From f9e3f87f2244cd937589a7e05a081e2cc45fb8b1 Mon Sep 17 00:00:00 2001
From: UditSoniJi <114995094+UditSoniJi@users.noreply.github.com>
Date: Thu, 27 Oct 2022 18:37:07 +0530
Subject: [PATCH 1/2] created CNN-based camera model fingerprint

---
 .../VGG16_lr001_dr8_finalModel_plot.png       | Bin 0 -> 23689 bytes
 Udit Soni/car-recognition-aws/evaluate.py     |  66 ++++
 Udit Soni/car-recognition-aws/matlabtester.py | 342 ++++++++++++++++++
 3 files changed, 408 insertions(+)
 create mode 100644 Udit Soni/car-recognition-aws/VGG16_lr001_dr8_finalModel_plot.png
 create mode 100644 Udit Soni/car-recognition-aws/evaluate.py
 create mode 100644 Udit Soni/car-recognition-aws/matlabtester.py

diff --git a/Udit Soni/car-recognition-aws/VGG16_lr001_dr8_finalModel_plot.png b/Udit Soni/car-recognition-aws/VGG16_lr001_dr8_finalModel_plot.png
new file mode 100644
index 0000000000000000000000000000000000000000..def0cceb4a58b12edde969c79301f9133b4df4cf
GIT binary patch
literal 23689
zcmeFZcT|&K*DV@Ax`2RyN)<&!sv^BfM^NyGQUe4~dhfkUQ4~a_N$(;h)KEhxBGN%h
zfB>O`^iGg=c6{IW`_8@NoOA!W_ph5V7!H%C?q~0{=9+V^=gl)sRVs=b6c7l6>am)V
zHUvT#41o~Lkdc6IL<T1pz(1GV6du1I10R2~S5e^q<j!gaZV(8K8UBypt6YH%_)@}M
zS>OG+lcl?tnX3iF!OY#+-pSqG#+<{`!qv^j$&urMz<q%Sd>q#9?#_~eg8#jNfRn40
zpa|vhPY8qq@>og!g?H-ul#l*ri<2Kaf|<dErjE5r`S~`<g=2G$%oYf1W@Nmj&7z&I
zZAz0NYWS`05Ti(UmHnX6J>KrS)sw$md3s$RQ!zzx5bHhj4h%ASe~Y5)L0faru|&*r
zQgURnm7}e#=H_C37)yuC6<S)_xM14#HIgXsQR~!(Tm|2d3sD4ff)9ZT!80W#rQ6MS
zA#cIYm#z^%!fy}Bf!qbZ(wSa`D1)B~-w|fwcMksN|KIulg8_RY0TavQ>+b9{TkcDH
zQEHT&naKeYQw0x#qS=MzaD7}cP11vmn!2Y+h`8B&tj>*laB#56F#<fIt1SozjQB>?
zkpI->jAszOh^}~EqC*1kY`0zPzyWeO9z1vu6%!-7E5jCb6%ySdvd&CPn}vVc+CMJs
z?w2&P;281q!}R}!`H|$><QNPtlB{_eSr&+7MO`?m_NO}^j5IkWU;Ld~adh+IWdVMh
zAC6A84!utf=QDzqk|~0xGyhIU?rIzQVk7eL@eiL~b_^cH%EX+HgCm)RNv&*P5-(1M
zi^;9hh@XBL8!Ug>c@`6*^mlIW?qi&UUXt{rlFMDJ*R3V+l2S3(E2i^qr+F~GC|RvL
ziaX`OPku7fwo1EE{h5^eM585c7ksxb^~tf{(|tWNGavim!)bqH`1rF@-fa<U$YS@&
z=FwIBN{I_z5-+ojCvLhfY&x%U)ZdjD*)15!?DX0W9&S70kJjAb`@7i#>v<LZCKgQz
z;(Iz>0WXTJYZ&f5dN!~F_h9iW;oWzbT0J+=zh>l<7v59^kGQ9!zOMI?F85zg<w*v0
z4VK!#fIH9x?2%Qb&4t_38bkD+HBvBS+UZdS=ou5o^ZM1C&G`lDkom5Ul@G&V8Qa^=
z=H}-5wa&BT8FIe+)_sy|eD_{_edO-`kcgCeVWuUFYuFInKr6G%jk!6#QxPX>2h4;+
z7|9g|3;F(b7t?+=HT4Re#v;A8Y-Pstdpe;PryahKi*t;Eg;8oNEOR@Bt=VQM|GsH6
z<VOsvMuX>u_xyXYm0x+3YojIe`$JlCo5x$I3)wJInx%~zY`y~qVLw(DVuJJ~*xv0H
zOmtoYkwue_o0QGxlFIvvj-x5kmp{fhqZ`v3f>7}=7uJ)*x-Gq8vNS10Z|^tf6xweu
zNyN<K>S@!RXTzC(ebeFn@t)ah!DuM|>F?1J=AE6L4jEL?#rf$W#>xL=qeit)W`97D
z57u;=D-$Vj{rYt!6_t<a>6%~>>;#d6JC}ZC*~;qGTxD~J8@HMrRNYI*a&s4KImulo
z?Ca_>Z-2}1{l||bNp@!D?vGp=yZ%p44mS+S&564c1xZc!mU?@Vg(FLi8+?&S5GjeP
z?U;;Q!oyL!i?hh1nevXS=egde=49d`BJu0%_SxClu4%Dv-*QSwNc5v8UAy^D&yKKF
zPFS`M7jb`_(8bwS&~<5V=ccVs$;po`Etldtz!j$-6N9VWEpuRY|4oQGr=X=3vST#z
zErdh4&YJ~&A;zIG$8JAeQs?{X`1H$y3o-(~l)vn#a9vcl?ETWdoDtObGw#+NGSJ^&
zIxa3wHD+bSIu-#_Qu=Pud&7F@M=LB%vdlTFHN!gwcBy%Fq==r5&eg~Xrmo)Qf(#Xi
zl{*)g@!ikV%8*U<+pKfnnohGy_ldyayj@*gcl~=m+*DN4(vowvm-3}Tf>j;alR~<I
zX^UZXJ{&J|FD%(-*tv=#3(m^Q`u_d<2k=?*S5P1X)1Y@x<m70JTU`7I&Q3=cIxoIZ
zBXEb0?;SXXQDjuqPgsxvdHm*-*{SzfMcxYf&$c4(_tZE7Rw={r1nIX1QCFj3Jp`Ao
zJ+!`pzdVn%(JGsv2y*&6ACi*PuuTC3E8~^i;9HprPTUu}QnTT&|5VwJf20&_{O~kI
zbO$#d>$Xviru&Hg;)|aJd@yP>%n9iM$?!L?d4I#;L&F>eZ=I`qKvSbQbt5jjWRtk|
z*wTBMHQttXYdZyUjfqLs(D24Bm3JIGJa522-?!>x0fR^S2&}hKHk4Ug1nt$U-^<L}
z#>!ukir5a%fQy*@U664&0_Wi5wCKS&wFO<+DJQ2qEN|_?`oB~}FpEv>tQK9lVXIhQ
z`pA2;u1;^q(yFaF<j|<>M6ijhCGfr`cU~NJm|&lHgqDt${E2~qkSU~Az29@q66sM2
z4}EtzrQ>KsPsq!06X{7{i;fu6rwn72Y!0KDpn#Pr^b?ULjKS$X?4OlD^!ffk-b7tW
zUwDNP_S!LDYiE|Ai@4m0s_Ann<{n@mVd5(@mW*MzLGe6)Yyj?^*o0z$L@Tp%(8e`a
zxbsXphi#NFUnghxhv*sW)Y`qWoM@%i<1nfu`Szl`E-X#br(o)1%Kx;_eKW%SEt*Jb
z1Dgab+469<_V`qLWROk~<AtHN(4X_e9Az*joYNnY$&au!PE9tkPPGJgo`!bC%eb8x
zR!h6F#<f1fV~);_zBEQ$Ag*QUL}u^Z;Vi?&-<0q%QaRebyaN_A&LX21Vo4j6Nudai
zr|QCTO^Q*VX{;4nx7?~F$8-87=x=^GM)#!+6fsSDv9rk9a~UKEghx;XOg*6O$D^XS
z;2JZc$pEooIL%g*oyb{w5L>>c%rT=Qr$yWn$PjAHhD|K#e$=w(V-bNZ2z$R(o{<Im
zi+p^TCTMq=S%dTAdNa0UgL*{}tu@9{*y^Xbxp-Wq_DJ0Snq^PfOpR}K8m}I;Ba(qF
z;A)>lN6iL3&trU|okLTz1b2R#ZD%suvy}4c;&JYm5pw284XoEn;hv-ayGf$wMun>Y
zypfQH$FtN>#CwJSx(F@#;)LOz)A+lAp67aSA}m5|IvDA<VTAPE9#C{f8yk-B68`;}
zJMe<l#G4@hNr+<Px>2PiqU@Fcbywn%!-<kQJw8kE41w`w^z{Ez*zj2+l3K`H$o+SA
zfS>YiHvSGRV&<FmU<TVKn^H%|Yo(QJC{YK0v}jz#BYc+ade;^?n-goF7?Tl%a9iwZ
z*{Er|727;+b%F@sv<Sz;$E#!0$n9=H=RswJvEk!!6#Jd071{#Onu)r%3{!bV**O}e
z{G8{H{0O3>em>62ZJv*X_BMKMWCmS)^uZSMdH=D2wqsDJaMTSE(*-rPC$Im>xq7bj
zRzU`g-$%IId6`KK7{M9&qhptOKC)@gl%r{k%gnUj29lspoQN&(+GgVrg+#R4U?C~E
z;FwBt;z@tuhmM!$JD|9R8Nwt9`UtAZ<w-Z1V(2nbwDG(?95rK_Bu*>pSVh=J5yaVE
zfFFk7gW7;I5hGuoDo3iJ0!7W3cx{8S=B^T>AQ|c7+eV(rz1<D7BwpnfT83p1lFMZ$
z{c=vksogd`>fc}D!8GCIY}<AV=|BaSkqa@1m^W`CDh`)3%=%bXhC7Y51aE&&ht-ra
zz?x1Rz0PE`z-^{DhOXP!9_U_s_|px|NQ+12beV@6VS2D|mG|}n*nT<8_9|HElPI3&
znLc}sFlY|RlN3v&>+^z%)JIz4&X9Vq$l({1xlT=m3Vob^u@s$Y;BmX{5{Ozf($1${
z6N5P$R@_6SX;l58RA>R1Nu#GkOfRod9o;v*$XK~ie>qyG!FLp(F8J{IRT~1V4uXL^
zc_OF*rikgq*Jrpj>T8f_#+G6OQTqm0t`|kRRiZI$R9i?Z#h#%8Ccf8i)dunF09P1r
zGV@8x{;t@~7Dji-XF(c07?q6Qgo6vC-X-D|j1%<hWL0PB@wipEAl+H77^4z_N#J!!
