Implement: STJL model training

STJL Version(Market1501+Duke)

global step 1000 : loss =  6.6439295 ( 0.12295246124267578 sec/step)
global step 10000 : loss =  4.989618 ( 0.12686705589294434 sec/step)
global step 20000 : loss =  4.7869434 ( 0.11679577827453613 sec/step)
global step 30000 : loss =  4.7598543 ( 0.12518644332885742 sec/step)
global step 39000 : loss =  4.6826773 ( 0.11968255043029785 sec/step)
global step 40000 : loss =  4.5441303 ( 0.14246749877929688 sec/step)
global step 49000 : loss =  4.6776476 ( 0.12763547897338867 sec/step)
global step 50000 : loss =  4.804378 ( 0.12951159477233887 sec/step)
global step 60000 : loss =  4.659108 ( 0.1263430118560791 sec/step)
global step 70000 : loss =  4.623295 ( 0.13466882705688477 sec/step)
global step 80000 : loss =  4.5173893 ( 0.13434863090515137 sec/step)
global step 90000 : loss =  4.4886193 ( 0.12137150764465332 sec/step)
global step 100000 : loss =  4.53279 ( 0.1429002285003662 sec/step)
global step 110000 : loss =  4.58663 ( 0.12114930152893066 sec/step)
global step 120000 : loss =  4.568182 ( 0.12872099876403809 sec/step)
global step 130000 : loss =  4.494215 ( 0.1232907772064209 sec/step)
global step 140000 : loss =  4.501731 ( 0.13323330879211426 sec/step)
global step 150000 : loss =  4.4914536 ( 0.12923026084899902 sec/step)
global step 160000 : loss =  4.5258636 ( 0.11639213562011719 sec/step)
global step 170000 : loss =  4.551574 ( 0.12645745277404785 sec/step)

model_0

  • single query: mAP = 0.559442, r1 precision = 0.779988, r5 precision = 0.903207(60899)
  • single query: mAP = 0.566793, r1 precision = 0.780285, r5 precision = 0.910629(81354)
  • single query: mAP = 0.566399, r1 precision = 0.786223, r5 precision = 0.909739(90544)
  • single query: mAP = 0.571777, r1 precision = 0.790974, r5 precision = 0.910629(100000)
  • single query: mAP = 0.569773, r1 precision = 0.778800, r5 precision = 0.910926(140202)
  • single query: mAP = 0.577163, r1 precision = 0.789192, r5 precision = 0.917162(149677)
  • single query: mAP = 0.576389, r1 precision = 0.793943, r5 precision = 0.912114(152046)
  • single query: mAP = 0.572002, r1 precision = 0.783254, r5 precision = 0.911817(154414)
  • single query: mAP = 0.575645, r1 precision = 0.793052, r5 precision = 0.912411(170000)

model_1

  • single query: mAP = 0.565738, r1 precision = 0.784442, r5 precision = 0.904097(100000)
  • single query: mAP = 0.570187, r1 precision = 0.786223, r5 precision = 0.908254(150000)

model_2

  • single query: mAP = 0.560733, r1 precision = 0.778207, r5 precision = 0.905879(100000)

model_3

  • single query: mAP = 0.567453, r1 precision = 0.786817, r5 precision = 0.910926(100000)
  • single query: mAP = 0.580884, r1 precision = 0.794240, r5 precision = 0.911817(150000)


STJL Version(Market1501+CUHK03)

global step 22000 : loss =  4.9631658 ( 0.12742328643798828 sec/step)
global step 30000 : loss =  4.909469 ( 0.1299915313720703 sec/step)
global step 40000 : loss =  4.7835627 ( 0.12056827545166016 sec/step)
global step 50000 : loss =  4.7345533 ( 0.11754298210144043 sec/step)
global step 60000 : loss =  4.7335835 ( 0.12260150909423828 sec/step)
global step 70000 : loss =  4.7573004 ( 0.11776137351989746 sec/step)
global step 80000 : loss =  4.716571 ( 0.12890052795410156 sec/step)
global step 90000 : loss =  4.7394953 ( 0.11937379837036133 sec/step)
global step 100000 : loss =  4.654492 ( 0.11723089218139648 sec/step)
global step 110000 : loss =  4.6596985 ( 0.12869596481323242 sec/step)
global step 120000 : loss =  4.5664363 ( 0.11740231513977051 sec/step)
global step 130000 : loss =  4.6673927 ( 0.12252688407897949 sec/step)
global step 140000 : loss =  4.6504707 ( 0.12412047386169434 sec/step)
global step 150000 : loss =  4.665951 ( 0.12623000144958496 sec/step)
global step 160000 : loss =  4.7147837 ( 0.12892818450927734 sec/step)
global step 170000 : loss =  4.6009784 ( 0.11673927307128906 sec/step)
  • single query: mAP = 0.553986, r1 precision = 0.770487, r5 precision = 0.899050(71714)
  • single query: mAP = 0.556472, r1 precision = 0.775534, r5 precision = 0.902019(81915)
  • single query: mAP = 0.562012, r1 precision = 0.779988, r5 precision = 0.905582(92170)
  • single query: mAP = 0.560011, r1 precision = 0.776128, r5 precision = 0.905879(100000)
  • single query: mAP = 0.561778, r1 precision = 0.780879, r5 precision = 0.902613(128182)
  • single query: mAP = 0.565374, r1 precision = 0.779097, r5 precision = 0.902019(138474)
  • single query: mAP = 0.569650, r1 precision = 0.786817, r5 precision = 0.912708(143622)
  • single query: mAP = 0.567477, r1 precision = 0.782957, r5 precision = 0.902316(150000)
  • single query: mAP = 0.571565, r1 precision = 0.788599, r5 precision = 0.905285(160256)
  • single query: mAP = 0.570951, r1 precision = 0.785926, r5 precision = 0.908848(170000)


