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ϵ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ƕ<EFBFBD><EFBFBD><EFBFBD>רҵ<EFBFBD><EFBFBD>ʿ<EFBFBD><EFBFBD><EFBFBD>ԣ<EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ǹ<EFBFBD><EFBFBD><EFBFBD>עͼ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>εľ<EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ƿ<EFBFBD><EFBFBD><EFBFBD>ͬ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ͼ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ƿ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ǩ<EFBFBD><EFBFBD>Ҳ<EFBFBD><EFBFBD>ÿ<EFBFBD><EFBFBD><EFBFBD>˵<EFBFBD><EFBFBD><EFBFBD>֪<EFBFBD><EFBFBD>أ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ʵ<EFBFBD><EFBFBD><EFBFBD>и<EFBFBD>ע<EFBFBD><EFBFBD>ͼ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ӿ<EFBFBD><EFBFBD><EFBFBD>չ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ƿ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>˷<EFBFBD><EFBFBD>Ǩ<EFBFBD>ơ<EFBFBD>
<EFBFBD><EFBFBD><EFBFBD><EFBFBD>20<EFBFBD><EFBFBD><EFBFBD>ͳ<EFBFBD><EFBFBD><EFBFBD><EFBFBD>кܶ<EFBFBD>ѧ<EFBFBD>߿<EFBFBD>ʼ<EFBFBD>о<EFBFBD>ͼ<EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ǩ<EFBFBD><EFBFBD><EFBFBD>ˣ<EFBFBD><EFBFBD><EFBFBD>ʱ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ͨ<EFBFBD><EFBFBD><EFBFBD><EFBFBD>ȡ<EFBFBD><EFBFBD><EFBFBD>ͼ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ɫ<EFBFBD><EFBFBD><EFBFBD>߽<EFBFBD>֮<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ɷ<EFBFBD><EFBFBD>Ǩ<EFBFBD>ƣ<EFBFBD><EFBFBD><EFBFBD><EFBFBD>߼<EFBFBD>һЩ<EFBFBD>ǽ<EFBFBD><EFBFBD><EFBFBD><EFBFBD>ѧ<EFBFBD>и<EFBFBD><EFBFBD><EFBFBD>ͼ<EFBFBD><EFBFBD>任<EFBFBD><EFBFBD>ͳ<EFBFBD>Ʒ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ɷ<EFBFBD><EFBFBD>Ǩ<EFBFBD>ƣ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ч<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>롣ֱ<EFBFBD><EFBFBD>2015<EFBFBD><EFBFBD><EFBFBD>Ժ<EFBFBD><EFBFBD>ܵ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ڼ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ӿ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ֵ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ǽ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ǿ<EFBFBD><EFBFBD><EFBFBD>ͼ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ȡ<EFBFBD><EFBFBD><EFBFBD>ܣ<EFBFBD><EFBFBD><EFBFBD>ͼ<EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ǩ<EFBFBD>Ƶ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>õ<EFBFBD><EFBFBD>˿<EFBFBD><EFBFBD>Ƹ<EFBFBD><EFBFBD>õĽ<EFBFBD><EFBFBD><EFBFBD><EFBFBD>
2 PyTorchͼ<68><CDBC><EFBFBD><EFBFBD>Ǩ<EFBFBD><C7A8>ʵս
ʵ<EFBFBD>ֲ<EFBFBD><EFBFBD>裺
<20><><EFBFBD><EFBFBD><EFBFBD>ȣ<EFBFBD><C8A3><EFBFBD><EFBFBD><EFBFBD>Ҫ<EFBFBD><D2AA>ȡһ<C8A1><D2BB><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ͼƬ<CDBC><C6AC>һ<EFBFBD>ŷ<EFBFBD><C5B7>ͼƬ<CDBC><C6AC>
<20><>Ȼ<EFBFBD><C8BB><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ֵ<EFBFBD><D6B5><EFBFBD><EFBFBD><EFBFBD>ݶ<EFBFBD><DDB6><EFBFBD>ֵ<EFBFBD><D6B5><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ͼƬ֮<C6AC><D6AE><EFBFBD><EFBFBD><EFBFBD>ݲ<EFBFBD><DDB2><EFBFBD>̶ȣ<CCB6><C8A3>ͷ<EFBFBD><CDB7><EFBFBD><EFBFBD><EFBFBD>ֵ<EFBFBD><D6B5><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ͼƬ֮<C6AC><D6AE>ķ<EFBFBD><C4B7><EFBFBD><EFBFBD><EFBFBD>̶ȣ<CCB6><C8A3><EFBFBD>