zF9F{arG*9j=$)!*#*}0R92xVDyY5p=Asn()X?VA^tG_Kz3qPO(OKeJhsV4;IeND1z
z29B*Do|`M)@S6O`RXJao(BQ+ox9f(kY3I+M00e*B-**R*!hebdVBNJ*gB8DvfU%~j
zfF^k&$^d}3tkQo^i9Y%AYSAz>Y-;h9rX^N3MFY3j?|go|eRgz(ZF9==?aB!hKlv<l
zE%U>hcIF*6(=_7mjpmHzL>~N!AHFvswN-T-?1XYygzA=7Y2QC~VOg2Nw8DD7h{isf
z)yjQ^ACu4b{f=6hoyKHqrMxEUTo?Z$I^%{v0&2*^8>jo9ddV9X8H;%IoBXHy17?Z6
zc#@NaXRC(P5ha`<Pb^Hq%8ROchzZXzk%0=8yA>RiA@@I)?4B*qW0Knvk=6|N#{8S3
z$kwP%u@}36)01^V-eprk2Rhc+riF>SocO?yHyy(YLIvY#1WjbndHGx~J$LI;FXFk%
z!D^FZ9yCWql>N#~J5xzIc2LxzVp_hV+*oN7S+jt-OSgi5ChCRnnfW<J;7)DGlSBDv
z#_d7XZHyCQhV$Vtx}*nlfNrR9UK{E5WF%m2+8m|w+=>u&C_d7X3W)kF4Z=>Gw&j8C
z+D*G(m!^FwsaZo~SZBLI<=p;M%%FVX*XFpD1FgCW$)xwKb*z(xN#GfSEgF$ES0)6j
z(Y`HotOH?@^5i>MLvu?@rzRy`V-zq-+H#Zf+I;ft*?@~Gy3|;steun7k&>xZM{k_1
zb|~M#xUtmOL14-Y+Y+yrP%>SyyI4|Fp+ZGj%@3VW#|}JeD0$w2=D$jA6i5##c$)Gf
zn1JY-r2E|>-I8p@NNV@rzrq2Yk<O5_8|C}>@nZ~&9lDh3_HD3C>kD5)1vo=CBiU!>
zqNkAlEiI!&wZ0<$=XE2UyhW>{)%+`=p{8vlQb&K*bRQZj`FA)G&V2nRXJx8EEa1<6
z7C_VQ*krx{lm?J(;_@Ciqk4&9bqotRIeDsU7Y}$?+UBDrVVhG90+XS{qr#s3GG&1>
z{<xc|Ug>b?%JoEJ%eEGId4!j9hHR5#Tog{3wYj*MPq$dV#Z>QUNKpHa9~G}UKN1?h
zocsO;4?<rQ>)*#Nzp}Lan4Ijh;G&|UGI)cJlYsv|?fKS4ag>{8%mMCT)nZvwZVFwp
z7o*Bv?F`L(3x{%Y&Tz{F7Zw)2va+IQV2EGyfAQjKXJ;pOz}fC%w?Ts^e}Lw$mCb+S
z&K2`&s1l)u?YEM54#Q-3v=G~9q*Sr79hdQ`$Z6f((>I^l0%`GhHkoxpyw7EM_kqit
z(iaiCAH+;X&aKN-*B`vz-`~I00gBtVBPO)um%*ZmqD2U9bqgx$GP-ZtyIolJhUPyF
zm{Oo6e;9cG5=eg@3JMA_Ya1GmA4`<#ZR-QbX|+0B`0m>Mx!f&!Mn;t+TgE6g0t<`r
z1czIyt=J$;T>}42SvK>R-J@?!Xm<ny2Z^rTxS{^!$z?G9b$#5oZx4Xe(>ERF#L99C
z3xC2mp{7!vUX6o%Oaq~g6jYux(x$UpJ;G+IFEHFkZ+Ss>`d)LIO?t(2AX9#SWiV{C
z1PZDZ<~vW409ND-I5{NFwCYbEE`cWEa5y7)int5sRJ})zVYO|OH6jWZm&$@$7xq4N
z(Xaf`H}cIQTFUnbQSS{_YzX#Jb8W5ouV25e-8aAV`0?Y1cBAhk+?JlfO=KW+^z`J^
zeN(e_W^E*l{D$Aru%M0TJJC@Y;_g3vkJA16JnJA)?_PJXRkAhHUuC*NyYq&l!-WMV
zmR;QSe-Dj3W+z9O)hQ-e)h_`^Ja1dTQOm8^M<d^w95SL7o?@|q!$fyEwY32<nd1BL
zt{iM6VrLjPe{rt2sHv2@{~Vwyp3F3<l6T#>9o-4ynRmj)HCugBSqj7%qFPG1%^-e!
zhi^>TA>FHO<-NTJOUqquNnbQ>Ovc4qEceMUk<dh>I7PVrkt+m&@((s43RhxKnQ5gp
zR@2X>S7<ITgv#-BXeiO8Qcy5mRL8V={<!Sio8X9@4#n!|6@HG6`fLnV&72oY$6J0H
zR@+!bg<?@G2;HaWCvRZvbmljOxY`3sK@lCef$@V+adL7g12l8m)e<&PK+|~9BonwT
z56x@v61>gb5SyDj8Ef2u4T9=Fd0fun-Oazfk~7d)=M*GJORI$kkf<FrYyc_6XKcH-
zYvM<>i9`qL=DlTo`Xv>213+N1Cz^=a8A*>utRuPe`1lo;VhLjH>{D}IaoRXoU<jo^
z*|yhep^#Up)G33?@h;;Z!)k5ZCpy5Coatych)VCh8@WcuLx2whOTFn-4j9AxOF!Pv
zf^hRu!++{t+$Y>DsWqA!1nJGSEXCRVq1miR(-Tivjpk`80AQaDVu%DAAyf6^>R1iI
z>aF1)?*a4$d(Bav@N)4}^~qropKYcm?-a${AU;@AgqxH(roo~20@cBgbmjIusa+F*
zX3mNuh&mEs6WegLLQxw!BQI7Kafoq!^Xc)SnF(@2_IsLC8$dXHQqFx6&b=aK)^TvC
z@owM=6<ZL?`7sErJ9x0Fy>xmeW)vv#B;DR^85S^tu;hhJqxtD22s+Ar0==dzDCIV5
ziDb{N-qWqhASDKo+K|HtaoMr9r7z%I!hOZGtvw@<LPDbHOh&!yxz_j~3I&JGC4%Xc
z`{ejQuXbBOzxE0zr)eL{N#FpXI?qT&<!lJ)WIGQj4I;hy^xU>?hOp>SQ=klhd#}S=
zK5~g-XH(r~Hfl<kCt)Qp8UUTnRM`(4pr@blBX=DZNbWGvoWMVj8Pa<M3Pf0q`oir8
z!n1N0<@HMP@<NvHbnXfMgz>+>(ZcN1JX#|3icCqxkZW%mxgi&dHgR)$KJRe-IzmSC
zvG>yTwcm}0hSr1Gucn3Zw5wX-vbd1%L!$tZ^>Ly!$(G&4fhthq-*Kh;{adtHDxm)0
zFaVczaEVd3#E|hV9eqiK&s2Sx8~0?8T;u-7WlTWvRKo@wo^%a|d%4ww&VRb8z8NdG
zIdw#V=kT<@NyNPWLd9auN{IApL&Gq-4xD)tpb9F##ix&7(t0p?Bcq+t>$GMT=Aup!
zw71qLb7EV$Gp;t&9d~tA97mNhbjC`XFE?_@PDk{>>nRLt4yR8piG$S5Q_Vt3LRGOv
z6l|3dE{1(Q@C=L4SjZiP*UWuJoth)Put8hnbDy)agzT^FuUTDWO?omH>(AD$HHgRP
zyvPuv0j$yuk(pRI(@F7u6w}ygj2E{0Wc6(C9{%9*uCD714*S-Ar+?PA$6gw)-?V71
zut1m}^~*U>V?T0DCwDA;(ehr5F9Up0^o~0LsIff-Yc5u9`CT48sr9R|!cwlok!_JT
zMHvO^f9|0ZZ{g?1aAzMPYkFsXL&|YZi873ni&J8JX<}pAf&E7-Znt}qIGsrlk=-|~
zU7A{wAhXN@Sg<R#UT~|@vlh}#lUlS%<(F`;X>zL!hw9^WpC7_o^SjF#jG_eT-mR7B
zg!xMnnveO}cjf+P!s#_U<<=Y%z&7bJ8&dqYzpI2fB1h0AbSMv;P_e;Fd52EAhNn+@
z&QjIm?Kh20RA1$uoTtl-Dkh|DnBe{?)b2cGgVq=Z<bJ+N8Nwc9e?h=7(i-kOZXt@^
z%8Im~ebb#PF;HC;pKpL-TOHx`+#@*Qb($2X9KZ&yCEG12VzCs#?b$RRqO9t^S@@rp
z`?V}kiQ(#X=JXysC=MK$Yxm@Xdiuv2IF|)Lq7AL!R$AU`B_H8W&f91Z(Q20YPt*BV
zS7)f0?c2BHE}9R}&<|9yw#@@Bd@Ow_R&F#{KRC_-+r)Z9w5qwzG=%@)h%+FF7s41V
z`-n5$Z6-JzDTu0y2~QR-`6N#`nZPefyFVijeGDDQQ6Tm^8!g#PN47_o)NGBxkx^un
z(i;=wE-l2PZi&e@NZ;$%J#z~hwwArzS1-}Xt~7Lh8_I9tuNgZ7Xq!OQEH{5bF8XQ8
zdPdvnv2zzshn&;MJzny8@$LE8c38<tGZ7+Bt3Z<xf_(DdW3Y~~Ixt!8t=!reuPh0Z
zCrp|jGpRko!m1=3Zr;4Y=Ez?9QpDPn*WEZ|YrQgB%*b#;`rvc#6eW2%>Z<^%(G)J_
zJ#gLzZM@be-sXEXCwn(2us>bM;_ZQCUy9!TK+5M=d$)y!9ahkvit_fdww#EY%?C-<
za&cg+C((25qrZ3CnRF$e66za%aR?H%OOSCy_vwajj(@s{i!-a~AZglN%+lfYnz6$$
z+M)%Hv4yuKVm5@O!3i7%jBRfoDY?4Zi+ioUNI}=W>g<GlN+PpLvuJTSv`_KVzZyo;
z(k+NVWrR~|<Q~2F&rPfCtWQ|DEhY-qY^B1Rv=<|T#k@*O9zIpCU?{z%qMtP%dN{DS
zpdXvXds#)bAu^2Uv1R_jDZi05(oXIaWb~zr(=0^?9j7|i!>!Crq_bj#en{V)r}o}R
zqq^4WJvgd3fRuXd<i`)j%>(7!EvNVp+Bi;e=jMD1<cR=4d6YbqYg?!dDa&J5%*e}W
z>2OrA=S*-fQNyk}Wzf?v`qtMU<T|%i0=&GYQzP)uktJ!C#-4Md*Z@&x-p#l&+UvWz
zyL*V){3zhmWKer^tR*bV*8`J*E2*Kt_9dEB1RXJ%=x=A)?vNco7{ks(KRoj2mpPP~
zrDLP+O;PMl@T{^WUQV}L-uw9Z)${kGC6zoXSbw74FY!+>c~=eV=p0e3fhWCeM_V-g
zGKoiXL_DbnYa4<^Qw<Jp&Q7+L+04(Dn2WZ&aAg~O4_sf(V&|*T_N9RNLwM)9E`}mE
zw@?nHv*%}`NAur`C@!Y)L-4PsP8e5oN&ndgenz9pA1as_20Fs0DTdeF!&@2o?QbsT
zCAR)}Kl*rYqek?-T31vm%i5vC@K+qtZhU;Zl~Q;AXy7L1fy1xcYs6Rklf`ICYg-kj
zFV5Jdk8{&54)jv!ljF?VMZxN`-rJ&Ptxy<d7Go4ps*i|$X>QNuLSC9WdQ?MbTE0Sb
zJ7g&%sJK9Yhn$2OkS2UIr#q24;reyI0#d$wITE?|qKq~}xUXy*ohgs!*;!(UWPJ4n
zwU2jo@8r5!@lFjQ@P)#*ox_xEks6agRR<bI<J4%@_GZq9o?9V9*7E{+v<@$$6s!to
z*}y=xO!f5#;TJm{hg)R|#78hM?vDwU7xZHdOtNRgU69Wl|A`KC)*uYN6&ABGzbS~K
z=I97vm5QMW#~{Z0`=y5@x$uP0#vFl`oP&(C5074-ZHve1E7b;1QKl{Zn3n_#O((@`
zwc1?Xiwr9x8q6|&n3u6Fd;N;mU(Zx)9n%gLz-pm)iEeYrR3wG=1udtE!nc*!f+o90
zK(%y6ll}T)YV8C{P^Q06@bS@u0Le-JN~#S31M@Vw@c1vbSc&B{tD?ZX%gGY-j*Wwx
zsGSJ@t2jV?xvU@Bw6utBZc)5umDKlg!8RS55w3}(K2V-&=HlWeb*%3+HYqR^fm*vQ
zC8yV&S6GN*h3=B+!Vzy;e8#K0eJ>t7Sa<U{J4pArP#E~HCj^pEMBvb`x3XGBU1jmS
z4oTb11eith{gwHQN)NluN3}j~n*azIIAK9zfZ`TDjo#83tGKBx4e&Wjw}{=$UVll!