STJL Version(Market1501+Duke+CUHK03)

global step 1000 : loss =  7.0658684 ( 0.11960005760192871 sec/step)
global step 10000 : loss =  5.598107 ( 0.12567615509033203 sec/step)
global step 20000 : loss =  5.342411 ( 0.12709975242614746 sec/step)
global step 30000 : loss =  5.3258862 ( 0.1216890811920166 sec/step)
global step 40000 : loss =  5.193895 ( 0.12981653213500977 sec/step)
global step 50000 : loss =  5.201321 ( 0.12166690826416016 sec/step)
global step 60000 : loss =  5.2354083 ( 0.12090516090393066 sec/step))
global step 70000 : loss =  5.1289186 ( 0.13580751419067383 sec/step)
global step 80000 : loss =  5.14295 ( 0.12595915794372559 sec/step)
global step 90000 : loss =  5.0982275 ( 0.12741780281066895 sec/step)
global step 100000 : loss =  4.992791 ( 0.12177371978759766 sec/step)
global step 110000 : loss =  5.1492605 ( 0.12134599685668945 sec/step)
global step 120000 : loss =  4.983779 ( 0.1263585090637207 sec/step)
global step 130000 : loss =  4.9668055 ( 0.12137293815612793 sec/step)
global step 140000 : loss =  5.0940356 ( 0.11754059791564941 sec/step)
global step 150000 : loss =  5.0502443 ( 0.11811256408691406 sec/step)
global step 160000 : loss =  5.048631 ( 0.13070178031921387 sec/step)
global step 170000 : loss =  5.012842 ( 0.132490873336792 sec/step)
global step 180000 : loss =  5.063532 ( 0.13053083419799805 sec/step)
global step 190000 : loss =  5.001734 ( 0.12381863594055176 sec/step)
global step 200000 : loss =  5.0319147 ( 0.13091158866882324 sec/step)
global step 210000 : loss =  5.1374083 ( 0.12871623039245605 sec/step)
global step 220000 : loss =  5.124153 ( 0.12740373611450195 sec/step)
global step 230000 : loss =  4.8920755 ( 0.12305855751037598 sec/step)
global step 240000 : loss =  4.968564 ( 0.12494254112243652 sec/step)
global step 250000 : loss =  4.946948 ( 0.1261603832244873 sec/step)
global step 260000 : loss =  5.0086174 ( 0.11992311477661133 sec/step)
global step 270000 : loss =  4.991828 ( 0.12592720985412598 sec/step)
global step 280000 : loss =  5.0628977 ( 0.12367725372314453 sec/step)
global step 290000 : loss =  4.9893484 ( 0.12531042098999023 sec/step)
global step 300000 : loss =  4.959217 ( 0.12381482124328613 sec/step)

model_0

  • single query: mAP = 0.504263, r1 precision = 0.736639, r5 precision = 0.881829(51855)
  • single query: mAP = 0.515017, r1 precision = 0.750297, r5 precision = 0.891330(72629)
  • single query: mAP = 0.524238, r1 precision = 0.759204, r5 precision = 0.891924(77813)
  • single query: mAP = 0.517201, r1 precision = 0.742280, r5 precision = 0.886580(88162)
  • single query: mAP = 0.528697, r1 precision = 0.756829, r5 precision = 0.888361(100000)
  • single query: mAP = 0.545325, r1 precision = 0.778800, r5 precision = 0.900534(149136)
  • single query: mAP = 0.557859, r1 precision = 0.778207, r5 precision = 0.909145(183625)
  • single query: mAP = 0.555436, r1 precision = 0.782660, r5 precision = 0.906176(193470)
  • single query: mAP = 0.554574, r1 precision = 0.773456, r5 precision = 0.900238(200000)
  • single query: mAP = 0.561017, r1 precision = 0.781770, r5 precision = 0.907363(229503)
  • single query: mAP = 0.562381, r1 precision = 0.783254, r5 precision = 0.907363(249246)
  • single query: mAP = 0.554391, r1 precision = 0.772565, r5 precision = 0.900534(259113)
  • single query: mAP = 0.554389, r1 precision = 0.779988, r5 precision = 0.897862(273916)
  • single query: mAP = 0.566512, r1 precision = 0.785036, r5 precision = 0.906770(278853)
  • single query: mAP = 0.556057, r1 precision = 0.777910, r5 precision = 0.902613(283753)
  • single query: mAP = 0.561916, r1 precision = 0.781770, r5 precision = 0.904097(288730)
  • single query: mAP = 0.564381, r1 precision = 0.792162, r5 precision = 0.909739(293663)
  • single query: mAP = 0.559526, r1 precision = 0.778800, r5 precision = 0.901128(298601)
  • single query: mAP = 0.566351, r1 precision = 0.790974, r5 precision = 0.908254(300000)