<20><><EFBFBD><EFBFBD>󣬽<EFBFBD><F3A3ACBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ģ<EFBFBD>ͣ<EFBFBD><CDA3><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ͼƬ<CDBC>е<EFBFBD><D0B5><EFBFBD><EFBFBD>ݺͷ<DDBA><CDB7>ͼƬ<CDBC>ķ<EFBFBD><C4B7><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ȡ<EFBFBD><C8A1><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ͼƬΪ<C6AC><CEAA>׼<EFBFBD><D7BC><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>뽨<EFBFBD><EBBDA8><EFBFBD><EFBFBD>ģ<EFBFBD><C4A3><EFBFBD>У<EFBFBD><D0A3><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ϵ<EFBFBD><CFB5><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ݶ<EFBFBD><DDB6><EFBFBD>ֵ<EFBFBD>ͷ<EFBFBD><CDB7><EFBFBD><EFBFBD><EFBFBD>ֵ<EFBFBD><D6B5><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>С<EFBFBD><D0A1><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ͼƬ<CDBC><C6AC><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ںϵ<DABA>ͼƬ
<EFBFBD>ȵ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>صİ<EFBFBD><EFBFBD><EFBFBD>
import torch
from torch.autograd import Variable
import torchvision
from torchvision import transforms, models
import copy
from PIL import Image
import matplotlib.pyplot as plt
<EFBFBD><EFBFBD><EFBFBD>Ŷ<EFBFBD>ȡ<EFBFBD><EFBFBD><EFBFBD><EFBFBD>ͼƬ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ͼƬ<EFBFBD>ͷ<EFBFBD><EFBFBD>ͼƬ<EFBFBD><EFBFBD><EFBFBD><EFBFBD>ͼƬչʾ<EFBFBD><EFBFBD><EFBFBD>£<EFBFBD>
loadimg()<29><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ǵ<EFBFBD><C7B4>ͼƬ<CDBC><C6AC>·<EFBFBD><C2B7><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>£<EFBFBD>
#--------------------------1.<2E><><EFBFBD><EFBFBD>Ԥ<EFBFBD><D4A4><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>
transform = transforms.Compose([transforms.Resize([224,224]),
transforms.ToTensor()])
def loadimg(path = None):
img = Image.open(path)
img = transform(img)
img = img.unsqueeze(0)
return img
content_img = loadimg('images/1.jpg') #<23><><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Լ<EFBFBD><D4BC><EFBFBD><EFBFBD>ͼƬ<CDBC><C6AC>λ<EFBFBD><CEBB>
content_img = Variable(content_img).cuda()
style_img = loadimg('images/2.jpg')
style_img = Variable(style_img).cuda()
2.1 ͼ<><CDBC><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ʧ
<EFBFBD><EFBFBD><EFBFBD>ݶ<EFBFBD><EFBFBD><EFBFBD>ֵ<EFBFBD><EFBFBD><EFBFBD><EFBFBD>ʹ<EFBFBD>þ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ϊ<EFBFBD><EFBFBD>ʧ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ڴ<EFBFBD><EFBFBD><EFBFBD><EFBFBD>ж<EFBFBD><EFBFBD><EFBFBD><EFBFBD>ͼ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ʧ<EFBFBD><EFBFBD><EFBFBD>£<EFBFBD>
class Content_loss(torch.nn.Module):
def __init__(self,weight,target):
super(Content_loss,self).__init__()
self.weight = weight
self.target = target.detach()*weight
self.loss_fn = torch.nn.MSELoss()
def forward(self,in_put):
self.loss = self.loss_fn(in_put*self.weight,self.target)
return in_put
def backward(self):