z=%=~489>7I&FaAx&r)aGHn2gJmfL&%bv4PnH~*aLLiiX1f^S~|+4n<|42IE#4{CW!
zO<Vo;?J<h3n~Iu^c8dRSakU&TW0KC}^;)em82G9jFbPsu94D=p`z5<0f}<NelCGxB
z>p#Bc^!ev_<M>h(i{xi7W7NL;+8G9kL6kHcTlUvvLH@%uKjui19?#zgOb@TFz7G=*
zmu^Xg*>D(XNz!H_%-8erK?sqk|H*tTBG>Qtuk6^Vtx20%jyrdp?e>T%(0y#ZDbTH*
zK@S$=J+&0kY7;efhZZ2vFoZJ3(W^bFqLnM7x_<RsJq|2X&FPk~L#}#@=<pWhXlb{A
zj`D?z`PgUG-wsY0tmZ7!Bt_p9Z!cd&P=(#onMs5d8#Z>Q%jilrAFu>X1l3&z=siw)
zapBR{w5ZlecJxe~s{a(H1SfZg`d9mD`!*Xu5Iv3s<Ol_(RGG2;zH2VkU_9^g^gjx@
z9N%<3;Ee2|Jxi>F?^x|J27kPFD9yy-VdH|$q&P2?ZdT|SEnIF?BW^jPhOMql2&z0@
zTTVWg)%+hXz{tq&(cZ&gBv2W4l-JsM-~JvdF%=pk&3ew=Uu{gNcyWH%(b<<ax>n<q
z5pYKy8I<=LM%)aP9&J=$Kx<pun>TOFvWl9mWj(j1djpQ9R5vE7$!TZ`0G|Lfs)E;s
zyxP7FlVq{7wnrI3)9T@mdF}_vUS_pDg8B8TmjpYch`x;2D4-m*z$Z+MirMKt{iI9e
z6ubC<v9x}!P-`oP?DVGGQEJKQ)2G$x3qS{~aiDLGC7c1e%QYZG09M3e<eQHB%0P&K
zas3Bxdxvj978VvR$BT)T9&5UmJxMKp4%U43dc_L;DpnUc0|lfow!52Hr@ZMu2YVjT
z3sJ06GYL?8vRXYs<{UXgHL`ctU&;u`%QIeqqo!pp>3vv(d8P56*RI~f<Uh^*3P?TE
zm1*Q=|F40pduPwpalQ9u@ahi0$M+p4Yi})o5p__nah!_sI%9DCyffdi*Geh4eX@mm
z-sFGGi7e2ieIw_4?!()!TUWvVt4OaO&3`X@-@VA7(!Y-1Eapx5bj3HFfT@aX33oD}
zA$S^!Z%G&Z2<YNDmJ9lEx%SMHnXYMukoiHWf$+uh7S>4-DR*`W;6dKNg9{O!(s0VG
zm#PgxKdu8|`s`@s<)3pBRMf12fkCl;d1y-*>GFR|ioNUOmEb8+9Zh+$X9WJ?!K+21
zP^ZQ64z^pjGwR5blam2C!{(D37dKl6Q^>svm;g^}!Zo#f!n6G@_LIK9wJB9J62?@H
ze9i)(sHX6iq-#aWdb$d;FJldG(NSg2UfJ8E2!&crR?JI}V^?VBCs^_|z1TS|x$(^_
z83Fm*-06c^VH>g!0_qup1c+y~-T@_Wc=$y{W#t>md@|dA(K<d&mUz80MJyKM*zg|k
z3}r6!Dw_>kvefLdJ@i`Aam#jd^Yhnb{cAfgg4#D;S&R?Ei}`B**me`;jh>(BroZZ>
zqYThUoH)&XIRd8vmDFq-qob%3!j%V#(2tV*<5T)<MN<9w-Bj7B)o&psKY3P0je#01
zmm^@j?&XAG#3zn&7KiU-ezCiY4()J+CB4>{2gK(`az(cV^CDoE3AQHdKGl1y0eXvW
zqRPf?;S}i4y`OoW<ABT>>-Tj2$9oCyZ4M~HUgFohnC<Xg6Yp7Kz~1fpXEgdka&mHr
z>M0FQJLt9a7h}Kal<XDj^22IYuI(+ib+AJ-kB_Y=Jya<vE7_0;2mbwP8@E%za=S!^
zeB<)gfYTk-=`F7x*AY4L?{948_#Vt}=@rgRy<bPeYp`}(CeY-)Nw;v6iu&7wO9D{K
zi`4ul<tEq5v;}J>&5yV>UpH*-8ex>xd8i0Z9@Cf}td0P-Nusnykc(@v)<iyWqS_8W
zZH?NHK)mFVkVpoEp2FVP+C3ncmX!(hi<;zK#qFqqva&(W(TL}A(B4u@UgDChUO~*p
z%ZdE=;Y|trT1~1`g#{BHF-@bT#)zCBZy824Rw%-O^y2bhv1`2S_h3tzfpG1XtQ3sp
z<gkTGOdq#4-0Atx-M3#vHZo>QY7A*+dc`ogr<!S##z-2O+-+{UN2i>Sh=>S4HYoyC
zcdEfl&Q<p2&5*Uy27!Kbl^r^UMiv*Gn#vM-F%{+87C|*!YAn^i%{>RFW+}_UuG_DX
z?GBEna*LY3o-zrnEAe+2BFWB?8QSw)EksNnf7F;<_S{0Gx5{0-MqMCg7lD!B9A6XB
zAERNOwNEk}3GA977{;~X=ag81ju5}i^3Ab{YO~ehxsr70#j8cv#hp1C|9p#oQ~0dK
zlkWFKwY5w%r?#g=pgG_GHIH`})85c?O9H2Y8_DHsGXUNUj@Nq>&5Oqj<*3qFZO^o*
zre`E2t-Sd5qVqQvg{llZ^IhKSm-Dh6E&eTiF#Mc2upbBrq|U#FhGfo9qCS6?4`U%^
zDbVHbc>MIh!)_Pw9wKG_J{{wTUUW(4#*BQ&c+-@NZi#E^>q{NSGejLe#}*wngRK*n
zW9YB{7H^A`(lon}+xfG0_~*lwj^lH`{%Q&C4i^*O&Z#ELUJ9&e+yhN4HfEklCK%%4
z;*y(}ci;Rw8JNM=qs7<VmU^<LQQg<*>FHzl_X(6#Rre0iFk$DJM_<I8Gy$dP0~9Bp
z_3{q?1Kud__w$bozr4409G_#Gnyesh`qP!ZHwtZMFW;v?mn{QFgiXOQg1>`;GOXTX
zLwtgWrkOiq%<+aDdSWB#{`Pj8^TpY-!e4LT4fe?gY@s8IzXw{}9H=QOOge!arMdV?
z%eEb^2HA~P=4i@(oU8o>)E<;F#<QC%A8I^`o^p5hBad2fJ6ykdS`F%c=`N6vLeI)7
z5bO?`F%75TcJ~brR&{^n!BCAFb-DS5HeVbt_UG9B-_ypgIO&4_3s=~DNOOo*gasZi
zKvAQGFQcKcvVD@>l1b_oKWb-5cr{!x4%^dyO^#0oEv(ui*(W^pQ$)w>4699lUP)PI
zV%rRy*3oN0!=N-%T5^S6n%^n^bL2O0R%z=LnVwXeSi6xf*3DJ47gp}f-K*~|J)qv*
z`pQr2(Dhr17Lhbi@&wvjlf3WNMyof<H|x6(HCbgR8LjxBB+Kb*LrZUQy#E2Za~c<?
