From 0639032b6c74d54661c0c86836df71926eaae41b Mon Sep 17 00:00:00 2001
From: bishe <123456789@163.com>
Date: Sun, 23 Feb 2025 21:15:41 +0800
Subject: [PATCH] use kun's ctn method

---
 checkpoints/ROMA_UNSB_001/loss_log.txt        |   1 +
 .../roma_unsb_model.cpython-39.pyc            | Bin 20066 -> 19935 bytes
 models/roma_unsb_model.py                     |  17 ++++-------------
 3 files changed, 5 insertions(+), 13 deletions(-)

diff --git a/checkpoints/ROMA_UNSB_001/loss_log.txt b/checkpoints/ROMA_UNSB_001/loss_log.txt
index e2b8c6e..741a618 100644
--- a/checkpoints/ROMA_UNSB_001/loss_log.txt
+++ b/checkpoints/ROMA_UNSB_001/loss_log.txt
@@ -33,3 +33,4 @@
 ================ Training Loss (Sun Feb 23 18:59:52 2025) ================
 ================ Training Loss (Sun Feb 23 19:03:05 2025) ================
 ================ Training Loss (Sun Feb 23 19:03:57 2025) ================
+================ Training Loss (Sun Feb 23 21:11:47 2025) ================
diff --git a/models/__pycache__/roma_unsb_model.cpython-39.pyc b/models/__pycache__/roma_unsb_model.cpython-39.pyc
index cb9f3451854184b9dd7c8c10760781f83c7b950e..0c9cd4b7b8435d3c603e9599112a983eb428a6c2 100644
GIT binary patch
delta 2232
zcmZ{leQXp}5Ww%<^?KK9dp%m((o##IrS*WL*9KHdTPUU9EhgF^JuA!hZfUpN?OkX0
zti@_ES_L8oc`At*XhRJ0moaNl5To%AqhiDl)MO)K#1B*?U;<){A2aU=A4V>DznwR4
zcHX>sGkf2@z)tO9UO7A4W5eHM=^KL|-2aMq%Ev<H%f19-eP&tiQkECSMJscY0-6$)
zbak@=1Lja(r9GV%Ai|%?A7y<!tDt$UM2Zar8R1Wd5Xtuv`Uy3JKEi#3`B$-&q)Nhi
z#2y@a8p6%IPFT;Y(`T7)7F@P7i61K3RTRe+%c;pQs9Vl(T+<CmpbADzxp*yON&fn5
ze~^rX+H!@5;%ZpYVFXp1GXeokF<K=p{F9AgjCM#_nlnm9T;<#7t?h(sv5ByDpzd$d
zKgC}xZ(z^y)8z>^$$Kl76mB7P7sB#rVMU6{V+!an&i7RmvqAGnML%QD^TNvhTsoTN
z2pgK;2}|guNprGtpQBJD<RC3N4Ukn04$-1%E?@AM9rsukSjWCJcLly;*@MC;9F1!V
zjPeJoC)ktbpVgVH<Se~GF@TToBf{@WZE}XstEpzc@VjdIy~U_^<7mdA`^=*?Q?7|D
zBWJjIc=fn#xPGc-+D6ZaJyNv85pfmS9xB3l>r~r#(KH*kN$umdsSeb%i&>}9g>ns!
z&PVOI+AZewSLZ!wU5T@99KrGYX$RV@iR9jG^A{(pFoGYp9XK=1=hVGgdb5ll#2dnl
zU#MGMup3|4>>W@x#z865;;KGYR|e1WuI2OCPOdGlTY4T1Qnz+9Zx@9(GTTM?5W(lx
z=QU5#Y(IhW3nvLVgu{fd38xU27nG5t3W_cbf*g_H8(KL^*vm)j>xwDsP)Nup6d){z
zAtzy)AFJ<VyLmywB3ayXnsUD!g$6Q6Qw=2sqx3>I;d@egK-MCf6jib5RMNgg+B8Fr
zDewe|*`UZ#sYw!vs10R^rxcSD@HF4w(8{**pBuU+rpVN#D~4<s09$D0mXk)D!iPI?
z3ilA+BeW4(2=61TjExd@mZ4|{RN*`{b5w?ef``6gU+)#|MtPGWtqO*GH_WqwQmwx@
zRfQ+{bYtGE%jBB%hSF3mtONC*WgpNXjepV@U@`trV`U%%m7#3mkit|cg8%d)^~+8)
z(?8;g74LfNrFO5w>+mrjZ(Vu(jtjWz&w^j^0b2<ottLt28{Q&m86uP!jqAF!TB6P1
z6z^#&VIT8hWGTPUn&&)@HgnB=O+~Cx+$<&jI*%T5mW{A7qVYjJ8B-oeq&in@BW{$o
zfzSA*RYB*cC{~*boBQm;zoAGT#jh8rA)$kiMG)~<!9|jH6GVAZ=2DwegIj4-AmF0i
z9l8+-d7^S`AR<d`O>5w~qw;et{@}IXuZ8Rxkd5$A>SHdxHgoATBIaq{*6N?|(fkr^
zIs>Jv{_*1_>H%ibI(<&yRz%31ssO1r<ciixlvns1_5RG%?H7=>IKmu~st9F-d4zIA
z$l1~>t?RQ=cQTkBjStAt5N}^m>O6stmYErCCs^KmGN%NoEA2L0uipXZ`1$r*{J*0L
z_7X(eMBL&86nrXyev%gO+d9h9{~*Im-rq5MLUiw8nv1RvkdbJ3FG)EFe~vhU<;HeQ
zecAtSi-l+z$`bXYCuBp#?p{nPj}hObXu1@}Dqcj2;#ks9qgUD}$S-wlXB+lE(#a~t
z(a_>b6*x!|MM33(c}yLVALvq94WH57<6MBokMVGK%fyX;vR0Ok{jP>%iZK+Az~kho
zhH#AV1>rQ|EJ1WhA4yd0Kt&3aZ>UA@KwDhbe4-7-zoCXE@8Hl&5xlA=|MpzB*X?mv
zxhsIy{3ZDRoQx&LK;-&6RKXE~C|<V_hw#v1O@p_&-ZQr7VOnt!UL#Zy0thR;Cmu;g
P6<AD~^#qO2TGRP24Cg+0