self.loss.backward(retain_graph = True)
return self.loss
<EFBFBD><EFBFBD><EFBFBD>ϴ<EFBFBD><EFBFBD><EFBFBD><EFBFBD>еIJ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>
target<65><74>ͨ<EFBFBD><CDA8><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ȡ<EFBFBD><C8A1><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ͼ<EFBFBD><CDBC><EFBFBD>е<EFBFBD><D0B5><EFBFBD><EFBFBD>ݣ<EFBFBD>
weight<68><74><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>õ<EFBFBD>һ<EFBFBD><D2BB>Ȩ<EFBFBD>ز<EFBFBD><D8B2><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ݺͷ<DDBA><CDB7><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ϳ<EFBFBD>ͼ<EFBFBD><CDBC><EFBFBD>Ӱ<EFBFBD><D3B0>̶ȣ<CCB6>
in_put<75><74><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ͼ<EFBFBD><CDBC>
target.detach()<29><><EFBFBD>ڶ<EFBFBD><DAB6><EFBFBD>ȡ<EFBFBD><C8A1><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ݽ<EFBFBD><DDBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ҫ<EFBFBD><D2AA><EFBFBD><EFBFBD><EFBFBD>ݶȣ<DDB6>
forward<72><64><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ڼ<EFBFBD><DABC><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ͼ<EFBFBD><CDBC><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ͼ<EFBFBD><CDBC>֮<EFBFBD><D6AE><EFBFBD><EFBFBD><EFBFBD>ʧֵ<CAA7><D6B5>
backward<72><64><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ݼ<EFBFBD><DDBC><EFBFBD>õ<EFBFBD><C3B5><EFBFBD><EFBFBD><EFBFBD>ʧֵ<CAA7><D6B5><EFBFBD>к<EFBFBD><D0BA>򴫲<EFBFBD><F2B4ABB2><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ʧֵ
2.2 ͼ<><CDBC>ķ<EFBFBD><C4B7><EFBFBD><EFBFBD>ʧ
<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ͬ<EFBFBD><EFBFBD>ʹ<EFBFBD>þ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ϊ<EFBFBD><EFBFBD>ʧ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>£<EFBFBD>
class Style_loss(torch.nn.Module):
def __init__(self,weight,target):
super(Style_loss,self).__init__()
self.weight = weight
self.target = target.detach()*weight
self.loss_fn = torch.nn.MSELoss()
self.gram = Gram_matrix()
def forward(self,in_put):
self.Gram = self.gram(in_put.clone())
self.Gram.mul_(self.weight)
self.loss = self.loss_fn(self.Gram,self.target)
return in_put
def backward(self):
self.loss.backward(retain_graph = True)
return self.loss
<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ʧ<EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ĵ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ʧ<EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ĵ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ƣ<EFBFBD><EFBFBD><EFBFBD>֮ͬ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ڴ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Gram_matrix<EFBFBD>ඨ<EFBFBD><EFBFBD><EFBFBD>ʵ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>˷<EFBFBD><EFBFBD><EFBFBD><EFBFBD>ʧ<EFBFBD>ļ<EFBFBD><EFBFBD>㡣<EFBFBD><EFBFBD><EFBFBD><EFBFBD>ʵ<EFBFBD>ֵ<EFBFBD><EFBFBD>Ǹ<EFBFBD><EFBFBD><EFBFBD>ķ<EFBFBD><EFBFBD><EFBFBD><EFBFBD>Gram