zFEI!K!&f6((>@)>eZ7MFD`v>RsRpQErd(%ND9|kn3+%r#@sVv#+Dz3x&)2VDqJ<9}
ziCxM#<>Azxu%EK!+GgYr{i&8Dh3;>~F7=AOq8OE0a#B!qSn--y@!k;O-K%&R8IEcS
zU;wQ1+NmW&nOdfPbyxlyE^bWr7`*A;j}>dp7@d$(qjAym68*s(2mXrX^vYvrJrkSg
zN2faaNeu6r)&wx&^Kg{u7*JQ#l?a0gGu5s?@DiBx&T+WxvRSvinAoAm!SfmHCnIV6
zO|OPL>|k{<07|FGg{f)Fee8*4_q*8D&H|GE)|zgV8U$$WJh$#>MKo+7JT4^~nGelz
z-imf<sd@-?hm7t4I1|tD_-XY>flr}!J1ZHpXw`HP3Dw)dDCtHVCyV4IRw;#(!*nDf
zDhdLati1Gjeg1_G_L`2GDH11Cz^;Rn)+&&Pcq>6Qb%A+z8S}?}ZM4y8=$CzT)kx6>
zr0AccAuR;Dm^(MO`8ByGOfid-(*($;+HYB9?GHk+{zGguZ1!eCwYA+fp6g}pq%%#X
z@%KDay-cRz;|*e794s;ujQ4@^<iGcxl#a)<mxy%cbZ@qz`)!a6a3p*Jqr+3}#Qwd#
zg$brF>vy0-8O8>nAp@;FmaX2MXUy)NPQ}QbQU>lsBSIz<GD@ZigvHAaM)=bP&pO55
z16g@h52~Z0f}UN!7N>oei&)YNvCYKhxOXPacn&yRh5|^x^z;49kz8{B?GR+(r1;T*
z!#~KwFI11k>`2f?&l~ODMExWO*0s+Mn8g%QL`AECD<Da%pF-BJ=i)5mJ-^p)bH8Q=
zpn>W3T(f6bn`>8Y=$l50o>GpU=&j#(q@|UhJ%j1yUVW?<jBqHSc0K!3_UcWzeC+We
zlUK+hFs$^5L|zU_ajb}M5Sui$Frrv{FJ^D|N#*^zA#g0{;2%kI1eNnf^%Z729es0i
z)8E?GlVfd85ABoqd7F+ct;c@}qc2lb)CT|@e{&wlRyS4U!RpF`HS5;Be(jxa41{|@
zK=rZzS!RdIx%5eFk6R@K{5gswrsTNS)#ik7FFbjw60^Tv&8zzw*hlWKWdw%$ZB=I)
zTmh;o=8KZ59pyvXI6B}@m?Mg$ZsCQGMJ*@Y7v<XeR%F_H1G)h%{1}NHaG8_}++SS`
zFIBeh`I8gNXR4Vzmp<gh3HI3Q$DDPr?LFbS4n~8A_)#3)g826FGU7cZj0Pk<iRCdy
zCzwCM(-hy|@9F7^q4ysSGlqC~b0s!w2#%jcvLpTkbMKy?=sp*^*LB#&0ImgCDYX?w
zi_0Yf(MCjTXbhrZ{>OO!jq-96|M?uhi$4dIS-MWXGX%Bs)IU4R>mq_q<Bs+!DyN`>
zGW#DkB6)!Xp+NBLd$+c>y}D5}<0h3bo6q;Ior4X^#YW?pfKfioYKb<N*OmSJOpa@E
zYy!2-Z>7A}EG<l8MBMUA0woS!oRHoG2LU$K?;6x>ub0!qg!`**bV^9*axK<i?I)nG
zUXefKl)@kjHk(79<*1N1@@%e-RNJ;pRBy7$GoeX3Op8qK=?NrzzKJ3OCYw11z?A@W
ztMl}5y}UvI8$c>-`6!HpYwQEh^O+-TmaPu6CI4fOHAAw?RMnWAHR?d1cit{Gy1t1*
zy~1#D;x*><!^K2|{Er0wX9d(x(l2)>Dt5$hO@8Qk;=TL0LC)Km+(Uvb=PC0MTZ@)7
zL8aC5=Ghv$KO$IJDOU;1Z|?yRs96^PqCEFane*&mm7N&3Z@mOHgBHXvFzv@rBWIgV
z*3W_CWDVX1>g^}e&Y4>iT%2@xc}e<vwUv`){*-2#?Ll0CIac<F4XGtPd%}PaL|smk
z9QUrZ`w2v`#%CLbz+Q_VdZMNi>e8kO#z6NCilKb@<6T+%fu#G0(^|8(U;n`1{ys^v
ztUwSw!+|0Za|d`LfB{JRNi?Iuij*^?wDA|C@9Xg6GI_Abz0lY0w{|?<zbj(fT#zww
z$lVsv0+MUJGVlv^V;FuaolelmJxUZ1o%0>HK#Yk<H%2kJv5Ymb{`zHK1PpW8#>>ac
zH)W<uUAp;|?cXI<S)0+{Qdat20PgGUN(cMwTi)UyJMB!+W!;ec2?^@cgT#@JlLNh@
z4LC|cB|?V#9))8g?c7|ZaYHDa^D$7LU}f<LS%Uyze8@`!;fuD|$Ocnm@#;||HCTFp
zLG_WoFLHqUP~%zRK$x27x>!uP14r3cudl(OGDr{!guG|N5ko40>Aqb*vN2WVkKcep
z`H;D?bnJNC>iYKX-DS>Ow;pS0adUFEqI=`$c7b7{lPh6%0uc^x`r~ygdg0X?qdRMn
zuEB?Hjb|SmXnz%7Pj!nqLAbm;85hHjseg=>^E{#;94@jtS^Lp0U!+^tHj}{bJZj*)
za-}DOCQZt3k3B3(EjjPs6x-cxz%RBT3k#jihTNNC{8FM|j0Q6p1pgLOnv2Tj2XyiE
zhc?<}J99lxK<aFrSs$-E8kX66;&n~>$Y^DRayo!X05wTL2TWS&|A7GUH)CFk0NaOy
zi?VasL3*0C_{idnygu>^Ru-eQ{15trtI)^MQ#HcG=i%;0d!eCxW5&(|sX<K}7XXXB
z__mOYooyya67Fc7S<b+`OuX~d^vw4}Cg_wFC?`FBzy@>kKfu634PDBZ9>Z|;mDzwt
zZ7VRLs*tSFh-C^>s^yFF>$le!sgL;5GVU9Nz;$@<7WTJHBN+(`o>n549|&)p9-!NL
zlDURRXyYEJ|C{Q4cD>vor2*nXYoC2HEtIK3xInZw=VP*%FV%>E1dZ+JzUi&#o448$
zIN&A@?eTUayB^U>O6np3L3RvfmqL&CHV__mP@}j{^Rg9{ZgR3qlL_Ci2>kXSk@8BI
z2l`mbarIZ-oX^C?QDiHFZPdU%NeJ9|7h+E7Qzo9kV9SWJ5VlglHX&wm9@FCvp_m>8
zU^y?|E4l~RikQ5>vp=sRBLPZnYYnH|3)o!f{6tPcA-O5;y=_X4&~NN>nIzc$s6ny+
z(YRse5K}_!X8X;-<i1n-h$Zl(?~wkV3>p3~dRkg}ZEfw+Ta7<YydM`<`Be52@WG}%
zhsEF2xs_yhoor$L#9o}yB=Cn#RBL3GfSJqn{r}{^*s==^4gIQ9q+6<Z`mX3`D?563
z{9D`l6y?R4ppg&P`k2{463~jKh7}YwzW(cQf`q2Jx{SQK#T=rdqUhtXG5rKSgO6Kl
zhnrIxNZ$jzF~9cDYLP*;ZG272J+9tr$iiIoUO-p2vS^P~V53aaMIh$;e0H>YCVq{|
z&M*-O{o6nslbp-Iz@QAQx9M=3U%BL;xV0=nCxAb0HWG~K5U8$19Ve5vPVi<9VD?@6
zpuK*3?*b$MRrYXTsfm^2kepy7N9}fzOHl&jb(b{CkeJ22r<6Kop2e$gF*{mdE<2~*
z8<d->z<1((Frw$P7_SFh&%ngd7D;nt6xQAlcrkT2M@1u7Dg|_wqA$B^@Iye^T$8DP
z7X2E3Og_{KL9q!^^5)H(9S4Vbxw$N8-lU9~8Ph(wi&OeT2=#y2I8F}(<A1e`oS;it
zUxyPo<AmO^*ra%!{wW%9J>P$x^ziYl7U(*O1Z2d(z%$^hRZ~;@>4C2=fdhka<wneW
zneTzEZH8%`mA?Kp(5_HN&C?!p{rv3A%NoN7l(^6tDzWB?YJg2-?Q?*ym<HE-Wjj+o
zY-KQqv;<m{4{GTrNkSptzbpMY+@O5apDt4ftq}koa>!+N+0<x(`Y4t6H^49gNYU5|
z%)LHm2k3xfyQqSNh2~IV&x^U}<cH)Sm-__id~3FYx0KXO=r)5V7V^W1(A3C?#m2@4
z7)?TRb8q8KEU6N1U-jT@5J1#jB5jb8_5wVe%+aKxtLp=B4!k)T80FI`K`z%rg?j%I
zL^aD@JjzpF|2p->Ptp*MWzcJeI{nQYA>bTM^b6uK^J^-fbtazFU-dh-xb~2&-rdZ)
zTS9v8a+MPeHVAGaM;rGDZ)Hy=y6c)6w0yDQ#NG`|W`*)$z?7n-1T34UgR^a;BczQe
zl$4E)^N;I5UIm~IAgOb!MfIOQo8J_-uEc7Q&V-P1-2UFNK?m%iTvs9O7pGFEDb9Ui
z#ZV@tJau3%m390C8iRQ7J1%3-f>gIwiS=>M<E+3#5hH%IBWPeyW=)H7V}zZ1lE}-8
zo&tcw=|~Yg2TKr)JUh8gFMF*loSlmKPwi9iLmuce!X^6b;Ub^=EDr=ht_d}FY=9&2
zgo9I~dNTS`CVNBHh5DpuBt3RL!oZ!Vv!li-G+z^uidT6{z;^^7M5N!gE5Q|+0B6ei
z%T78=*7`3SO3y3v+xTm6_44?W0N{>`J>K(44uR~iF)8UM#bR!O_7`;kHL$qZA(Z|P
z<<|+_`A#rcjS2`~`+yw2D6^u;2|`{DhBmMQX!@^<YQGA}^QLA$^q_={9qwrLI6Fd2
zo6FQ?9p8iV{|2Z)`Di)gS&hK{`tP-b?LTWDo=OOmo-r?{Ew7D&Ir$k3gSNX<BLNT3
z_K2RW!;hu_Av0xO2+8Dbq80{Rs5awWi25L?gFf9EKO27Wv+vO0`7|pk)_6Sb;VXJ$
zQqe;LDKG?G6vU5szDT%nNQ`ev#e_d%|JfOGUvL?+ZQ5J-S_!@L`0qw^_>F3&yUuWe
z*vrViG|Dxv7k5J4eT_dYeG<iAjR?3JZ%hr8hNBs9MGoq30mJ57HFY@4e-Ad^$L#?6
z$aw6)9d85oRymrQHPb>vYB`Lp85R@{M@5)c{O-3KsB?Q$BM*j*PaE>NwX-wi4gE%7
zhAc~A?K;@-{>ya)bH+22&!rb|lexdgiw4JYUg`}if!=&T2cU}WXw@FqBvH~Ay!-qt
zr13Hn%qV+H5DY~E4@7;o@i>2K+LJle^=$)b*=wSi*W}LI9N%A;<grH2e$-IkKf?#1
z3Vf5%=bMFv+dCbeC@uB|{?;pH^UXcUz+;{W+uT&*`&!J0IVLiFv{ZSCm=qw<@CI=I
zYLMrU&!H&~+9@7&e6M8+y6Q?c<^uXSX2O})U7md~F4bU?5~zhJC}{e>|2<^3tofC}