model_1

global step 1000 : loss =  7.130766 ( 0.11600422859191895 sec/step)
global step 10000 : loss =  5.7195334 ( 0.12104916572570801 sec/step)
global step 20000 : loss =  5.3503776 ( 0.11776971817016602 sec/step)
global step 30000 : loss =  5.217521 ( 0.12575197219848633 sec/step)
global step 40000 : loss =  5.3631873 ( 0.12670445442199707 sec/step)
global step 50000 : loss =  5.1465087 ( 0.12709450721740723 sec/step)
global step 60000 : loss =  5.193937 ( 0.12746381759643555 sec/step)
global step 70000 : loss =  5.206148 ( 0.11620140075683594 sec/step)
global step 80000 : loss =  5.1347294 ( 0.1254732608795166 sec/step)
global step 90000 : loss =  5.1385555 ( 0.11708283424377441 sec/step)
global step 100000 : loss =  4.945298 ( 0.12205672264099121 sec/step)
global step 110000 : loss =  5.1626496 ( 0.12178158760070801 sec/step)
global step 120000 : loss =  5.1032085 ( 0.12511992454528809 sec/step)
global step 130000 : loss =  5.016778 ( 0.1231393814086914 sec/step)
global step 140000 : loss =  5.065547 ( 0.1253063678741455 sec/step)
global step 150000 : loss =  5.0726404 ( 0.1224985122680664 sec/step)
global step 287000 : loss =  4.9665976 ( 0.1260669231414795 sec/step)
global step 288000 : loss =  4.980357 ( 0.12383174896240234 sec/step)
global step 290000 : loss =  5.0085516 ( 0.13026738166809082 sec/step)
global step 294000 : loss =  4.959462 ( 0.12028694152832031 sec/step)
global step 295000 : loss =  4.9149723 ( 0.12395691871643066 sec/step)
global step 300000 : loss =  5.061846 ( 0.1275169849395752 sec/step)
global step 301000 : loss =  4.9229937 ( 0.13216423988342285 sec/step)
global step 303000 : loss =  4.935828 ( 0.11971735954284668 sec/step)
global step 305000 : loss =  4.9140754 ( 0.12793803215026855 sec/step)
global step 310000 : loss =  4.8928385 ( 0.1263582706451416 sec/step)
global step 313000 : loss =  4.955532 ( 0.1397106647491455 sec/step)
global step 318000 : loss =  4.916205 ( 0.12558603286743164 sec/step)
global step 320000 : loss =  4.924187 ( 0.12659668922424316 sec/step)
global step 323000 : loss =  4.9170866 ( 0.12105250358581543 sec/step)
global step 330000 : loss =  5.043866 ( 0.11905956268310547 sec/step)
global step 333000 : loss =  4.920663 ( 0.12226295471191406 sec/step)
global step 336000 : loss =  4.9310083 ( 0.13434505462646484 sec/step)
global step 338000 : loss =  4.897778 ( 0.11655092239379883 sec/step)
global step 340000 : loss =  4.936147 ( 0.12731051445007324 sec/step)
global step 346000 : loss =  4.8914347 ( 0.13010644912719727 sec/step)
global step 350000 : loss =  5.0154543 ( 0.12137627601623535 sec/step)
  • single query: mAP = 0.535280, r1 precision = 0.753860, r5 precision = 0.891627(98843)
  • single query: mAP = 0.548744, r1 precision = 0.766033, r5 precision = 0.894596(148703)
  • single query: mAP = 0.544599, r1 precision = 0.757126, r5 precision = 0.889549(150000)
  • single query: mAP = 0.552475, r1 precision = 0.766924, r5 precision = 0.897268(250000)
  • single query: mAP = 0.555462, r1 precision = 0.764252, r5 precision = 0.897862(294778)
  • single query: mAP = 0.555270, r1 precision = 0.771971, r5 precision = 0.899941(299736)
  • single query: mAP = 0.564761, r1 precision = 0.774941, r5 precision = 0.902910(304690)
  • single query: mAP = 0.561744, r1 precision = 0.772565, r5 precision = 0.901722(350000)

