From 338b1b97aeb50eee6223ff49b4565bad7c7fb7a6 Mon Sep 17 00:00:00 2001
From: lisb20 <2692465509@qq.com>
Date: Thu, 27 Jul 2023 19:27:38 +0800
Subject: [PATCH] detr

---
 .../data/__pycache__/dataset.cpython-39.pyc   | Bin 3261 -> 3197 bytes
 NLOS_detr/data/dataset.py                     |   4 +-
 NLOS_detr/models.py                           |  42 ++++++++++++------
 .../utils/__pycache__/trainer.cpython-39.pyc  | Bin 15641 -> 15705 bytes
 NLOS_detr/utils/trainer.py                    |  13 +++---
 5 files changed, 38 insertions(+), 21 deletions(-)

diff --git a/NLOS_detr/data/__pycache__/dataset.cpython-39.pyc b/NLOS_detr/data/__pycache__/dataset.cpython-39.pyc
index 02ebd99c379b494f930fa2f5c1c402ca89d72fcb..0ce2a7cb3784d67f34190bcce54bf8c6451d8291 100644
GIT binary patch
delta 317
zcmdlh`B#EBk(ZZ?0SE-1A58hYk@pxQ<DSV^7-y^1uoP`5;b~^5VTk9cVaVdi;svt!
z@;FNPvjjk_6oz01O|H#rnT!~jGzBNWWR9Mkz*5erGWiCJ0we3>zbw9dx0rKM^NN&#
zhHEkwX>N{SJ<7yrFj<ZxjnQy&D~CO!+T^1gevImqML1pAG=Zv$3@68PTFU4FnYUQ-
z3R3fLF(;O#-(oDd#axnGaEmds$ZYZ!&fVPFAho(6!e??lkND(?T&FdIz?w=@i%T?F
zir7JN0U$yLL@)yhO}3&C5Ib~oD|f0^1W1Ib2&Abf7{nGT5(Tl%fy6D=ywco)$|7};
cfb(P#9uL-A%mpb)lM{I)*jzz^5kOn*0QziEr2qf`

delta 382
zcmW-bO-lk%6o&7eFUQfb$&w6&jG`Jv`f{TST^VjFLizv~V$BpvPQ&$96t0qf!J$<x
z>L-MY79p*QqV25iqQB5D2)fs^czJlw`*O}-ayV@^P18VpUT)r;-)Hj!;{XDDU!qaR
zpk@!XwLb1k7}40{*b~C2r;tx-^-&KsEWWkc<DBB2PAA%vf*e#sc4TlTtzqL@b6TWZ
zOxp?glItm^`lRJiCZD<<lYQrqg}W|oP|`edVP3h!OYoqiRmsA-1Vd#rMi%&EWR8oC
zNp6r--zB^bKcQoshl*at8MxKAa25i+qUA(14>RFHYLz2XJl{QIa`WPZVV}vg<1^eM
z1)$;A$g;?;2-${mbP9HiZ+mxCqHdSk3JHl_O~~j2nUk%NePJ9xCE{l4!jXt5kv+k2
jk`P?v!c_0P<6n`iFg75FR8_Lb8D^tNtt!kqzgqeSGlF78

diff --git a/NLOS_detr/data/dataset.py b/NLOS_detr/data/dataset.py
index 83bcc8a..426e827 100644
--- a/NLOS_detr/data/dataset.py
+++ b/NLOS_detr/data/dataset.py
@@ -37,8 +37,8 @@ def __init__(self,
             tmp_dir = [os.path.join(str(peo), d) for d in os.listdir(os.path.join(self.dataset_dir, str(peo)))]
             self.dirs += tmp_dir[:600]
         