zCpcwDE>yQ9Q$3#ZTpK*F5LLxyZK0U?(zE5&voM7N$kIad*-=9qY$)Gz9F#6`x3V0d
z-JfcvZ4HFs8j>*$``rOlD9+D{H_jQr=-vj8C@D|FY)bO$GlKIy_M_t*KL{%~Hqpwe
zZORKX-~a&Atwy}Gg*+3TbpKSsd%Vn!+g4l|AF5b9UT&k;8pD81KXNf%YipaGMrBk)
z+Y#VzE$fm*A9+*t;%j{G*SmMOWn|JoyAw>TqXAD_oFnI{i-=ZSw==^Pf{hz?S>DRH
z|1>$Ajl9Bgd8?7A1{-(^Ws$uMlweq(O~*H#x*(J4drC@UoC~rpe%@||mp=1aZGlRw
zBTxq<X(xh%vox9mNLQ!DE6`ixd9m35da0x*wXZKoSf(VCySOl(cg0tBFEE@p2TnK1
zNS{3fAQTz)=}t?p7_{_snVm8SLX)0m>LpP72%iOVKU9$^X*#aH7(=mtd076&x8s|^
zkir}+NPhl@swTTQ1gMd)0Ae#U(Bowk;L(Cbcsi_%bjP$snvFle9OsxRS|1+p`F&iv
zcmq7>UkEc(Q;QB^8M2}wTl)X}Q)ctLG`<<Q39Heqr^liS(te^wo(6*{&q`UU?5s#*
zh!mGFGyrac+Tba{Gfn#gTd2oBO^qczKcnq#{zNG@|NiYfqJuEaakT(GDvapZOHuo>
zo!K%W&|eeA1W*h4Wr|=|^U{Ez29K~CBD;Zo|1^Z;=Cy#3{+2=1$!XIlC4c#!y^D=A
zJIv@32WUf6%L1$Z9JB)*O;?_%qR*v9Q3TuY64KgY;bLRTks?I}!Yb?8aL?WptD1Wk
z_%LysBKRXJS=j8z@vGj}cY2)d=qo9YI``2q`!Ty{v8H2^K3NQ<LszHE5ybX91Dqia
zr*l2&|EeF&wuTe?P7N162Y@UaTH|QqXm)*m;yB04iFYP3mVdb57PivN$yig4wdrD7
zW0qzH4Rg?%d#+i2ZxDY8*8eAHM&GpAFL$v3jB5E|s(L)Q5Gv-GhAlqeEcIwD@L&lf
zhgv^s+YQ=1J91u1HnC5%!__E~l2IMZF@q<D6GV{*=cv5T21#@O_fRBA-48%JTXs&)
zbumYyCU|FG-x31j_!750`_OtI^AZ$FQp*?w;F*%@^6}^Q%s`77OQ%zLKDraIvN&9{
zyy$6z(*v_=4Di>M78d2^?G)xOU%I8mzJG7g`H}0m$?<y>Eod1En@0NiHN3It&Qjdm
zn2wn`o`*?!DcBdspW6j$ic+pTNef_hn2oAgSkQ;xw*I@!?&I1yKm1!fagD!qtoL8)
z$<AE+lT`7zks>`NT)+ci;m)2O>0R+#w{onh1U6(ZCT;&HE6bxxJI=h19M`T*){l5D
z;{K$V+Ky^O@~wMJuaW@gpcoB#NCH1{j}44B!N0q|#`9Y7-CqPiMv4S>)HE@y|0jS}
zm~ejplmY#2Y!caf*KhFe8AZ$a9#jwxs7g3Dd*bvfNo+?dO2)Hoi5I^SuZT!_jyaSV
zB1Ax_h`Ra)<{zyKH%SBK=e+L;z|%2n43|L9l-yXo2QS*ltvAjM^sqrb^XkO_*Yyx+
zbDB31d-d}Q3DpgEyn{I6BD@(;wtFUJBPdk~|7#$Q{k7t)+o<Miswj`!T})AZ=k?sI
z!qj^pkN$l#gr?Tkckkc7e;asgk$y+FmO&?4?PmR2j9=iVJ5Tj1EuWUU0S)0Lz-jNU
z3*W&$V*>tNef=aWX4nMq=(}x5dPS(bk4g1pS;6342A-xVGR%EaTVK(2h%nwaY>xoR
zL>p(m6Ad8~vL4{@-CvOh)IVN%qrdZn<Np1)rhpT|fl4c_)zK0S0C;C-XMv-Z54~3O
zqL9#e@;LY-aAKYL_4Ki~MU2OWhRwFoWIZo<X#)MI*zB?Dy(I(EFk$fj&Xl`_B__v1
zzl_YJa#M`g?K=~9nt&;HVV@V?EZmRxGR=O~)D_c%>@Qs@^To#+=~gE~jz_4#vlTQP
zmM#Y^Gi@P<Nsa3q-u_U>7_`4az;F*TiaDAI^YmT#>0i`pO{Nc~^yPmNlafAPO_?E8
zjR-=W-`K2e(>&OsQH)HVIannCO;kL>07gAYHYd*Pmt+4TW_9e!Of2h8HaBP5bM0qN
zc1+6OlXqPvepJrY1e&pqwwh87)@A{jhLx>de*k*mpr<Yh(<YZHyaa10oOnU|0%*n~
zG|lYp3@JztJy^>!SN3#2r717}kU&xbZ7Vj=JzUH+A`s&5r6BSBQbb~!r~a(2Vi^uH
zIuMX-X``-c8-C%(;dV73o(Ifb@nX81_XpJ>eaX+Gigil9c+f8`JhJ_^F*^JBZ-(5t
zb^syy#-)*|Oi39uJG&kfK!@hL-ZN4CtO3CfDf}nr3YwCf0FCgSzDA;p?;6{=M%?^p
zUR=obvySUDaHxsQZO>Uu?&=6K3Q}`E_=WJ6nt#V>gsHyrotZa;;qS<+5D0w$iiJ)p
zyYJ7X-|g)!T5|Vol-*=cF1NPLR`!Di8%`5!3Z?wtoFEOhes+qxRb^{YeXgK5dw>Qr
zL&9JEQooFqmd`5B%0Sk&t=V3~;8JA{K!b5|w&yEzocFo#tyA52rF^!Ej{H4z^G+d)
zXfg!C+=qWJ07>w5VPJ@~r|Li+G43H74Je=CTTD3Fav&@HzlI#P5Wj~uT#h!k4>1U|
zU;@?r{Dvy9oipgZ&AS~Dyf(aZS`Fq8;k$_7vys&i-Jec>-0$j9gDx7&K54r1=B~!m
z-DqOe`I*di`t%LLzRr-`+YNBGzpLXpO(x1^V|AyQ2q2|`3~J!CQ)Xg4dZD)k!!`jY
z1?e;fs@TlKgRtfjAKN=1#-HZ{VWu4D<l3wwoesjXd2QI9)an#@PCjQ6c-<cDAoH_1
zgv~d|iHe{hA}A-jSnN4iCvTM%o6pVr*)YvVXGbM>_uTqh*2jG!<btiz%P~@0`!iMb
zIcpHfm`H)QjQD18@1*+Yr}Dm%pam8v$kPs$R*K7g$Fc#y>37J&ayeFduEn#Ka3)K!
z%ug)U_nTE|5uFfz!1CCdD+Ktd1dzk6*usL=8f?1IDS`A!@f4ty+9KHOHN3@@R3-wn
z1TGCyn1UK3RvW?xSXa=xVb)H8iS#=67Y9Xa)2lWTFP)^8Fyz1;C%#LASIWR8sq#P)
zbU?dKV3XbGSl8HAQ`lag)bH-}py~W&;{cEQ2!APBu<_rH!=1Z;g@~r3cNp^;x+A{v
zFiDK&hKS_vZiG`L-t4Lv;a%u3{gwDDzPD`dpH%`rJ7B_%VdXysLOC`Fv{{pG|I$Fn
z<~--N#*Aug4B;<8t*`*1pHK{Qn%(M`<KGWBN(*l$RC^juPQTuS_PuD{lL-d4ptt6K
zJ12Lp5I0wtZy2uCW6FsD5@?0FyNAHeyLc1w3=yx#^aT+>!VE5@6Dmk0q@_)-%Lm5~
zeh%TI1Xi720zqd$ykq1Go~Wig-Iej18PVG~zc?qAmp8G=+u7&*kP{h%oNxxgwK)r-
zKfw<r0V>bIIk1v5`hz6;Rikky;(fNL<K?Lazrr+ubT-1#Q#t%#gA?=uL91J^W`7jH
zfBQD*DG0Cs-V%nu5dqL`pC}IB-M><XZySv20Hdu7RBZ;mG2BKwQ`=<Srm-l+Y?>D+
zc)qny1~xEF`0(k`%m8WQcUkSf)}<yv*~34pyektUgu?)8F~LAN9p2P4+op@VJ~JD6
z^WkvYMvul~*A+&~Bbob`i~o+k3q^3=AQNnC473Ng7Gx-dQz+N>sInW3gWl*YEq0ok
z`*QN~>BeMphQ$Bc;r;}9?!z0xKh%-?zlj#RnSRs75cSL29kKggxu?5tdZKyfUY95D
z&MO+zG|6^U0=O5TW92<}7CJpbwby?KrxI<6zt<9Q<nz^-c(E{?Lqg(joZcY>W`LU=
zAz($Eoc(DjEJ$!Bk`}p?`o-1{O-#)2Z8HlTDm95Pll{-RpML%DsQG1<l`s<TZ;Y7$
zJ6ZN}(u7BYKI)yTkh5q0jZ@P;+{51%!SA+Y`n-TtS87xs80at7H{&8wRKQ!EhRL+3
z2>wQMZ_ufk;ZkynTx7H!XjXmi3=EZCWMkD3rN28$s}l^wo*ps-b@Ygo=Fm(L-~ItC
z=Kq^EEUVrx${^1cYUfcf@apC&M>9T2evo$wx@JI1-3%xbiW%^l1j}Ckv1@veCmNL-
zYH#ta21?&~Pwzl9Q$*hr5RnMEuRH_39_2`CR{X04KpORtMcIygyUZ&2IZq=Iyhs4$
zyqxCc3jQ8=ZBEiT)>%p8IEqai8UjHFXDsvc>Te?tF5ynLvSF(KI(8^{CyM_RnN}_E
zSlWzyV*u6=Dc4ef1>Dz0-+QAg#y(ktZ;cx~Ef#*p5&H(JsE`1XCbysf@QeTCdiu3W
zZ9RgU=n4SnUVG}AA<|x=%g;<`QcYo4b$3GKn@$+8iM<__-S`FaPms1@6Dpt+wCYz@
zROo|u2b7nW*Y$P$s}1W@)3ggq)7c?(6b{qsh+*9eFtN0}1AOCk)G%n}52Ve{Pg)5d
zw@KUK#NNO~E%%HDzZxQs+jX7)-FdZ0!FuubUoREFqVoz0p4T`U19S-tBxc|(5JrKa
zml)XD`?qJ?WJWjVx_$;TJC+r@w=)RYC&8iapXWvipQu-0ww=Ffw+lM}f`_(Q3;#}m
z&`?(yPnP9cH`OAex(7vi@TCs+pao#Aw*+lW;5`Ez*kzjkBJvzC(j)r{ZlJM|S=90M
ziZ#GOM^z2lS+z33s{I#IEht4ieIxz9AO^jw@Wi0>-V=ynm34~~>P!UyuQVyIzP)2J
zGqcy>;XcJv?K@q(LcpB%;c!qDfy;Tp<4BYFucYwQW>xrJ#s7jF#7V}?e@IUL{d>u0
z@>d?PD|lYyCC+3=`(x6j8EP0a|ArJaP8rY~|4)42W*kEcz|mEXQ(q#em_NqHw*w;2
z8+f(cN;n1HpPw^3HjQ_!P5BGs<!;8swoi(93UKZpPz%k^&8cZ<yac7hZ7#0CVguo&