model_2

global step 1000 : loss =  7.168634 ( 0.13558173179626465 sec/step)
global step 10000 : loss =  5.528709 ( 0.13562488555908203 sec/step)
global step 20000 : loss =  5.414777 ( 0.1252753734588623 sec/step)
global step 30000 : loss =  5.2316475 ( 0.13341617584228516 sec/step)
global step 40000 : loss =  5.1650457 ( 0.12183976173400879 sec/step)
global step 50000 : loss =  5.2512503 ( 0.1226351261138916 sec/step)
global step 60000 : loss =  5.1913385 ( 0.12853550910949707 sec/step)
global step 70000 : loss =  5.021592 ( 0.12180423736572266 sec/step)
global step 80000 : loss =  5.152728 ( 0.11937189102172852 sec/step)
global step 90000 : loss =  5.162696 ( 0.12498116493225098 sec/step)
global step 100000 : loss =  5.135518 ( 0.13256406784057617 sec/step)
global step 110000 : loss =  5.097643 ( 0.12162947654724121 sec/step)
global step 120000 : loss =  5.1232347 ( 0.1366441249847412 sec/step)
global step 130000 : loss =  4.968847 ( 0.12698817253112793 sec/step)
global step 140000 : loss =  5.092561 ( 0.12272047996520996 sec/step)
global step 150000 : loss =  5.06645 ( 0.13303208351135254 sec/step)
global step 160000 : loss =  4.9991093 ( 0.12385416030883789 sec/step)
global step 170000 : loss =  4.979229 ( 0.1217641830444336 sec/step)
global step 180000 : loss =  5.0365458 ( 0.12199831008911133 sec/step)
global step 190000 : loss =  5.1220293 ( 0.11827874183654785 sec/step)
global step 200000 : loss =  4.9968386 ( 0.1278698444366455 sec/step)
global step 210000 : loss =  5.062701 ( 0.1260511875152588 sec/step)
global step 220000 : loss =  4.9834375 ( 0.12623143196105957 sec/step)
global step 230000 : loss =  4.938182 ( 0.12402153015136719 sec/step)
global step 240000 : loss =  4.976327 ( 0.1277775764465332 sec/step)
global step 250000 : loss =  4.957697 ( 0.11791801452636719 sec/step)
global step 260000 : loss =  4.965317 ( 0.11903858184814453 sec/step)
global step 270000 : loss =  5.0312824 ( 0.12018823623657227 sec/step)
global step 280000 : loss =  5.021184 ( 0.12349724769592285 sec/step)
global step 290000 : loss =  4.9589043 ( 0.1296100616455078 sec/step)
global step 300000 : loss =  4.9984503 ( 0.1328732967376709 sec/step)
global step 310000 : loss =  4.9793367 ( 0.12455058097839355 sec/step)
global step 320000 : loss =  5.003694 ( 0.12512612342834473 sec/step)
global step 330000 : loss =  4.966714 ( 0.1330428123474121 sec/step)
global step 340000 : loss =  4.99354 ( 0.1234135627746582 sec/step)
global step 350000 : loss =  4.9873514 ( 0.12108373641967773 sec/step)
  • single query: mAP = 0.545830, r1 precision = 0.761876, r5 precision = 0.898753(250978)
  • single query: mAP = 0.559310, r1 precision = 0.773159, r5 precision = 0.904691(290375)
  • single query: mAP = 0.551953, r1 precision = 0.769299, r5 precision = 0.897268 (300229)
  • single query: mAP = 0.555358, r1 precision = 0.768112, r5 precision = 0.898753(329784)
  • single query: mAP = 0.552775, r1 precision = 0.762173, r5 precision = 0.895190 (350000)