delta 2347
zcmZ{mYitx%6oB{ccDu{AT}pwLZWk=Hv@Y5g50yT6m6k^f7;Olw)nU3j?QWT!S<Xy>
z)+Ln?D_}t3`oI)Ji~0kjF=kN-h>t%e8jTQ9abg4&A0#S1qT&lZ=MGRHvYYw#+;i@|
z=bU@)y|bqV*%y16U(U(#Iq>tS<4Bi(<BR^GTxO+T$W1a9P8a9RVq>Fdw6fMJped?k
z#QPM8q+?^toS9?*g+DTW0}J!){Kl1QNFftC5TQ)Ci{y5~-GmxKm~ana>SfF#sf-{Y
z2Jta65MJhYgT$*cC!~+&pLen~{N=*kg-NtnZcT<R!*WM-%`jmdssPe*U?pSg`Jtjv
zHFXkd%M*#|@rYsopsH_FMTMrA&5{;5>tGn86YQSpF~N17lI=9=LBdFEBCMULyFmJf
z_%kI9>~a2ONs<llw$hmen@Qb*uzXrXkyN=?0RuMieWd{wOCK(6XKaucl(pwkYL+Wv
zY9Tkwq(KAe`^)yZ3dDjuOGdW|a$JK~$apYaH|-xM`j}m@ik(dFuK1MY#6+ivs%r{t
z<o8zfvmNQbs<K${dm2G8KrZ21M95=1Im@TiRI%^)9X0L#AnLvNG~#2lrC+TX^7P+s
zI-`XtS9ivsBN|9Ky6Z`LGGhy44zv`Zr4ids(H`Bsu)8Vc8fqSPq+C&FG`PbR^%Ocb
z6`HmMM++PNUSXwK7Y!{=1%_F>Gv$z$;IlO47-~WN(%NB$%~lj^@tINJz}7Nz`NekY
z7U!lS>{^yWc2}py54*@7&AZbPn#hmOer?kAigf^si^Y1rc47V=%(l<pp>*n?6l;3i
z=&db=$GE3%3fsk})z$X@gdOaHUvIWw#cb4#6W*g4b3u_+X+d?R5tEZjaYJ!mUU6}!
zyjHnN7c>;Lv+`0`P}cRtK{2E*kfXK5A!cPqphl5Rg*3@nXax+AgQp2pZTO5ZnsAtK
zig21As~eZ~If+LJ`}n!K+5i<D3JBu~`3TEp$~~~3H`Xs@d-%@!n-_>}mZ7YdRcN3N
z8F5odz(yL-O8A<TK9IGjCaG~e^WCI9K-vs5o={*b|GRz?+s%s_n%Oqq(a<t^n2a7n
zF=f*P*i5aQtMcTpH;nIpo90<WSWRdmyo0btbxL@KO+_;;uf2NGbSOtlC}?kBu>yA9
zG)ScV2aRNH;f%NzF#xx!!*X^Q;3k2YmOHAbrc6JYp0M0;&E%;$<Cw;G%_(c~VW%)g
zh{dBS#L;s&M|i%}v~o7c$)pmM?0Bxd3+TK)jFT8A_?bCx<+1?Fbq1W1oa0=6`tpNw
z-@ol=N+274poq5-#OfX-iTuMGB+W*IvsB$Mq=mv4R`$reAUn=a&I=Y!p?=p~L&tDL
zc{*o)A)6!0y_KdJnLmq$_z~79Rqrx-63WJ;eVYiUqRiwJPVnf0YWD{y&PqSAV70UR
z7Zkt2M||4Ai@&ghkWCQTl*4Z%(`m9dDHXwXPM40ALAwUL<lSYM(XdaI6CF`mS~Pz-
zT#-0$Zwgh91V0k8uR}H?G5ZB0ZosLOjfi<af2k?dKZe@BlhZtuF3*n_s@N}(Pxb)W
zZzL2@CyzQ^q6LPJP!-CucT^#D6w#NEbQ7VJP(~<6gxyVT(yG;#U7Rje?~v61OeXU+
z75Nb7otqwLKEcL@$V^KWpXYQq+CnZk#}kVut~gK8>?4Q;6d8zIC=GilD@dwDg#8y+
z%}B~-Tvjcgsz-3AYCQ=9{-imw`6r8u65`y88$le88X7Oo$Y_#s5uwo{S<8zD*?uMc
zpZ6;AnJu>9MUOC@R6bK*t7wK4(M+i3@{+x*^Qj9<7z^=#m)?+RXT(ce;%pY**|N+{
zp>OA>TbjaGzwxbX1Fwa6B%zowJqiy{0^(1=yCi)~I71NcFL7AtZ~<*|pxp_yyl@N7
z4wbyWbpq4*p4Q2?PVoA@K5x0V6m+zQg7|NsCz0&67w9V*a)cnZrPtIU606iSc$5Fw
n+FQ|277yVNfo>Y8Kv<c}^k|Q&KqYA;g6`n=wdV8Zmd*GNuli@^