zo@7Np07KNCJh_eg`q&k`OCkq`3be;IrBN{48_0c3Kkc_AGdDK}EE=gVJVimLggt0N
zO(S+f>bx!Ju=xw{hIdmWgPWfH@_3!@PY1d+pKFFJc7wi^#`6T_8gt?*M~4$!eXEff
zAOWM9Fqb+~9C!Usz{`D13#l_RdI&CGa$5#nbDlw`IBmA`<|(42kLDT)ly=LS_?Abv
zzf<55LD>`VZjp#UR?<x*lkfhvM>It<1B8Y>!nlFr$gsnO^br+k(2nv>B{bzpjJLRz
zX=8JRo+wB4q+Xf$D&vyCjL}#)ht2;e=E~!#+`j%kheX#cq&kL>NODn1#txUvQzaA*
zZYML*HC3ilKNTS*MI@Q0OvyYq7{fK5A|YdDIcAP`J#|0#^S-}#`1k#M`a{oop1s#z
zd#~?z?Y-Ap_5-G&S(sb#1L;Sg#K#ly{NJQ~OPxc`+HiUt2y|?*3gu?wp7?YU_IQDK
zD|?t><2f#?M$R<L{-SBo*9qPD&L6y-1H7CJDJ5B7&>hyBnK?_&-uIHy3eXRCty~+I
z%h^-3>q;?hY36oIY+t5-)3dKTagMOht=}s0+R8gih2$?wY9jTdb9&Z$zU>$8t&}Nu
zdroSf4pfqGOn9_HE+*6v>NP2!j&8BB#J9bdQvYBhPq4dmf=+STO%~*KWPCU9>f-Nt
zPLB@H%iD<`wX?rzAxB&vDo?b{ITNt%6Shv|Z%FQ6|5D`TLR<~-UR)??9iv1Dpp<xF
zoutB(4YQFJ{v>XXk}K3r)l|ww94k`RW=9==<@gEfMGHCiaQ=ETGONZLXTh98!6D7`
zCJ!gk#IA1bEqYNz?6kY^z92OLgp!2SQAejWKV4|=O@tn#FjF*R`Cl<7RawgW$llOJ
zj7UHjd-jpp%F>c{k|S>$u*o?sEr$s_svg09xrG;%l)RzXZa;7Vu`&vUiAa0k(7>OJ
z%1QU@+y*<UedxOoCE_3wERb+Nz>H)2`#%I6G(F9R9-8he-6!zR<0w4S-;b5tcgV77
z>YYM}f&w|q9*A0zTdpv?Wd@-|RncjO_v%PPYMT!&&3)oLEs9+kp>uaoU}xzA_Uezc
z*~cc@B_-*ZEC@wnU0=e@`ENa9Vruaf{{`O6y#;R`5q((xWqdwfb*%CUVn&obAhNOX
zlM^<&v=yI@>X(QKRqwzHoO4O?#=XK55E6soRy>^9R?m49A!88z#z7JbrlNmd<3PBd
zZSyoRTgm#I6u=&6ieilj1uM(TCn)9TC-VHx06I8;1Q*p`F<+kBd-S5a4hSFQWt#Lw
zGllXAG=Lnjt}Y-&cTIV>L#xuiK{M6d(2V4go<ORBjz-v_symp>(jfEmR*^mXG{Wwh
zF{@y*M2GC4K<==Sn%eC@-Q0{wq^$i4o~+)}5N4Lcb=)ScxcI{R_wP%G?kmv3{5RfS
zDv)riye+^8T6x(Jg^KMv4oROlVF#gXaAy+<LALR0Yd3DI#3xZ*dJFZcHbx{1Aov!-
z6YEOh#%T{LlZU|G)2{ZhE(hudjH(e5NLH*R47syMsyI7mWn^Rk%$2T`@VJzp&9`ke
zHZ<(G?ytlVgh|od+#C%V)qxiIsi_)jYHAZRGu?2J;_o`0Jr``L>AxCG`|Q>pE*uob
zb#P=%N5nvTeM#$)h=SokRK4fgRtrqDtL)3}ZsTqLstH)Au{r~d@wT}=lfA0Kz`$@s
zRFrEenojpyq|t;vA5|HP#?Q>$zoQf+YK72m^?~Tx%G@7ng6xmqr*Ksn8XKpyw6w6F
zf>2Gx_^U&;-|j2S5SF4-QVzlG*n1b)&KIYZl19u8tMk6~f4w9s*=nJOQ1N+n#%e-X
z-6Pk2!tr5qjMTa(bE>jszFbo+DJgktQRMVV%EQa6&SiRfdLJp@@trz)l!dp!F~j-^
z{j-#!6aV`4d5avcdOJ%K6Y7JrHM3^<Hlfe2k8)LP(buEpJtpLGnj0F<>g%_tQ75$z
za^Zv!9$BfWhoChdqLGP7W^%IdppTADa0-;unVQ|CFIQ0^L^aE+-`7N1S~^{#qYsw8
z`9aE$)qF36fQv{1|7HRX&5ZT*LMJ9BUPC(XZIuJ!=EY~iPk7Ckhw{isR0>{$fKk)D
zOSdh&r=|6t$sQx#xNzpoi(9wwb@csQljgz7nE6&`gs@N`NU-hUK@D1`p+S0LqUz(v
zk9%HO2RAk~32Vil(Y$!kcDo{E%V)w&L8P?O;C16{^Yq$U^_b)`RVT)|cP!~eCfvNQ
zcJYPKe-5p7iea2_iM>>Lb+rwxbeXg#Y1y_?J2kVi+Ra?Fy!xljeGxJl*i=v@G6Xig
zbM2_s#W5hOJvg={Au|@mqpnLt029nsb45N`P-0`-a&GZRYlA!ev=kv%F1)~$)=4}e
zhQ43HxG_Niw8=1A{Nq&&sL<q(?*!QjDPVE(@Uk40^C;FJSY+;9Z0)q;v94n9hp?I0
z4}Zg?hVeaqGdO<rH^hW*>10HdU)a$f23PZj7uCad?Y}p(RfnNgH}=rj`0EK*gRoBO
zz6t}(TEVENxXamu3b*N=eHalpYmP0e4~J8Kp)|4b;dmEG%VZjel?lCz!Ek<<sUUij
z4R~x4>AvPpi`oN5cR>vWkp88C44#MZ9mfMM7%?h;l)lDZKR?t)#zWu6&^L8JO@L@g
zi^6c@=kGgYC(B5I%#^<y$U}tnLu^6V@7CD&0ED(e9Gl-DR$(t-lJT6LK(3gA*%47W
z@BUe(^-bs)L%t~<#jkZV$)5wXoN}XZVEWq5hW_Bk{?LG~`3>P+*?Wk2{l~gF3$4q@
zV#?~mA&Sm_c>yDU>a#w(nWti8kFA}mwzfX^M|?*YLId*f)!4GmY=y|<C-Njf_UzLr
z;P{oQzHd;0(N&zt09G|1g=q*yHl(3q&uLSm^XCuC$e46>b@jlxbS|~9px|5O`b@{W
zpxq<$*JyW@Hfk;<!}~4&H}b^+q`-k86bH*3US1bB`2{-S4LepVOL~`l#>^};BjX6%
z{z~}G)bhA=!l7wbXQz8hf2{1)1a5Bb&cFV0{XOSXM@M2pf*^o_Bc$vrrCmYDYTs->
zmlU3`^KMs9PlnaeC|4qZ)d5QL?KIj3RQz1(5>FIdk&6LZI?c)GY90|2GqklG6ZVe9
zP;Gg+Zg!bQGk-M(Q<E!vP#uvK`3&Cpl0Q6f+n1Q0t_6rX-y6WPTjG-&1419e*>So5
z1~$QY{|TNVr%G@LIW+{L@rBj%z#31UJPBA37Z;Bv`mRBr5OW>McS?Q`dUGvl2y8{#
z?d{1ri*s*xWu~O4dw6*CKwg4M`~#?zaaap*JsXNm?6YL&-FdeFj@Vb`Z!&Cu?Y(?u
z4=1PH_u+>7n+l-zbK+$D064$y!}aW^Ao_V?eEf{Hb#kH8r-bj{|I|8nnDd-*_a|H{
z)OwrHP+B{?L5FMa*EY@E&ClcvdkdZONkjx3$d$kA#qN(9uxy)lfl5!{#{({S+9&{H
zOYr9{JIjRKWJi*3AYnr|Xxa+Qi*>(}pE;;UKXMR!Bgr>UkK4M>4g9`m?_P`cO{U~=
zOA)~PGjL4rcsKP?JSjYpbVs9CNTZB3tWZnhQO|oq;!s_f7#p+D%U!O!)sdtMcn!5-
z{BiA6tFouVc9HOJu<c|+o5b5h=kTTo&#RO5mLUpcun<ZxJ1{viukHQ#@#DBVyJpMe
zB7}Zl<d(kobgky!zDy)ZZ4XukHJRVRa0f*m0K@_5F{7%s#;g`es3Kj}|3>z*r0^kb
z9v)cX99R6R`oFz;xen9h0Xqy3x_)SWyjiAS_u6MdBH-{gJZ@xU#2#dDaCCG_Gc9YO
zo(xm`UjKkd>xiRw#yd>gJ2<q1qd*4rOk`wcx5N5gjN94}8N|p{{Nk(A$;ru$t*z8n
zf4bY|a@Cec+9d_`c8PqLrgmFCpSj=f9x1eiU~ynG2cNfNrKR!}aI7_7*Z0v<iXC^f
z*{SEX)~~<*YFg=2;yK-ujJYV4l@{@5>c6$~dO>neW@6$&zutV?unysiR4NrqKT0nt
zk^44;v2A`UF>`d(W^*BKE24jv8te6iYwOE|u4iT8!=$*)DO(m_yQU}Nd&S&qm!^9Q
z?Aio3p(36M@TGnn5t5;xr^*#+=)I@>dOifYq=gaJG%G49N+I8XGw3w$OT_{hO89H5
znp*)xELa~K%f`w)l+t)iTpix+`@KP=bh@zDz4g@cyR9g9vVVgnll_Q@(rVDfV%I<3
zl<+`l0mb=uyLTEz@B#DL!!^^WJD0s$_^0ygZX0x1c_IC?G7uSaFHXT@B0}`Y5h;KA
zLR{?OwqyM+Jmiq45Cc^*_S~B{<FK#0hX<FI4eao03FQ0x6>K1xrWuTEh{_8&v6ro_
zp91Mq)1|W3{mKaB50IkOOb$CWYGw2-vAFnSIUGySB3a)sTXD18ee2-tY;=^ME6dNy
zRNywIe>X#$AKEn(U^@d|BQ|Hg5SXGL8DYFM+n)oDQ2_nXqeqD#ZM%PEzH{$Ml1G<O
zDO{w0V(?)4a-?vNzP`Tjr0{SKt#gi!nM_Pf9BX>e99VPaTJv(R@MxEbrIEdT`V|P^
zJw;y`$|mUX|Hcty2gf86laq!d(*8~1f8{>*;+FGS$rXGU*x1;}(GvOiae97!M|?)R
zDmarCeyi3buRf18sKmI~+fO&{NHs7qGmFVC|E`_dr1MK2G5qD053CsdhzMh7bg`OY
zS6X?wIaFw&-rFPQ-s@bHUQl2J*G&46wzrR|>qJU)i5GoNXl&HhYV#Wc)H64K=9(2l
z@7L;<2!|q%yV4U9RQ#$eEq&&1KM^-eX(g`f0wC@*cC34T%KN;DNvdJ0e)@|S$}uu7
z-Fh;8Nsz8mR#|m<2i}VN`!z+D<<MiQEhT!szP=mR8Y4tAGBZ1*spg517B{q@Z;#wp
zG}SGwD=p9D=7&HAAPr1fII$?4saUqyE|zZ#_j_v_3$2T78vL=#rZS5$gFV(Yx3>Lw
zo#Z`TkddC=2GtqC%RbU;Z3@YsRLA9+K6%fnDEd-SmHn6U>pgG-l}jy42cc3Z9hxu!