matrix<69><78><EFBFBD>Ĺ<EFBFBD><C4B9>ܡ<EFBFBD><DCA1><EFBFBD><EFBFBD><EFBFBD>ͨ<EFBFBD><CDA8><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ȡ<EFBFBD>˷<EFBFBD><CBB7>ͼƬ<CDBC>ķ<EFBFBD><C4B7><EFBFBD><EFBFBD>Щ<EFBFBD><D0A9><EFBFBD><EFBFBD><EFBFBD>ʵ<EFBFBD><CAB5><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ɵģ<C9B5><C4A3><EFBFBD><EFBFBD>ֵĴ<D6B5>С<EFBFBD><D0A1><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ͼƬ<CDBC>з<EFBFBD><D0B7><EFBFBD>ͻ<EFBFBD><CDBB><EFBFBD>̶ȣ<CCB6><C8A3><EFBFBD>Gram<61><6D><EFBFBD><EFBFBD><EFBFBD>Ǿ<EFBFBD><C7BE><EFBFBD><EFBFBD><EFBFBD>ڻ<EFBFBD><DABB><EFBFBD><EFBFBD>㣬<EFBFBD><E3A3AC><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>뵽<EFBFBD>þ<EFBFBD><C3BE><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ͼ<EFBFBD>д<EFBFBD><D0B4><EFBFBD><EFBFBD><EFBFBD>ֻ<EFBFBD><D6BB>ø<EFBFBD><C3B8><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>൱<EFBFBD><E0B5B1>ͼƬ<CDBC>ķ<EFBFBD>񱻷Ŵ<F1B1BBB7><C5B4>ˣ<EFBFBD><CBA3>Ŵ<EFBFBD>ķ<EFBFBD><C4B7><EFBFBD>ٲ<EFBFBD><D9B2><EFBFBD><EFBFBD><EFBFBD>ʧ<EFBFBD><CAA7><EFBFBD>㣬<EFBFBD><E3A3AC><EFBFBD>ܹ<EFBFBD><DCB9><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ĺϳ<C4BA>ͼƬ<CDBC><C6AC><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ӱ<EFBFBD>졣<EFBFBD><ECA1A3><EFBFBD><EFBFBD><EFBFBD>Ĵ<EFBFBD><C4B4><EFBFBD><EFBFBD><EFBFBD><EFBFBD>£<EFBFBD>
class Gram_matrix(torch.nn.Module):
def forward(self,in_put):
a,b,c,d = in_put.size()
feature = in_put.view(a*b,c*d)
gram = torch.mm(feature,feature.t())
return gram.div(a*b*c*d)
2.3 ģ<>ʹ<CDB4>Ͳ<EFBFBD><CDB2><EFBFBD><EFBFBD>Ż<EFBFBD>
<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ʧ<EFBFBD>ͷ<EFBFBD><EFBFBD><EFBFBD><EFBFBD>ʧ֮<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ǻ<EFBFBD><EFBFBD><EFBFBD>Ҫ<EFBFBD>һ<EFBFBD><EFBFBD><EFBFBD>Զ<EFBFBD><EFBFBD><EFBFBD>ģ<EFBFBD>ͣ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ʹ<EFBFBD><EFBFBD>VGG16ģ<EFBFBD>ͣ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ģ<EFBFBD><EFBFBD><EFBFBD>С<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ҫ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ǩ<EFBFBD><EFBFBD>һ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ȡ<EFBFBD><EFBFBD><EFBFBD>֣<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ز<EFBFBD><EFBFBD>֣<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>£<EFBFBD>
cnn = models.vgg16(pretrained = True).features #Ǩ<><C7A8>VGG16<31>ܹ<EFBFBD><DCB9><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ȡ<EFBFBD><C8A1><EFBFBD><EFBFBD>
# if use_gpu:
# cnn = cnn.cuda()
#ָ<><D6B8><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>зֱ<D0B7><D6B1><EFBFBD><EFBFBD><EFBFBD>һ<EFBFBD><D2BB><EFBFBD><EFBFBD>ȡ<EFBFBD><C8A1><EFBFBD>ݺͷ<DDBA><CDB7>
content_layer = ["Conv_3"]
style_layer = ["Conv_1","Conv_2","Conv_3","Conv_4"]
#<23><><EFBFBD>屣<EFBFBD><E5B1A3><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ʧ<EFBFBD>ͷ<EFBFBD><CDB7><EFBFBD><EFBFBD>ʧ<EFBFBD><CAA7><EFBFBD>б<EFBFBD>
content_losses = []
style_losses = []
#ָ<><D6B8><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ʧ<EFBFBD>ͷ<EFBFBD><CDB7><EFBFBD><EFBFBD>ʧ<EFBFBD><CAA7><EFBFBD><EFBFBD><EFBFBD>õ<EFBFBD><C3B5><EFBFBD><EFBFBD>ں<EFBFBD>ͼƬ<CDBC><C6AC>Ӱ<EFBFBD><D3B0>Ȩ<EFBFBD><C8A8>
content_weight = 1
style_weight = 1000