diff --git a/models/roma_unsb_model.py b/models/roma_unsb_model.py
index 45b220e..05cfaa7 100644
--- a/models/roma_unsb_model.py
+++ b/models/roma_unsb_model.py
@@ -527,26 +527,17 @@ class RomaUnsbModel(BaseModel):
                     setattr(self, "fake_"+str(t+1), Xt_1)
                     
         if self.opt.phase == 'train':
-            print(f'real_B0.shape = {real_B0.shape}  fake_B0.shape = {self.fake_B0.shape}')
-            print(f"self.real_B0.requires_grad: {real_B0.requires_grad}")
-            # 真实图像的梯度
-            real_gradient = torch.autograd.grad(real_B0.sum(), real_B0, create_graph=True)[0]
             # 生成图像的梯度
             fake_gradient = torch.autograd.grad(self.fake_B0.sum(), self.fake_B0, create_graph=True)[0]
             # 梯度图
-            self.weight_real, self.weight_fake = self.cao.generate_weight_map(fake_gradient)
-            
+            self.weight_fake = self.cao.generate_weight_map(fake_gradient)
             # 生成图像的CTN光流图
             self.f_content = self.ctn(self.weight_fake)
-            
-            # 把前面生成后的图片再加上noisy_map
-            self.fake_B_2 = self.fake_B + self.noisy_map
-            
             # 变换后的图片
-            wapped_fake_B = warp(self.fake_B, self.f_content)
-            
+            self.warped_real_A_noisy2 = warp(self.real_A_noisy, self.f_content)
+            self.warped_fake_B0 = warp(self.fake_B0,self.f_content)
             # 经过第二次生成器
-            self.fake_B_2 = self.netG(wapped_fake_B, self.time, z_in)
+            self.warped_fake_B0_2 = self.netG(self.warped_real_A_noisy2, self.time, z_in)   
             
     def compute_D_loss(self):
         """计算判别器的 GAN 损失"""