model_3

global step 1000 : loss =  7.04678 ( 0.2009446620941162 sec/step)
global step 10000 : loss =  5.3856635 ( 0.18761372566223145 sec/step)
global step 20000 : loss =  5.22773 ( 0.19115328788757324 sec/step)
global step 30000 : loss =  5.063499 ( 0.19239163398742676 sec/step)
global step 40000 : loss =  5.0455146 ( 0.19614768028259277 sec/step)
global step 50000 : loss =  5.045067 ( 0.19063067436218262 sec/step)
global step 60000 : loss =  5.046233 ( 0.19148945808410645 sec/step)
global step 70000 : loss =  5.086785 ( 0.18630337715148926 sec/step)
global step 80000 : loss =  5.026585 ( 0.1829357147216797 sec/step)
global step 90000 : loss =  4.9325647 ( 0.19374656677246094 sec/step)
global step 100000 : loss =  4.9528813 ( 0.19474244117736816 sec/step)
global step 110000 : loss =  4.92362 ( 0.18029260635375977 sec/step)
global step 120000 : loss =  5.008074 ( 0.20298123359680176 sec/step)
global step 130000 : loss =  4.9537654 ( 0.18552637100219727 sec/step)
global step 140000 : loss =  4.953067 ( 0.19675636291503906 sec/step)
global step 150000 : loss =  4.915715 ( 0.18825840950012207 sec/step)
global step 160000 : loss =  4.847957 ( 0.19228315353393555 sec/step)
global step 170000 : loss =  4.9900312 ( 0.19379043579101562 sec/step)
global step 180000 : loss =  4.9385934 ( 0.1805553436279297 sec/step)
global step 190000 : loss =  4.887682 ( 0.19129061698913574 sec/step)
global step 191000 : loss =  4.85817 ( 0.19269442558288574 sec/step)
global step 197000 : loss =  4.854458 ( 0.19942831993103027 sec/step)
global step 199000 : loss =  4.869013 ( 0.19484639167785645 sec/step)
global step 200000 : loss =  4.923344 ( 0.18692803382873535 sec/step)
  • single query: mAP = 0.537164, r1 precision = 0.763361, r5 precision = 0.896378(109060)
  • single query: mAP = 0.549250, r1 precision = 0.769002, r5 precision = 0.902613(131521)
  • single query: mAP = 0.547300, r1 precision = 0.770487, r5 precision = 0.902613(150839)
  • single query: mAP = 0.554844, r1 precision = 0.776425, r5 precision = 0.903504(154056)
  • single query: mAP = 0.551156, r1 precision = 0.769002, r5 precision = 0.894596(157261)
  • single query: mAP = 0.549954, r1 precision = 0.766924, r5 precision = 0.901128(160473)
  • single query: mAP = 0.551575, r1 precision = 0.767518, r5 precision = 0.901722(166887)
  • single query: mAP = 0.551306, r1 precision = 0.769893, r5 precision = 0.899050(170092)
  • single query: mAP = 0.555924, r1 precision = 0.771081, r5 precision = 0.904394(173298)
  • single query: mAP = 0.559138, r1 precision = 0.776425, r5 precision = 0.904394(176504)
  • single query: mAP = 0.553942, r1 precision = 0.773753, r5 precision = 0.907957(179717)
  • single query: mAP = 0.554061, r1 precision = 0.769002, r5 precision = 0.900831(182921)
  • single query: mAP = 0.552965, r1 precision = 0.769299, r5 precision = 0.902910(189335)
  • single query: mAP = 0.557376, r1 precision = 0.772268, r5 precision = 0.903800(192548)
  • single query: mAP = 0.554273, r1 precision = 0.774941, r5 precision = 0.899644(195760)
  • single query: mAP = 0.558506, r1 precision = 0.774941, r5 precision = 0.902316(198973)
  • single query: mAP = 0.556326, r1 precision = 0.777613, r5 precision = 0.905582(200000)