<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ͷϷ<EFBFBD><EFBFBD><EFBFBD>ͼ<EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ǩ<EFBFBD><EFBFBD>ģ<EFBFBD>ͣ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>£<EFBFBD>
new_model = torch.nn.Sequential() #<23><><EFBFBD><EFBFBD><EFBFBD>յ<EFBFBD>ģ<EFBFBD><C4A3>
model = copy.deepcopy(cnn)
#deepcopy<70><EFBFBD>ƣ<EFBFBD><C6A3><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ƶ<EFBFBD><C6B6><EFBFBD><EFBFBD><EFBFBD>ȫ<EFBFBD>ٸ<EFBFBD><D9B8><EFBFBD>һ<EFBFBD><D2BB><EFBFBD><EFBFBD>Ϊ<EFBFBD><CEAA><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>¸<EFBFBD><C2B8>嵥<EFBFBD><E5B5A5><EFBFBD><EFBFBD><EFBFBD>ڣ<EFBFBD><DAA3>ı<EFBFBD>ԭ<EFBFBD>б<EFBFBD><D0B1><EFBFBD><EFBFBD>ƶ<EFBFBD><C6B6>󲻻<EFBFBD><F3B2BBBB><EFBFBD>Ѿ<EFBFBD><D1BE><EFBFBD><EFBFBD>Ƴ<EFBFBD><C6B3><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><C2B6><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ӱ<EFBFBD>졣
#copydz<79><C7B3><EFBFBD>ƣ<EFBFBD><C6A3><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>һ<EFBFBD><D2BB><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ķ<EFBFBD><C4B6>󵥶<EFBFBD><F3B5A5B6><EFBFBD><EFBFBD>ڣ<EFBFBD><DAA3><EFBFBD>ֻ<EFBFBD>ǽ<EFBFBD>ԭ<EFBFBD>е<EFBFBD><D0B5><EFBFBD><EFBFBD>ݿ<EFBFBD><DDBF><EFBFBD><EFBFBD>һ<EFBFBD><D2BB><EFBFBD>±<EFBFBD>ǩ
#<23><><EFBFBD>Ե<EFBFBD><D4B5><EFBFBD><EFBFBD><EFBFBD>һ<EFBFBD><D2BB><EFBFBD><EFBFBD>ǩ<EFBFBD><C7A9><EFBFBD>ı<EFBFBD><C4B1>ʱ<EFBFBD><CAB1><EFBFBD><EFBFBD><EFBFBD>ݿ<EFBFBD>ͻᷢ<CDBB><E1B7A2><EFBFBD>仯<EFBFBD><E4BBAF><EFBFBD><EFBFBD>һ<EFBFBD><D2BB><EFBFBD><EFBFBD>ǩҲ<C7A9><D2B2><EFBFBD><EFBFBD>֮<EFBFBD>ı䡣
gram = Gram_matrix()
use_gpu = torch.cuda.is_available()
if use_gpu:
model = model.cuda()
new_model = new_model.cuda()
gram = gram.cuda()
index = 1
#ֻʹ<D6BB><CAB9>Ǩ<EFBFBD><C7A8>ģ<EFBFBD><C4A3><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ȡ<EFBFBD><C8A1><EFBFBD>ֵ<EFBFBD>ǰ8<C7B0><38>
for layer in list(model)[:8]:
if isinstance(layer,torch.nn.Conv2d):
name = "Conv_" + str(index)
#ʹ<><CAB9>add_module<6C><65><EFBFBD><EFBFBD><EFBFBD><EFBFBD>յ<EFBFBD>ģ<EFBFBD>ͼ<EFBFBD><CDBC><EFBFBD>ָ<EFBFBD><D6B8><EFBFBD>IJ<EFBFBD><C4B2>ģ<EFBFBD><C4A3>
new_model.add_module(name,layer)
if name in content_layer:
target = new_model(content_img).clone()
content_loss = Content_loss(content_weight,target)
new_model.add_module("content_loss_"+str(index),content_loss)
content_losses.append(content_loss)
if name in style_layer:
target = new_model(style_img).clone()
target = gram(target)
style_loss = Style_loss(style_weight,target)
new_model.add_module("style_loss_"+str(index),style_loss)
style_losses.append(style_loss)
if isinstance(layer,torch.nn.ReLU):
name = "ReLU_"+str(index)
new_model.add_module(name,layer)
index = index + 1
if isinstance(layer,torch.nn.MaxPool2d):
name = "MaxPool_"+str(index)
new_model.add_module(name,layer)