-        print('dirs: ', len(self.dirs))
-        pdb.set_trace()
+        # print('dirs: ', len(self.dirs))
+        # pdb.set_trace()
 
         if use_fileclient:
             self.npy_loader = npy_loader()
diff --git a/NLOS_detr/models.py b/NLOS_detr/models.py
index ab80acb..38806eb 100644
--- a/NLOS_detr/models.py
+++ b/NLOS_detr/models.py
@@ -184,8 +184,15 @@ def __init__(self, pretrained: bool = True,
         for m in init_modules:
             m.apply(self._init_weights)
 
-    def forward(self, Ix: tuple):
-        I, x_gt = Ix  # (B, C, T, H, W)
+    def forward(self, Ix: tuple,):
+        I, x_gt,freeze = Ix  # (B, C, T, H, W)
+        if freeze:
+            for para in self.transformer.parameters():
+                para.requires_grad = False
+        else:
+            for para in self.transformer.parameters():
+                para.requires_grad = True
+        
         delta_I = torch.sub(I[:, :, 1:], I[:, :, :-1]).float()
         B, T = x_gt.shape[:2] ## (b,T,6)
         if self.warmup_frames > 0:
@@ -232,9 +239,9 @@ def forward(self, Ix: tuple):
         cost_matrix = torch.zeros((npeo,npeo), device=x_gt.device, requires_grad=False)
 
         pred_matched = torch.zeros((B,T,npeo * 2), device=x_gt.device, requires_grad=False)
-        x_loss_tot = 0
-        v_loss_tot = 0
-        m_loss_tot = 0      
+        x_loss_tot = torch.tensor(0.0,device=x_gt.device)
+        v_loss_tot = torch.tensor(0.0,device=x_gt.device)
+        m_loss_tot = torch.tensor(0.0,device=x_gt.device)      
 
         class_gt = torch.zeros((B, npeo, self.max_peo + 1), device=x_gt.device) ## class: idx==0 -> empty 只分为有无人
         valid_peo = []
@@ -256,12 +263,15 @@ def forward(self, Ix: tuple):
                 pred_peo.append(npeo - torch.sum(p==0))
                 for i in range(npeo):
                     for j in range(npeo):
-                        x_loss = self.criterion(x_pred[b,:,2*i:2*i+2], x_gt[b,self.warmup_frames:,2*j:2*j+2])
-                        v_loss = self.compute_v_loss(x_pred[b,:,2*i:2*i+2], x_gt[b,self.warmup_frames:,2*j:2*j+2])
+                        if not freeze:
+                            x_loss = self.criterion(x_pred[b,:,2*i:2*i+2], x_gt[b,self.warmup_frames:,2*j:2*j+2])
+                            v_loss = self.compute_v_loss(x_pred[b,:,2*i:2*i+2], x_gt[b,self.warmup_frames:,2*j:2*j+2])
                         # m_loss = self.criterion(class_pred[b,i,:], class_gt[b,j,:])
-                        m_loss = self.cross_entropy(class_pred[b,i,:], class_gt[b,j,:])
-                        cost_matrix[i,j] = x_loss * self.x_alpha + v_loss * self.v_alpha + m_loss * self.m_alpha
-            
+                            m_loss = self.cross_entropy(class_pred[b,i,:], class_gt[b,j,:])
+                            cost_matrix[i,j] = x_loss * self.x_alpha + v_loss * self.v_alpha + m_loss * self.m_alpha
+                        else:
+                            m_loss = self.cross_entropy(class_pred[b,i,:], class_gt[b,j,:])
+                            cost_matrix[i,j] = m_loss * self.m_alpha
                 match_res = linear_sum_assignment(cost_matrix.cpu().detach().numpy())
                 match.append(match_res[1])
 