z<rcFKsuD^|OYekC6qEe?DpPcyCP2*?Ig)lyVAj2qt&k5;V9?0IVkVmS8q%IB7Q4-Q
zAcqA2j$dhZd412$yB@$`0h4|0E_WOwFGJ@Mv9Z(lh@ij|*S}xJh5_sY1!+3?w^lT$
z%aOV&NcSKy&Hz|gHD#Zvsp$#dHFsEY2Ai9kYt$-wGElQT*1Mk3uQN)6t`~ZXTvE4$
zt^z3(6cyR}@JrR&fhKMYh04%sw)3X}jSkLFOgLEE+KSL(?%&_(3$dzPeJ(9jiH6q&
z1@=}}0z>P%Qndu&$g;w!%YrUsGC8qBZQx~FqJ|^X-{h9L@(i{{iyw}++QB*E4fx#S
z{T>rp(sMGO|Hjpk;o*8hqGZ*^x!`@0-I11l>2M=imO;AV?G03B;o<4&u(`IFk(C8X
zdj4iGH@o=j4%kS5yZvo#GRH)Sqs_%{foH$0t1{0?r}~xFlGse=a*LejKHs$fLyy3s
zhe?W7@~;Fd>HPfpS(|e{08<6*rq(7cw=$FkUVQmtvADF<IJtO4LBR|*67tvPUPQ;l
zNIUnOz#Is;REIRJ&RM@@1;~dSm91Ir8OekMd^flJ?+XhsL6}W9$0=C?jr|3?f!Mrh
zK`jH~YLX8zF65Yliwo7Q^;D{f#dS3dLSf+!H`2o&@|yaFxUB%_Uk2wrv~}#bk&zKg
zFGq0~5Hz+V^gKNaSs3x7aG)gZ?Ch)`Y262fheDGM6kT5!<sv7T)x0z>X<ym&1E?Th
zsaQ0I{T`ofaN)4Iv)k8qEW25mnxl@9b)l6K7;1C*^#$F-oPU;1Md)-N3jYu+SxKlG
zg~&}CnCg_Iq<6z3BTpfYL~N*c)MuJw$L$YU5Z|J$tsMfA!F_S)>SB(+^L<{quTFz|
z71zcS=a*j`BiBY@)sxDXzj*)Tmxpx(vJ;a&Fr{p8p6UO}p^QH_2EflvY(MxJ+gtzR
iOa70v`~URG=oZ79eskJ|?qC=U(o)kuovmVZ>puV*>GJ;o

literal 0
HcmV?d00001

diff --git a/Udit Soni/car-recognition-aws/evaluate.py b/Udit Soni/car-recognition-aws/evaluate.py
new file mode 100644
index 0000000..b5a9bd1
--- /dev/null
+++ b/Udit Soni/car-recognition-aws/evaluate.py	
@@ -0,0 +1,66 @@
+'''
+Created on 10 lut 2018
+
+@author: mgdak
+'''
+
+import os
+import argparse
+import sys
+from pprint import pprint
+import shutil as sh
+import random
+from keras.preprocessing.image import ImageDataGenerator
+from keras.models import model_from_json
+from keras.optimizers import SGD, RMSprop
+
+def main(args):
+    pprint(args)
+    
+    test_datagen = ImageDataGenerator(rescale=1. / 255)
+    test_generator = test_datagen.flow_from_directory(
+        args.test_data, 
+        target_size=(224, 224), 
+        batch_size=args.batch_size, 
+        class_mode='categorical')
+    
+
+    # load json and create model
+    json_file = open(args.model_name + '.json', 'r')
+    loaded_model_json = json_file.read()
+    json_file.close()
+    loaded_model = model_from_json(loaded_model_json)
+    # load weights into new model
+    loaded_model.load_weights(args.model_name + '.h5')
+    print('Loaded model from disk')
+     
+    # evaluate loaded model on test data
+    sgd = SGD(lr=args.learning_rate, decay=args.lr_decay, momentum=0.9, nesterov=True)
+    loaded_model.compile(optimizer=sgd, loss='categorical_crossentropy', metrics=['accuracy'])
+    score = loaded_model.evaluate_generator(test_generator, 3957/args.batch_size, workers=6)
+    print("%s: %.2f%%" % (loaded_model.metrics_names[1], score[1]*100))
+
+def parse_arguments(argv):
+    
+    parser = argparse.ArgumentParser()
+        
+    parser.add_argument('--test_data', type=str,
+        help='Dir with test data to evaluate model on'
+        , default= './car_ims/tst')
+        
+    parser.add_argument('--model_name', type=str,
+        help='Saved model name location', default='./vgg16/tst/vgg16_lr001_dr7_decaye-4_finalModel')
+    
+    parser.add_argument('--batch_size', type=int,
+        help='Batch size', default=16)
+    
+    parser.add_argument('--lr_decay', type=float,
+        help='Learning rate decay.', default=1e-4)
+    
+    parser.add_argument('--learning_rate', type=float,
+        help='Initial learning rate.', default=0.001)
+    
+    return parser.parse_args(argv)
+
+if __name__ == "__main__":
+    main(parse_arguments(sys.argv[1:]))
\ No newline at end of file
diff --git a/Udit Soni/car-recognition-aws/matlabtester.py b/Udit Soni/car-recognition-aws/matlabtester.py
new file mode 100644
index 0000000..723f78c
--- /dev/null
+++ b/Udit Soni/car-recognition-aws/matlabtester.py	
@@ -0,0 +1,342 @@
+'''
+Created on 22 sty 2018
+
+
+3D Object Representations for Fine-Grained Categorization
+Jonathan Krause, Michael Stark, Jia Deng, Li Fei-Fei
+4th IEEE Workshop on 3D Representation and Recognition, at ICCV 2013 (3dRR-13). Sydney, Australia. Dec. 8, 2013.
+
+@author: mgdak
+'''
+
+import os
+import argparse
+import sys
+from pprint import pprint
+import scipy.io as sio
+from keras.preprocessing.image import ImageDataGenerator
+from keras.layers import Dense, GlobalAveragePooling2D, Dropout
+from keras.models import Model
+import shutil as sh
+from keras.applications.inception_v3 import InceptionV3
+from keras.applications.densenet import DenseNet121
+from keras import callbacks
+from keras.applications.vgg19 import VGG19
+from keras.applications.vgg16 import VGG16
+from keras.optimizers import SGD, RMSprop
+from matplotlib import pyplot as plt
+from keras import regularizers
+from keras.layers.core import Flatten
+from keras.layers.normalization import BatchNormalization
+from keras import backend as K
+
+train_data_dir = 'train'
+val_data_dir = 'val'
+nb_train_samples = 8144
+nb_val_samples = 8041
+
+#Creates an array with following values: picture name, picture category ID, train/validation label       
+def readData(matFile):   
+    content = sio.loadmat(matFile)
+    data = [(_[0][0][:],_[5][0][0],_[6][0][0]) for _ in content['annotations'][0]]
+    return data
+
+#Creates an array of all classes
+def readClasses(matFile):   
+    content = sio.loadmat(matFile)
+    classes = [(_[0]) for _ in content['class_names'][0]]
+    return classes
+
+#Movces raw data (pictures) into respective category subfolders with train/validation division 
+def dataPreprocessing(dataDir, labelsFile):
+    data = readData(labelsFile)
+    classes = readClasses(labelsFile)
+    print("---------------")
+    for recData in data:
+        if recData[2] == 1:
+            #validation set
+            os.makedirs(dataDir + "/" + val_data_dir + "/" + classes[recData[1] - 1] + "/", exist_ok=True)
+            sh.move(dataDir + "/" + recData[0][8:], dataDir + "/" + val_data_dir + "/" + classes[recData[1] - 1] + "/" + recData[0][8:])
+        else:
+            os.makedirs(dataDir + "/" + train_data_dir + "/" + classes[recData[1] - 1] + "/", exist_ok=True)
+            sh.move(dataDir + "/" + recData[0][8:], dataDir + "/" + train_data_dir + "/" + classes[recData[1] - 1] + "/" + recData[0][8:]) #train set
+
+#serializes the trained model and its weights
+def serializeModel(model, fileName):
+    # serialize model to JSON
+    model_json = model.to_json()
+    with open(fileName + ".json", "w") as json_file:
+        json_file.write(model_json)
+    model.save_weights(fileName + ".h5")
+    print("Saved model to disk")
+
+
+def prepareDataGenerators(batchSize, srcImagesDir, labelsFile):
+    classes = readClasses(labelsFile)
+# this is the augmentation configuration used for training
+    train_datagen = ImageDataGenerator(
+        rescale=1. / 255, 
+        shear_range=0.2, 
+        zoom_range=0.2, 
+        horizontal_flip=True)
+# this is the augmentation configuration used for testing:
+# only rescaling
+    test_datagen = ImageDataGenerator(rescale=1. / 255)
+# this is a generator that will read pictures found in
+# subfolers of 'car_ims_dir/train', and indefinitely generate
+# batches of augmented image data
+    train_generator = train_datagen.flow_from_directory(
+        srcImagesDir + "/" + train_data_dir + "/", # this is the target directory
+        target_size=(224, 224), # all images will be resized to 299x299
+        batch_size=batchSize, 
+        class_mode='categorical') # since we use categorical_crossentropy loss, we need categorical labels
+# this is a similar generator, for validation data
+    validation_generator = test_datagen.flow_from_directory(
+        srcImagesDir + "/" + val_data_dir + "/", 
+        target_size=(224, 224), 
+        batch_size=batchSize, 
+        class_mode='categorical')
+    return classes, train_generator, validation_generator
+
+
+def getVGG16Architecture(classes, dropoutRate):
+    # create the base pre-trained model
+    base_model = VGG16(weights='imagenet', include_top=False, input_shape=(224, 224, 3))
+    for layer in enumerate(base_model.layers):
+        layer[1].trainable = False
+    
+    #flatten the results from conv block
+    x = Flatten()(base_model.output)
+    
+    #add another fully connected layers with batch norm and dropout
+    x = Dense(4096, activation='relu')(x)
+    x = BatchNormalization()(x)
+    x = Dropout(dropoutRate)(x)
+    
+    #add another fully connected layers with batch norm and dropout
+    x = Dense(4096, activation='relu')(x)
+    x = BatchNormalization()(x)
+    x = Dropout(dropoutRate)(x)
+
+    #add logistic layer with all car classes
+    predictions = Dense(len(classes), activation='softmax', kernel_initializer='random_uniform', bias_initializer='random_uniform', bias_regularizer=regularizers.l2(0.01), name='predictions')(x)
+    
+    # this is the model we will train
+    model = Model(inputs=base_model.input, outputs=predictions)
+    
+    return model
+
+def getVGG19Architecture(classes, dropoutRate):
+    # create the base pre-trained model
+    base_model = VGG19(weights='imagenet', include_top=False, input_shape=(224, 224, 3))
+    for layer in enumerate(base_model.layers):
+        layer[1].trainable = False
+    
+    #flatten the results from conv block
+    x = Flatten()(base_model.output)
+    
+    #add another fully connected layers with batch norm and dropout
+    x = Dense(4096, activation='relu')(x)
+    x = BatchNormalization()(x)
+    x = Dropout(dropoutRate)(x)
+    
+    #add another fully connected layers with batch norm and dropout
+    x = Dense(4096, activation='relu')(x)
+    x = BatchNormalization()(x)
+    x = Dropout(dropoutRate)(x)
+
+    #add logistic layer with all car classes
+    predictions = Dense(len(classes), activation='softmax', kernel_initializer='random_uniform', bias_initializer='random_uniform', bias_regularizer=regularizers.l2(0.01), name='predictions')(x)
+    
+    # this is the model we will train
+    model = Model(inputs=base_model.input, outputs=predictions)
+    
+    return model
+
+def getInceptionV3Architecture(classes, dropoutRate):
+    # create the base pre-trained model
+    base_model = InceptionV3(weights='imagenet', include_top=False, input_shape=(224, 224, 3))
+    
+    # InceptionV3
+    x = base_model.output
+    x = GlobalAveragePooling2D()(x)
+    
+    # let's add a fully-connected layer