<EFBFBD><EFBFBD><EFBFBD>ϴ<EFBFBD><EFBFBD><EFBFBD><EFBFBD>У<EFBFBD>for layer in list(model)[:8]ָ<><D6B8><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ǽ<EFBFBD><C7BD><EFBFBD><EFBFBD>õ<EFBFBD>Ǩ<EFBFBD><C7A8>ģ<EFBFBD><C4A3><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ȡ<EFBFBD><C8A1><EFBFBD>ֵ<EFBFBD>ǰ8<C7B0>㣬<EFBFBD><E3A3AC>Ϊ<EFBFBD><CEAA><EFBFBD>ǵ<EFBFBD><C7B5><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ȡ<EFBFBD>ͷ<EFBFBD><CDB7><EFBFBD><EFBFBD>ȡ<EFBFBD><C8A1>ǰ8<C7B0><38><EFBFBD><EFBFBD>Ѿ<EFBFBD><D1BE><EFBFBD><EFBFBD><EFBFBD>ˡ<EFBFBD>Ȼ<EFBFBD><C8BB><EFBFBD><EFBFBD>һ<EFBFBD><D2BB><EFBFBD>յ<EFBFBD>ģ<EFBFBD>ͣ<EFBFBD>ʹ<EFBFBD><CAB9> torch.nn.Module <20><><EFBFBD>add_module<6C><65><EFBFBD><EFBFBD><EFBFBD><EFBFBD>յ<EFBFBD>ģ<EFBFBD><C4A3><EFBFBD>м<EFBFBD><D0BC><EFBFBD>ָ<EFBFBD><D6B8><EFBFBD>IJ<EFBFBD><C4B2>ģ<EFBFBD>飬<EFBFBD><E9A3AC><EFBFBD>õ<EFBFBD><C3B5><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Զ<EFBFBD><D4B6><EFBFBD><EFBFBD>ͼ<EFBFBD><CDBC><EFBFBD><EFBFBD>Ǩ<EFBFBD><C7A8>ģ<EFBFBD>͡<EFBFBD>add_module<6C><65><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ݵIJ<DDB5><C4B2><EFBFBD><EFBFBD>ֱ<EFBFBD><D6B1>Dz<EFBFBD>ε<EFBFBD><CEB5><EFBFBD><EFBFBD>ֺ<EFBFBD>ģ<EFBFBD>飬<EFBFBD><E9A3AC>ģ<EFBFBD><C4A3><EFBFBD><EFBFBD>ʹ<EFBFBD><CAB9> isinstance ʵ<><CAB5><EFBFBD><EFBFBD>⺯<EFBFBD><E2BAAF><EFBFBD>õ<EFBFBD><C3B5>ģ<EFBFBD><C4A3><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ƕ<EFBFBD>Ӧ<EFBFBD>IJ<EFBFBD>Ρ<EFBFBD><CEA1>ڶ<EFBFBD><DAB6><EFBFBD><EFBFBD>ģ<EFBFBD><C4A3>֮<EFBFBD><D6AE><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>д<EFBFBD>ӡ<EFBFBD><D3A1><EFBFBD><EFBFBD><EFBFBD>
print(new_model)
<EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ľ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>£<EFBFBD>
Sequential(
(Conv_1): Conv2d(3, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(style_loss_1): Style_loss(
(loss_fn): MSELoss()
(gram): Gram_matrix()
)
(ReLU_1): ReLU(inplace=True)
(Conv_2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(style_loss_2): Style_loss(
(loss_fn): MSELoss()
(gram): Gram_matrix()
)
(ReLU_2): ReLU(inplace=True)
(MaxPool_3): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
(Conv_3): Conv2d(64, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(content_loss_3): Content_loss(
(loss_fn): MSELoss()
)
(style_loss_3): Style_loss(
(loss_fn): MSELoss()
(gram): Gram_matrix()
)
(ReLU_3): ReLU(inplace=True)
(Conv_4): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(style_loss_4): Style_loss(
(loss_fn): MSELoss()
(gram): Gram_matrix()
)
)
<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Dz<EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ż<EFBFBD><EFBFBD>Ĵ<EFBFBD><EFBFBD>룺
input_img = content_img.clone()
parameter = torch.nn.Parameter(input_img.data)
# <20><><EFBFBD><EFBFBD><EFBFBD>ģ<EFBFBD><C4A3><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ҫ<EFBFBD>Ż<EFBFBD><C5BB><EFBFBD><EFBFBD><EFBFBD>ʧֵ<CAA7>ж<EFBFBD><D0B6><EFBFBD><EFBFBD><EFBFBD>ҹ<EFBFBD>ģ<EFBFBD>ϴ<EFBFBD>ʹ<EFBFBD>ø<EFBFBD><C3B8>Ż<EFBFBD><C5BB><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ȡ<EFBFBD>ø<EFBFBD><C3B8>õ<EFBFBD>Ч<EFBFBD><D0A7><EFBFBD><EFBFBD>
optimizer = torch.optim.LBFGS([parameter])
2.4 ѵ<><D1B5><EFBFBD><EFBFBD><C2B6><EFBFBD>ľ<EFBFBD><C4BE><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>
<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ģ<EFBFBD>͵Ĵ<EFBFBD><EFBFBD><EFBFBD>Ż<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ķ<EFBFBD><EFBFBD><EFBFBD>󣬾Ϳ<EFBFBD><EFBFBD>Կ<EFBFBD>ʼ<EFBFBD><EFBFBD><EFBFBD><EFBFBD>ģ<EFBFBD>͵<EFBFBD>ѵ<EFBFBD><EFBFBD><EFBFBD>Ͳ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ż<EFBFBD><EFBFBD>ˣ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>£<EFBFBD>