@@ -269,10 +279,14 @@ def forward(self, Ix: tuple):
         for b in range(B):
             m = match[b]
             for i in range(npeo):
-                m_loss_tot += self.cross_entropy(class_pred[b,i,:], class_gt[b,m[i],:])
-                pred_matched[b,:,2*i:2*i+2] = x_pred[b,:,2*m[i]:2*m[i]+2]
-                x_loss_tot += self.criterion(x_pred[b,:,2*i:2*i+2], x_gt[b,self.warmup_frames:,2*m[i]:2*m[i]+2])
-                v_loss_tot += self.compute_v_loss(x_pred[b,:,2*i:2*i+2], x_gt[b,self.warmup_frames:,2*m[i]:2*m[i]+2])
+                if not freeze:
+                    m_loss_tot += self.cross_entropy(class_pred[b,i,:], class_gt[b,m[i],:])
+                    pred_matched[b,:,2*i:2*i+2] = x_pred[b,:,2*m[i]:2*m[i]+2]
+                    x_loss_tot += self.criterion(x_pred[b,:,2*i:2*i+2], x_gt[b,self.warmup_frames:,2*m[i]:2*m[i]+2])
+                    v_loss_tot += self.compute_v_loss(x_pred[b,:,2*i:2*i+2], x_gt[b,self.warmup_frames:,2*m[i]:2*m[i]+2])
+                else:
+                    m_loss_tot += self.cross_entropy(class_pred[b,i,:], class_gt[b,m[i],:])
+                    pred_matched[b,:,2*i:2*i+2] = x_pred[b,:,2*m[i]:2*m[i]+2]
             cnt += 1 if valid_peo[b] == pred_peo[b] else 0
 
         return (x_loss_tot,v_loss_tot,m_loss_tot), (pred_matched, cnt/B)
diff --git a/NLOS_detr/utils/__pycache__/trainer.cpython-39.pyc b/NLOS_detr/utils/__pycache__/trainer.cpython-39.pyc
index 36466ef06e46fea5caf2746555463c42d39ccb71..89e6bb66a09e9f119348c5f0b1082c3258574552 100644
GIT binary patch
delta 1744
zcmZ9MTWlLe6ozMZoweg@e2cw~?{S>Cxz!|1OcR<)sRAk?;?hb*N@**N9MuIQ7K@5#
zwh^?Ud5ID<Pi+*{E)*UBAxIJ81qmq#!2=3Uh)VE)c%Tr<O`rnha?aRk6Jq7t|C}>>
zy!)RSUzq)3)_ua|vJ?2bNIzhwkG|#pD6MD~@p;;$Q{r&@(%42NaF#3`p@dUrJw=wO
zv5%7~;V$lGYEiAxX=Pe9eWu?GOe<#4Y%Rxc*KI{rjP-W(8@maWpGSe*$4SZ0nspK&
z+%L5qb%^kQj2z<r%s6$5rA+^@ODD^UF;z)alibIYn$oymV{V<qi0P9Zwj#gm0Fn4D
zGXfp*9nZ~qSj#yAef3i2069y|DDyI3wS)Ot0P<1H)o6&AG47~lSrFdR86qo$wZc7}
zETp3j^B7R)mQojstXNnZZ)&J0*BUCuZJ^>%P4H$=Nl+<nO*UkB)nB9V@Q!Mvh9;0w
z6#jE8Egs4SX}g%oCg|7Ve72Wn#5>tx?YakMG9fmzoGyuzov+g6HE&m1rB}sZ_gIsi
zCY61n(tU+4h)B;F`ii*FGv9uOk_A!<Fbzzj#*s=82G@tV#<;BYakNctA4fT|qi*)5
zpy#FDJ;OHH<z^*Z)$dpj{WTn2-TG}2SYMBXqi=-T#l3y=#~fg$U7G2HT7tna!A3WC
z0-am|dce{oYXPggU};_Xz|y+%gQZCpfNBsdt*cgUZK#mw?oZ6!2acwsqtFSyCC$R4
z1ba!XgNANWM?k|Gtie`;EDXJkx5-+YPGD=R#@kyltd-9$tI<(XX)k4}9S~(%tfrVz
z@T`+ZYgByP@0g3jYIst)HUS!*Rq7;Ydqby2$qZrX%hZfjx_AsWV*ETQQr6Bg=WIMC
zw`PLJt2x#I8Nw5cu<X2|6Rbd#)yWKOQ`|!pS9KTbyzOn*<Etv%*j<AYgk|~FVZDCU
z8`p?VimI%{8<|AonrvKCu)<PQ9wE8hl#zqYXHdW)<4crZA)W_Ji1&Ao3oRcHevhed
z5U(TNK)A(VzSGwNH|A_R(of_$^@{KESEx@c-*s?q6IUNc{D24|CJ-9p2I5D=XNaE=
zKO-!NUl8kv0^(Oh7;w#6e)8Ciae^5KG37+uMErrk-#Ft>#P5Wpi18Q7TZj<iZ^R_x
zAH=_6c3?t$IM91fGrWxP3Fhpm90(`Eg?I@I+uk2@9t3(|_{6Tkhv{p=96X@zSN4ka
z!K?J5cy`YVDex9fznxihVr*0#a8A201LxGjn^+T^)A*2{_%=_TUf*$6!~a_o*2*og
z=$0^s#^*F>+42m9qd&<=*#?o9S#@H%D?aYv&Z5F>ta(u}Jr#c`AdjD&+gHe{3UQEo
zn1j0@QkZj5IZ0uAxfbEqc9}la#|Z2S_f$?XKE?Ol*ur%=Ex9eOX{W`&@X=f=zPFs|
zaa{EnLeB5bz;ISZNW3?Em{Q>>gfwZVSM(Gfpt_hX#MOc_w6;=skkY)^y!&0cU%b5c
j&2ZUPc86^q<1*ZyqZ7HjaR?jT5T{0>G`Y4g@*@2YB2}c!