model_4

  • decrease cuhk03 train image, see cuhk03-replace2
global step 1000 : loss =  6.868814 ( 0.1978294849395752 sec/step)
global step 10000 : loss =  5.195675 ( 0.1843864917755127 sec/step)
global step 20000 : loss =  4.985582 ( 0.19842743873596191 sec/step)
global step 30000 : loss =  4.934865 ( 0.18244123458862305 sec/step)
global step 40000 : loss =  4.9312634 ( 0.1845855712890625 sec/step)
global step 50000 : loss =  4.958419 ( 0.20286035537719727 sec/step)
global step 60000 : loss =  4.8272038 ( 0.20019769668579102 sec/step)
global step 70000 : loss =  4.9248657 ( 0.19191217422485352 sec/step)
global step 71000 : loss =  4.9062996 ( 0.1883537769317627 sec/step)
global step 72000 : loss =  4.815995 ( 0.19318318367004395 sec/step)
global step 73000 : loss =  4.8316536 ( 0.1977691650390625 sec/step)
global step 74000 : loss =  4.8287992 ( 0.19582104682922363 sec/step)
global step 75000 : loss =  4.83317 ( 0.2025148868560791 sec/step)
global step 76000 : loss =  4.849081 ( 0.18788599967956543 sec/step)
global step 77000 : loss =  4.7891703 ( 0.1795940399169922 sec/step)
global step 78000 : loss =  4.8238544 ( 0.18093132972717285 sec/step)
global step 79000 : loss =  4.8750515 ( 0.18540596961975098 sec/step)
global step 80000 : loss =  4.861011 ( 0.18457961082458496 sec/step)
global step 81000 : loss =  4.7869496 ( 0.19111156463623047 sec/step)
global step 82000 : loss =  4.8510237 ( 0.19002652168273926 sec/step)
global step 83000 : loss =  4.828524 ( 0.18643689155578613 sec/step)
global step 84000 : loss =  4.778386 ( 0.19179940223693848 sec/step)
global step 85000 : loss =  4.815672 ( 0.20000672340393066 sec/step)
global step 86000 : loss =  4.809256 ( 0.18834304809570312 sec/step)
global step 87000 : loss =  4.78335 ( 0.19159603118896484 sec/step)
global step 88000 : loss =  4.806515 ( 0.17940425872802734 sec/step)
global step 89000 : loss =  4.856204 ( 0.1884472370147705 sec/step)
global step 90000 : loss =  4.7324634 ( 0.19449877738952637 sec/step)
global step 100000 : loss =  4.866296 ( 0.19835376739501953 sec/step)
global step 110000 : loss =  4.8441086 ( 0.18153643608093262 sec/step)
global step 120000 : loss =  4.7972007 ( 0.1905515193939209 sec/step)
global step 130000 : loss =  4.682397 ( 0.20338201522827148 sec/step)
global step 140000 : loss =  4.7430344 ( 0.19087910652160645 sec/step)
global step 150000 : loss =  4.771241 ( 0.19053220748901367 sec/step)
global step 160000 : loss =  4.757398 ( 0.19220209121704102 sec/step)
global step 170000 : loss =  4.78076 ( 0.21085572242736816 sec/step)
global step 180000 : loss =  4.7323074 ( 1.0468299388885498 sec/step)
global step 190000 : loss =  4.779081 ( 0.19121479988098145 sec/step)
global step 200000 : loss =  4.7093287 ( 0.20177030563354492 sec/step)
global step 247000 : loss =  4.668269 ( 0.19473552703857422 sec/step)
global step 250000 : loss =  4.760028 ( 0.20276761054992676 sec/step)
  • single query: mAP = 0.531866, r1 precision = 0.752375, r5 precision = 0.895784(51130)
  • single query: mAP = 0.529515, r1 precision = 0.754751, r5 precision = 0.894299(60769)
  • single query: mAP = 0.549877, r1 precision = 0.771081, r5 precision = 0.903207(70423)
  • single query: mAP = 0.552362, r1 precision = 0.773159, r5 precision = 0.903504(80070)
  • single query: mAP = 0.556278, r1 precision = 0.778504, r5 precision = 0.901425(92856)
  • single query: mAP = 0.555275, r1 precision = 0.774644, r5 precision = 0.905582(102412)
  • single query: mAP = 0.551166, r1 precision = 0.767518, r5 precision = 0.899050(112049)
  • single query: mAP = 0.555813, r1 precision = 0.779691, r5 precision = 0.899941(115268)
  • single query: mAP = 0.554484, r1 precision = 0.782957, r5 precision = 0.900831(118481)
  • single query: mAP = 0.563170, r1 precision = 0.784145, r5 precision = 0.905582(121706)
  • single query: mAP = 0.558799, r1 precision = 0.779097, r5 precision = 0.905285(124922)
  • single query: mAP = 0.555233, r1 precision = 0.772565, r5 precision = 0.896971(128143)
  • single query: mAP = 0.560554, r1 precision = 0.774347, r5 precision = 0.906176(131359)
  • single query: mAP = 0.558923, r1 precision = 0.777316, r5 precision = 0.903504(141037)
  • single query: mAP = 0.559966, r1 precision = 0.776128, r5 precision = 0.905582(150000)
  • single query: mAP = 0.559821, r1 precision = 0.778504, r5 precision = 0.904394(182150)
  • single query: mAP = 0.559085, r1 precision = 0.774050, r5 precision = 0.903504(201513)
  • single query: mAP = 0.564209, r1 precision = 0.779691, r5 precision = 0.905285(237119)
  • single query: mAP = 0.564584, r1 precision = 0.784442, r5 precision = 0.906176(240353)
  • single query: mAP = 0.565926, r1 precision = 0.780285, r5 precision = 0.903207(246821)
  • single query: mAP = 0.563388, r1 precision = 0.779394, r5 precision = 0.907660(250000)

model_5

  • single query: mAP = 0.557474, r1 precision = 0.771675, r5 precision = 0.901722(250000)

model_6

  • single query: mAP = 0.558671, r1 precision = 0.777910, r5 precision = 0.900831(250000)

model_7

  • single query: mAP = 0.561001, r1 precision = 0.780582, r5 precision = 0.904691(250000)

model_8

  • single query: mAP = 0.560639, r1 precision = 0.774347, r5 precision = 0.900534(250000)

model_9

  • single query: mAP = 0.561676, r1 precision = 0.782363, r5 precision = 0.906473(179301)
  • single query: mAP = 0.567241, r1 precision = 0.782957, r5 precision = 0.904394(239088)
  • single query: mAP = 0.564340, r1 precision = 0.781473, r5 precision = 0.904394(244065)
  • single query: mAP = 0.571256, r1 precision = 0.788302, r5 precision = 0.913302(250000)

model_11

  • cuhk-replace3
  • single query: mAP = 0.555267, r1 precision = 0.775238, r5 precision = 0.905285(251050)
  • single query: mAP = 0.562045, r1 precision = 0.781176, r5 precision = 0.904691(300000)

n(Market1501+Duke)