+    x = Dense(1024, activation='relu', kernel_initializer='random_uniform', bias_initializer='random_uniform', bias_regularizer=regularizers.l2(0.01))(x)
+    
+    # add Dropout regularizer
+    x = Dropout(dropoutRate)(x)
+    
+    # and a logistic layer with all car classes
+    predictions = Dense(len(classes), activation='softmax', kernel_initializer='random_uniform', bias_initializer='random_uniform', bias_regularizer=regularizers.l2(0.01), name='predictions')(x)
+    
+    # this is the model we will train
+    model = Model(inputs=base_model.input, outputs=predictions)
+
+    for layer in enumerate(base_model.layers):
+        layer[1].trainable = False
+    
+    return model
+
+
+def setLayersToRetrain(model, modelArchitecture):
+    
+    if modelArchitecture == 'InceptionV3':
+        # we chose to train the top 2 inception blocks, i.e. we will freeze
+        # the first 249 layers and unfreeze the rest:
+        for layer in model.layers[:249]:
+            layer.trainable = False
+        
+        for layer in model.layers[249:]:
+            layer.trainable = True
+    elif modelArchitecture == 'VGG16':
+        #train the last conv block
+        for layer in model.layers[:15]:
+            layer.trainable = False
+        
+        for layer in model.layers[15:]:
+            layer.trainable = True
+    elif modelArchitecture == 'VGG19':
+        #train the last conv block
+        for layer in model.layers[:17]:
+            layer.trainable = False
+        
+        for layer in model.layers[17:]:
+            layer.trainable = True
+        
+
+
+def initialTraining(optimazerLastLayer, noOfEpochs, batchSize, savedModelName, train_generator, validation_generator, model, modelArchitecture, lr_decay, learningRate):
+    # compile the model and train the top layer only
+    
+    rms = RMSprop(decay=lr_decay, lr=learningRate)
+    model.compile(optimizer=rms, loss='categorical_crossentropy', metrics=['accuracy'])
+    model.summary()
+    earlystop = callbacks.EarlyStopping(monitor='val_loss', min_delta=0, patience=5, mode='auto')
+    history = model.fit_generator(
+        train_generator, 
+        steps_per_epoch=nb_train_samples // batchSize, 
+        epochs=noOfEpochs, 
+        validation_data=validation_generator, 
+        validation_steps=nb_val_samples // batchSize,
+        callbacks=[earlystop])
+    plt.plot(history.history['val_acc'], 'r')
+    plt.plot(history.history['acc'], 'b')
+    plt.title('Performance of model ' + modelArchitecture)
+    plt.ylabel('Accuracy')
+    plt.xlabel('Epochs No')
+    plt.savefig(savedModelName + '_initialModel_plot.png')
+    serializeModel(model, savedModelName + "_initialModel")
+
+
+def finetuningTraining(learningRate, noOfEpochs, batchSize, savedModelName, train_generator, validation_generator, model, lr_decay):
+    # we need to recompile the model for these modifications to take effect
+    # we use SGD with a low learning rate
+    sgd = SGD(lr=learningRate, decay=lr_decay, momentum=0.9, nesterov=True)
+    model.compile(optimizer=sgd, loss='categorical_crossentropy', metrics=['accuracy'])
+    earlystop = callbacks.EarlyStopping(monitor='val_loss', min_delta=0, patience=5, mode='auto')
+    
+    # we train our model again (this time fine-tuning the top 2 inception blocks
+    # alongside the top Dense layers
+    history = model.fit_generator(
+        train_generator, 
+        steps_per_epoch=nb_train_samples // batchSize, 
+        epochs=noOfEpochs, 
+        validation_data=validation_generator, 
+        validation_steps=nb_val_samples // batchSize,
+        callbacks=[earlystop])
+    plt.clf()
+    plt.plot(history.history['val_acc'], 'r')
+    plt.plot(history.history['acc'], 'b')
+    plt.savefig(savedModelName + '_finalModel_plot.png')
+    serializeModel(model, savedModelName + "_finalModel")
+
+
+def model(learningRate, optimazerLastLayer, noOfEpochs, batchSize, savedModelName, srcImagesDir, labelsFile, modelArchitecture, dropoutRate, lr_decay):
+    
+    classes, train_generator, validation_generator = prepareDataGenerators(batchSize, srcImagesDir, labelsFile)
+
+    if modelArchitecture == 'VGG16':
+        model = getVGG16Architecture(classes, dropoutRate)
+    elif modelArchitecture == 'VGG19':
+        model = getVGG19Architecture(classes, dropoutRate)
+    else:
+        model = getInceptionV3Architecture(classes, dropoutRate)
+    
+    initialTraining(optimazerLastLayer, noOfEpochs, batchSize, savedModelName, train_generator, validation_generator, model, modelArchitecture, lr_decay, learningRate)
+    
+    setLayersToRetrain(model, modelArchitecture)
+    
+    finetuningTraining(learningRate, noOfEpochs, batchSize, savedModelName, train_generator, validation_generator, model, lr_decay)
+
+def main(args):
+    pprint(args)
+    if args.process_data:
+        dataPreprocessing(args.car_ims_dir, args.car_ims_labels)
+    
+    K.clear_session()
+
+    model(args.learning_rate, 
+          args.optimizer_last_layer, 
+          args.no_of_epochs, 
+          args.batch_size, 
+          args.saved_model_name,
+          args.car_ims_dir, 
+          args.car_ims_labels,
+          args.model,
+          args.dropout_rate,
+          args.lr_decay)
+
+def str2bool(v):
+    if v.lower() in ('yes', 'true', 't', 'y', '1'):
+        return True
+    elif v.lower() in ('no', 'false', 'f', 'n', '0'):
+        return False
+    else:
+        raise argparse.ArgumentTypeError('Boolean value expected.')
+    
+def parse_arguments(argv):
+    
+    parser = argparse.ArgumentParser()
+
+    parser.add_argument('--process_data', 
+        help='Divides the whole data set from Stanford (http://ai.stanford.edu/~jkrause/cars/car_dataset.html) into test and validation sets',
+        type=str2bool, 
+        nargs='?',
+        const=True, 
+        default=False)
+        
+    parser.add_argument('--car_ims_dir', type=str, 
+        help='Directory where all pictures are located or where subfolder train/val are located', default='./car_ims')
+
+    parser.add_argument('--car_ims_labels', type=str, 
+        help='Points to the file with all labels', default='./cars_annos.mat')    
+
+    parser.add_argument('--learning_rate', type=float,
+        help='Initial learning rate.', default=0.001)
+    
+    parser.add_argument('--dropout_rate', type=float,
+        help='Fraction of the input units to drop.', default=0.7)
+
+    parser.add_argument('--lr_decay', type=float,
+        help='Learning rate decay', default=1e-4)
+    
+    parser.add_argument('--optimizer_last_layer', type=str, choices=['ADAGRAD', 'ADADELTA', 'ADAM', 'RMSPROP', 'MOM'],
+        help='The optimization algorithm to use', default='RMSPROP')
+
+    parser.add_argument('--model', type=str, choices=['VGG19', 'VGG16', 'InceptionV3'],
+        help='The optimization algorithm to use', default='VGG16')
+
+    parser.add_argument('--no_of_epochs', type=int,
+        help='Number of epochs to run.', default=500)
+
+    parser.add_argument('--batch_size', type=int,
+        help='Number of images to process in a batch.', default=64)
+    
+    parser.add_argument('--saved_model_name', type=str,
+        help='Number of images to process in a batch.', default='carRecognition')
+    
+    return parser.parse_args(argv)
+    
+if __name__ == "__main__":
+    main(parse_arguments(sys.argv[1:]))
\ No newline at end of file

From 67d3d8410cbed59ff08474cd15f9c1d93ada46ab Mon Sep 17 00:00:00 2001
From: UditSoniJi <114995094+UditSoniJi@users.noreply.github.com>
Date: Thu, 27 Oct 2022 18:44:07 +0530
Subject: [PATCH 2/2] Add files via upload

---
 .../createTestSetFromValidation.py            | 62 +++++++++++++++++++
 1 file changed, 62 insertions(+)
 create mode 100644 Udit Soni/car-recognition-aws/createTestSetFromValidation.py

diff --git a/Udit Soni/car-recognition-aws/createTestSetFromValidation.py b/Udit Soni/car-recognition-aws/createTestSetFromValidation.py
new file mode 100644
index 0000000..ba25230
--- /dev/null
+++ b/Udit Soni/car-recognition-aws/createTestSetFromValidation.py	
@@ -0,0 +1,62 @@
+'''
+Created on 5 lut 2018
+
+@author: mgdak
+'''
+
+import os
+import argparse
+import sys
+from pprint import pprint
+import shutil as sh
+import random
+
+def main(args):
+    pprint(args)
+    
+    topDirs = os.listdir(args.src_dir)
+
+    # This would print all the files and directories
+    for classDir in topDirs:
+        print(classDir)
+        files = os.listdir(args.src_dir + "/" + classDir)
+        train_start = int(len(files)*args.train_split_ratio)
+        tst_start = int(len(files)*args.tst_split_ratio)
+        random.shuffle(files)
+        #move train data
+        for j in range(train_start + tst_start, tst_start, -1):
+            os.makedirs(args.train_dst_dir + "/" + classDir, exist_ok=True)
+            sh.move(args.src_dir + "/" + classDir + "/" + files[j],
+                    args.train_dst_dir + "/" + classDir + "/" + files[j])
+
+        #move tst data
+        for j in range(tst_start, 0, -1):
+            os.makedirs(args.tst_dst_dir + "/" + classDir, exist_ok=True)
+            sh.move(args.src_dir + "/" + classDir + "/" + files[j],
+                    args.tst_dst_dir + "/" + classDir + "/" + files[j])
+
+def parse_arguments(argv):
+    
+    parser = argparse.ArgumentParser()
+        
+    parser.add_argument('--tst_split_ratio', type=float,
+        help='Defines the split ratio from validation which has to be moved to the test'
+        , default=0.25)
+
+    parser.add_argument('--train_split_ratio', type=float,
+        help='Defines the split ratio from validation which has to be moved to the train'
+        , default=0.5)
+        
+    parser.add_argument('--tst_dst_dir', type=str,
+        help='Destination directory for splited test data set', default='./car_ims/tst')
+    
+    parser.add_argument('--train_dst_dir', type=str,
+        help='Destination directory for splited train data set', default='./car_ims/train')
+        
+    parser.add_argument('--src_dir', type=str,
+        help='Source directory where validation data set is located', default='./car_ims/val')
+    
+    return parser.parse_args(argv)
+
+if __name__ == "__main__":
+    main(parse_arguments(sys.argv[1:]))
\ No newline at end of file