delta 1684
zcmY*ZYiOHQ7(VAqnrm}u?q8crljhR2N!zSl8<wg+T8oUS!^=dhy24q8sYLb}>?kLe
z*;e<%V&#A^N+IbS4sYA&bmBl{2#UhgU$Tim#QqVTI#IfMo9}xPU7L?Q-}Aob{ygXX
zuFRg8wNKe>76QKyPJGDDPM))W98<39#8<H@jfu}QUQz0rb-!Bi%#)WglyFib+{W!i
zMa*Oq{SyO(%I9G~?&756W=55G2zN_vLhdKrBO|kziH}i>xESv@T2->FXfbg;?hdu8
zBt)t%ImW;mkz0;|NJJ9_81rc2*qn{o7hne^wG<yG^Ryad4(6;SSQ~ReK8U$G^^<Cd
zTWd+?=9F3dWQ8zXJ;l5#avu)?`nkT;%>pZWrt+31Yv+b0>)<9}L1+zeGq5nQ2sa*R
zQs1h(PT{!LT6-OjO-e!dr&*^;e1s{}igvWpQ|hfHAe&*)x>5~;Y*`-0QAdtK*1)JL
zdoJs=cg1=#3sQfZ91-CZr!R_?)N9ryN{Ds|T$xRaNX92FrmXAb?oEZB6~*2==%{$T
z_Y_?a`!mndHE?M$I!noMQt~hZ=t{K%6>r0@<2(FP_ip#A9N5-+AsdDHu4H$QfPXN<
zZ<Rnz<u0bw6?7KnHZX9r;nTJVd}>7E&b|UQi(`F@k6S=_vsB&+Z4q=7G`I0qKr2@O
z?Vx`rbAYZ+(7(ajK>r4Ff&QJ$4Xqx~zrnoR*knF&s6Wy>4)RB({4f)&5@Y^Rf^$h8
z01mTBu8K4L9dqp}nI<6j+q^?&+o6ZtrHnd8Nu{e4uO%Q#vLIL%1Z`40D2;}dWv>v7
zft^Vn1`d0ZJObR(<ej7BFk#U(S`AgYc?fQ4_ytm=b1cTX7EC-OFK&c~YiSmT4B-(*
zSYlC83AR9#P|0Dq#JF(9R_kHO|HVwL>R?=N9SRCt<<fq&@eVZKs!EEw#%7zDuzY)C
zEYeKCH7)9k%KaqWExsA(5L*Ld`VF{gqL7O?zXPJZiSQPJO+1xLxg2mA)21Ei(>YG<
zBC+cfb%@V)?cMhS_CAR4BZ3!U0>OZA4dEw*FA;u5_ys|aa2??zggnBpVrnp$-$Z%?
z0Y6UK7Q*ice<1vc;79liVIRWZ2>*z81}ErEu{oH_e~Jk+7Lo<QiqML15*xN15_5J0
zJgC+t4h`K$PmA}4#&x$Vd2w^-Jbgu+-~D`~EQ{U3^eWLVSFBKBNzA|(QO{f8i}=W}
zTG+^ur!VcOGT*Hpbvmh3jqo>B%?}Yq^zh8<C@DK()D3ZNcx<i(J`c`kC{*BGjFe3f
zwP8(H?G+cd^46lljLfv8R2>y}$s=pY%q=TqRR^(`JD7#rAX1ohNqK_8)oERV`dy|+
zdKdx6ZTF-V?J^$!i~rzz>Ckf!UNy{2iOR@i+KVzs9}i-$#}K4vw@eCP&L_4;9;8Iv
zoA(=}x=!(A{%-0M@8`q1VP$Cj>->I7`$T5X1$vv<*mEW@Z7SOXCWp2L!_#ykozot`
RiLQw^3PBoK|D^CT{THd(p{D=<