global step 1000 : loss =  6.6439295 ( 0.12295246124267578 sec/step)
global step 10000 : loss =  4.989618 ( 0.12686705589294434 sec/step)
global step 20000 : loss =  4.7869434 ( 0.11679577827453613 sec/step)
global step 30000 : loss =  4.7598543 ( 0.12518644332885742 sec/step)
global step 39000 : loss =  4.6826773 ( 0.11968255043029785 sec/step)
global step 40000 : loss =  4.5441303 ( 0.14246749877929688 sec/step)
global step 49000 : loss =  4.6776476 ( 0.12763547897338867 sec/step)
global step 50000 : loss =  4.804378 ( 0.12951159477233887 sec/step)
global step 60000 : loss =  4.659108 ( 0.1263430118560791 sec/step)
global step 70000 : loss =  4.623295 ( 0.13466882705688477 sec/step)
global step 80000 : loss =  4.5173893 ( 0.13434863090515137 sec/step)
global step 90000 : loss =  4.4886193 ( 0.12137150764465332 sec/step)
global step 100000 : loss =  4.53279 ( 0.1429002285003662 sec/step)
global step 110000 : loss =  4.58663 ( 0.12114930152893066 sec/step)
global step 120000 : loss =  4.568182 ( 0.12872099876403809 sec/step)
global step 130000 : loss =  4.494215 ( 0.1232907772064209 sec/step)
global step 140000 : loss =  4.501731 ( 0.13323330879211426 sec/step)
global step 150000 : loss =  4.4914536 ( 0.12923026084899902 sec/step)
global step 160000 : loss =  4.5258636 ( 0.11639213562011719 sec/step)
global step 170000 : loss =  4.551574 ( 0.12645745277404785 sec/step)

model_0

  • single query: mAP = 0.559442, r1 precision = 0.779988, r5 precision = 0.903207(60899)
  • single query: mAP = 0.566793, r1 precision = 0.780285, r5 precision = 0.910629(81354)
  • single query: mAP = 0.566399, r1 precision = 0.786223, r5 precision = 0.909739(90544)
  • single query: mAP = 0.571777, r1 precision = 0.790974, r5 precision = 0.910629(100000)
  • single query: mAP = 0.569773, r1 precision = 0.778800, r5 precision = 0.910926(140202)
  • single query: mAP = 0.577163, r1 precision = 0.789192, r5 precision = 0.917162(149677)
  • single query: mAP = 0.576389, r1 precision = 0.793943, r5 precision = 0.912114(152046)
  • single query: mAP = 0.572002, r1 precision = 0.783254, r5 precision = 0.911817(154414)
  • single query: mAP = 0.575645, r1 precision = 0.793052, r5 precision = 0.912411(170000)

model_1

  • single query: mAP = 0.565738, r1 precision = 0.784442, r5 precision = 0.904097(100000)
  • single query: mAP = 0.570187, r1 precision = 0.786223, r5 precision = 0.908254(150000)

model_2

  • single query: mAP = 0.560733, r1 precision = 0.778207, r5 precision = 0.905879(100000)

model_3

  • single query: mAP = 0.567453, r1 precision = 0.786817, r5 precision = 0.910926(100000)
  • single query: mAP = 0.580884, r1 precision = 0.794240, r5 precision = 0.911817(150000)


STJL Version(Market1501+CUHK03)

global step 22000 : loss =  4.9631658 ( 0.12742328643798828 sec/step)
global step 30000 : loss =  4.909469 ( 0.1299915313720703 sec/step)
global step 40000 : loss =  4.7835627 ( 0.12056827545166016 sec/step)
global step 50000 : loss =  4.7345533 ( 0.11754298210144043 sec/step)
global step 60000 : loss =  4.7335835 ( 0.12260150909423828 sec/step)
global step 70000 : loss =  4.7573004 ( 0.11776137351989746 sec/step)
global step 80000 : loss =  4.716571 ( 0.12890052795410156 sec/step)
global step 90000 : loss =  4.7394953 ( 0.11937379837036133 sec/step)
global step 100000 : loss =  4.654492 ( 0.11723089218139648 sec/step)
global step 110000 : loss =  4.6596985 ( 0.12869596481323242 sec/step)
global step 120000 : loss =  4.5664363 ( 0.11740231513977051 sec/step)
global step 130000 : loss =  4.6673927 ( 0.12252688407897949 sec/step)
global step 140000 : loss =  4.6504707 ( 0.12412047386169434 sec/step)
global step 150000 : loss =  4.665951 ( 0.12623000144958496 sec/step)
global step 160000 : loss =  4.7147837 ( 0.12892818450927734 sec/step)
global step 170000 : loss =  4.6009784 ( 0.11673927307128906 sec/step)
  • single query: mAP = 0.553986, r1 precision = 0.770487, r5 precision = 0.899050(71714)
  • single query: mAP = 0.556472, r1 precision = 0.775534, r5 precision = 0.902019(81915)
  • single query: mAP = 0.562012, r1 precision = 0.779988, r5 precision = 0.905582(92170)
  • single query: mAP = 0.560011, r1 precision = 0.776128, r5 precision = 0.905879(100000)
  • single query: mAP = 0.561778, r1 precision = 0.780879, r5 precision = 0.902613(128182)
  • single query: mAP = 0.565374, r1 precision = 0.779097, r5 precision = 0.902019(138474)
  • single query: mAP = 0.569650, r1 precision = 0.786817, r5 precision = 0.912708(143622)
  • single query: mAP = 0.567477, r1 precision = 0.782957, r5 precision = 0.902316(150000)
  • single query: mAP = 0.571565, r1 precision = 0.788599, r5 precision = 0.905285(160256)
  • single query: mAP = 0.570951, r1 precision = 0.785926, r5 precision = 0.908848(170000)