# <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ģ<EFBFBD><C4A3>ѵ<EFBFBD><D1B5><EFBFBD>Ͳ<EFBFBD><CDB2><EFBFBD><EFBFBD>Ż<EFBFBD>
epoch_n = 300
epoch = [0]
while epoch[0] <= epoch_n:
def closure():
optimizer.zero_grad()
style_score = 0
content_score = 0
parameter.data.clamp_(0,1)
new_model(parameter)
for sl in style_losses:
style_score += sl.backward()
for cl in content_losses:
content_score += cl.backward()
epoch[0] += 1
if epoch[0] % 50 == 0:
print('Epoch:{} Style_loss: {:4f} Content_loss: {:.4f}'.format(epoch[0], style_score.data.item(),
content_score.data.item()))
return style_score + content_score
<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>optimizer.step(closure)
<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>
<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>
<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>£<EFBFBD>
Epoch:50 Style_loss: 8.816691 Content_loss: 1.8809
Epoch:100 Style_loss: 3.377805 Content_loss: 1.7790
Epoch:150 Style_loss: 0.531610 Content_loss: 1.8476
Epoch:200 Style_loss: 0.143326 Content_loss: 1.7222
Epoch:250 Style_loss: 0.107568 Content_loss: 1.6353
Epoch:300 Style_loss: 0.099968 Content_loss: 1.6046
<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>붨<EFBFBD><EFBFBD>ѵ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ϊ300<EFBFBD>Σ<EFBFBD>ʹ<EFBFBD><EFBFBD> sl.backward<72><64>cl.backwardʵ<64><CAB5><EFBFBD><EFBFBD>ǰ<EFBFBD>򴫲<EFBFBD><F2B4ABB2>ͺ<EFBFBD><CDBA>򴫲<EFBFBD><F2B4ABB2>㷨<EFBFBD><E3B7A8>ÿ<EFBFBD><C3BF><EFBFBD><EFBFBD> 50 <20><>ѵ<EFBFBD><D1B5><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ʧֵ<CAA7><D6B5><EFBFBD><EFBFBD>һ<EFBFBD>δ<EFBFBD>ӡ<EFBFBD><D3A1><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>£<EFBFBD>
#<23>Է<EFBFBD><D4B7>Ǩ<EFBFBD><C7A8>ͼƬ<CDBC><C6AC><EFBFBD>
output = parameter.data
unloader = transforms.ToPILImage()
plt.ion()
plt.figure()
def imshow(tensor, title=None):
image = tensor.clone().cpu()
image = image.view(3, 224, 224)
image = unloader(image)
plt.imshow(image)
if title is not None:
plt.title(title)
plt.pause(0.001) # pause<73><65><EFBFBD><EFBFBD>ͼ<EFBFBD><CDBC><EFBFBD><EFBFBD><EFBFBD>
imshow(output, title='Output Image')
#<23><><EFBFBD><EFBFBD>sphinx_gallery_thumbnail_number = 4
plt.ioff()
plt.show()
<EFBFBD><EFBFBD><EFBFBD>֮<EFBFBD><EFBFBD><EFBFBD>ͼƬ<EFBFBD><EFBFBD><EFBFBD>£<EFBFBD>
Ч<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>У<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ͼƬ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>˷<EFBFBD><EFBFBD>ͼƬ<EFBFBD>ķ<EFBFBD>񣬲<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Щ<EFBFBD>ط<EFBFBD><EFBFBD><EFBFBD>RGBͨ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>̫<EFBFBD>У<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ż<EFBFBD><EFBFBD>ռ䡣<EFBFBD>ȽϺõ<EFBFBD><EFBFBD><EFBFBD><EFBFBD>ͼƬ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ɫ<EFBFBD>Ļ<EFBFBD><EFBFBD><EFBFBD><EFBFBD>ϻ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ͼ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ϣ<EFBFBD><EFBFBD><EFBFBD><EFBFBD>ͷ<EFBFBD><EFBFBD>ͼƬ<EFBFBD><EFBFBD>Ϊ<EFBFBD><EFBFBD><EFBFBD>ƣ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ͼ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ȼû<EFBFBD>з<EFBFBD><EFBFBD><EFBFBD>̫<EFBFBD><EFBFBD>ı仯<EFBFBD><EFBFBD>