diff --git a/NLOS_detr/utils/trainer.py b/NLOS_detr/utils/trainer.py
index 994175f..4704349 100644
--- a/NLOS_detr/utils/trainer.py
+++ b/NLOS_detr/utils/trainer.py
@@ -353,7 +353,10 @@ def train(self, epoch):
 
             if self.train_cfgs['amp']:
                 with autocast():
-                    (x_loss, v_loss, m_loss), (preds,acc) = self.model((inputs, labels))
+                    if epoch < 15:
+                        (x_loss, v_loss, m_loss), (preds,acc) = self.model((inputs, labels, True))
+                    else:
+                        (x_loss, v_loss, m_loss), (preds,acc) = self.model((inputs, labels, False))
                 
                 loss = x_loss * self.train_cfgs['x_loss_alpha'] + \
                           v_loss * self.train_cfgs['v_loss_alpha'] + \
@@ -460,9 +463,9 @@ def val(self, epoch):
             with torch.no_grad():
                 if self.train_cfgs['amp']:
                     with autocast():
-                        (x_loss, v_loss, m_loss), (preds,acc) = self.model((inputs, labels))
+                        (x_loss, v_loss, m_loss), (preds,acc) = self.model((inputs, labels, False))
                 else:
-                    (x_loss, v_loss, m_loss), preds = self.model((inputs, labels))
+                    (x_loss, v_loss, m_loss), preds = self.model((inputs, labels, False))
 
             x_loss = x_loss.detach().clone()
             v_loss = v_loss.detach().clone()
@@ -542,9 +545,9 @@ def test_plot(self, epoch, phase: str):
         with torch.no_grad():
             if self.train_cfgs['amp']:
                 with autocast():
-                    loss, (pred_routes,_) = self.model((frames, gt_routes))
+                    loss, (pred_routes,_) = self.model((frames, gt_routes, False))
             else:
-                loss, (pred_routes,_) = self.model((frames, gt_routes))
+                loss, (pred_routes,_) = self.model((frames, gt_routes, False))
 
         for idx, (gt, pred, map_size) in enumerate(zip(
                 gt_routes.cpu().numpy(), pred_routes.cpu().numpy(), map_sizes.cpu().numpy())):