3 С<><D0A1>
ʵ<EFBFBD><EFBFBD><EFBFBD>˻<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ͼ<EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ǩ<EFBFBD><EFBFBD><EFBFBD>㷨<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ͼ<EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ǩ<EFBFBD>ƹ<EFBFBD><EFBFBD>̵<EFBFBD>ʵ<EFBFBD>ִ<EFBFBD><EFBFBD><EFBFBD>һ<EFBFBD><EFBFBD><EFBFBD>Ƚ<EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ե<EFBFBD>ȱ<EFBFBD>㣬<EFBFBD><EFBFBD><EFBFBD><EFBFBD>ÿ<EFBFBD><EFBFBD>ѵ<EFBFBD><EFBFBD>ֻ<EFBFBD>ܶ<EFBFBD><EFBFBD><EFBFBD><EFBFBD>е<EFBFBD>һ<EFBFBD>ַ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ǩ<EFBFBD>ƣ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ҫ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ǩ<EFBFBD>ƣ<EFBFBD>Ҫ<EFBFBD><EFBFBD><EFBFBD><EFBFBD>ģ<EFBFBD>ͽ<EFBFBD><EFBFBD><EFBFBD>ѵ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ҫͨ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ݺͷ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ò<EFBFBD>ͬ<EFBFBD><EFBFBD>Ȩ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ʒ<EFBFBD><EFBFBD><EFBFBD><EFBFBD>ڵķ<EFBFBD>ʽ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ַ<EFBFBD>ʽ<EFBFBD><EFBFBD>ʵ<EFBFBD><EFBFBD>Ӧ<EFBFBD><EFBFBD><EFBFBD>в<EFBFBD>̫<EFBFBD><EFBFBD><EFBFBD>룬<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ҫ<EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ч<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ܵ<EFBFBD>ʵ<EFBFBD>ַ<EFBFBD>ʽ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ֻ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Щ<EFBFBD><EFBFBD>ʵ<EFBFBD><EFBFBD>ѧϰ<EFBFBD><EFBFBD><EFBFBD>˽⵽<EFBFBD><EFBFBD><EFBFBD>ٷ<EFBFBD><EFBFBD>Ǩ<EFBFBD>ƣ<EFBFBD>ֻ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ʹ<EFBFBD><EFBFBD>pytorchʵ<EFBFBD>ֵ<EFBFBD><EFBFBD><EFBFBD><EFBFBD>ϱȽ<EFBFBD><EFBFBD>٣<EFBFBD>û<EFBFBD>гɹ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ô<EFBFBD><EFBFBD>롣<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ǻ<EFBFBD><EFBFBD><EFBFBD><EFBFBD>Ż<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>һЩ<EFBFBD>ɹ<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>渽<EFBFBD><EFBFBD>
<EFBFBD><EFBFBD>󣬲ο<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>£<EFBFBD>
https://www.cnblogs.com/subic/p/8110478.html
https://zhuanlan.zhihu.com/p/23479658
https://blog.csdn.net/qq_17506541/article/details/80012589
https://blog.csdn.net/u014380165/article/details/76286047
https://blog.csdn.net/m0_46653437/article/details/108470002?spm=1001.2014.3001.5506
https://blog.csdn.net/xs1997/article/details/104503934/?utm_medium=distribute.pc_relevant.none-task-blog-2~default~baidujs_title~default-0--blog-104502633.pc_relevant_aa&spm=1001.2101.3001.4242.1&utm_relevant_index=3
Write Preview
Loading…
Cancel
Save