STJL Version(Market1501+Duke+CUHK03)

global step 1000 : loss =  7.0658684 ( 0.11960005760192871 sec/step)
global step 10000 : loss =  5.598107 ( 0.12567615509033203 sec/step)
global step 20000 : loss =  5.342411 ( 0.12709975242614746 sec/step)
global step 30000 : loss =  5.3258862 ( 0.1216890811920166 sec/step)
global step 40000 : loss =  5.193895 ( 0.12981653213500977 sec/step)
global step 50000 : loss =  5.201321 ( 0.12166690826416016 sec/step)
global step 60000 : loss =  5.2354083 ( 0.12090516090393066 sec/step))
global step 70000 : loss =  5.1289186 ( 0.13580751419067383 sec/step)
global step 80000 : loss =  5.14295 ( 0.12595915794372559 sec/step)
global step 90000 : loss =  5.0982275 ( 0.12741780281066895 sec/step)
global step 100000 : loss =  4.992791 ( 0.12177371978759766 sec/step)
global step 110000 : loss =  5.1492605 ( 0.12134599685668945 sec/step)
global step 120000 : loss =  4.983779 ( 0.1263585090637207 sec/step)
global step 130000 : loss =  4.9668055 ( 0.12137293815612793 sec/step)
global step 140000 : loss =  5.0940356 ( 0.11754059791564941 sec/step)
global step 150000 : loss =  5.0502443 ( 0.11811256408691406 sec/step)
global step 160000 : loss =  5.048631 ( 0.13070178031921387 sec/step)
global step 170000 : loss =  5.012842 ( 0.132490873336792 sec/step)
global step 180000 : loss =  5.063532 ( 0.13053083419799805 sec/step)
global step 190000 : loss =  5.001734 ( 0.12381863594055176 sec/step)
global step 200000 : loss =  5.0319147 ( 0.13091158866882324 sec/step)
global step 210000 : loss =  5.1374083 ( 0.12871623039245605 sec/step)
global step 220000 : loss =  5.124153 ( 0.12740373611450195 sec/step)
global step 230000 : loss =  4.8920755 ( 0.12305855751037598 sec/step)
global step 240000 : loss =  4.968564 ( 0.12494254112243652 sec/step)
global step 250000 : loss =  4.946948 ( 0.1261603832244873 sec/step)
global step 260000 : loss =  5.0086174 ( 0.11992311477661133 sec/step)
global step 270000 : loss =  4.991828 ( 0.12592720985412598 sec/step)
global step 280000 : loss =  5.0628977 ( 0.12367725372314453 sec/step)
global step 290000 : loss =  4.9893484 ( 0.12531042098999023 sec/step)
global step 300000 : loss =  4.959217 ( 0.12381482124328613 sec/step)

model_0

  • single query: mAP = 0.504263, r1 precision = 0.736639, r5 precision = 0.881829(51855)
  • single query: mAP = 0.515017, r1 precision = 0.750297, r5 precision = 0.891330(72629)
  • single query: mAP = 0.524238, r1 precision = 0.759204, r5 precision = 0.891924(77813)
  • single query: mAP = 0.517201, r1 precision = 0.742280, r5 precision = 0.886580(88162)
  • single query: mAP = 0.528697, r1 precision = 0.756829, r5 precision = 0.888361(100000)
  • single query: mAP = 0.545325, r1 precision = 0.778800, r5 precision = 0.900534(149136)
  • single query: mAP = 0.557859, r1 precision = 0.778207, r5 precision = 0.909145(183625)
  • single query: mAP = 0.555436, r1 precision = 0.782660, r5 precision = 0.906176(193470)
  • single query: mAP = 0.554574, r1 precision = 0.773456, r5 precision = 0.900238(200000)
  • single query: mAP = 0.561017, r1 precision = 0.781770, r5 precision = 0.907363(229503)
  • single query: mAP = 0.562381, r1 precision = 0.783254, r5 precision = 0.907363(249246)
  • single query: mAP = 0.554391, r1 precision = 0.772565, r5 precision = 0.900534(259113)
  • single query: mAP = 0.554389, r1 precision = 0.779988, r5 precision = 0.897862(273916)
  • single query: mAP = 0.566512, r1 precision = 0.785036, r5 precision = 0.906770(278853)
  • single query: mAP = 0.556057, r1 precision = 0.777910, r5 precision = 0.902613(283753)
  • single query: mAP = 0.561916, r1 precision = 0.781770, r5 precision = 0.904097(288730)
  • single query: mAP = 0.564381, r1 precision = 0.792162, r5 precision = 0.909739(293663)
  • single query: mAP = 0.559526, r1 precision = 0.778800, r5 precision = 0.901128(298601)
  • single query: mAP = 0.566351, r1 precision = 0.790974, r5 precision = 0.908254(300000)

model_1

global step 1000 : loss =  7.130766 ( 0.11600422859191895 sec/step)
global step 10000 : loss =  5.7195334 ( 0.12104916572570801 sec/step)
global step 20000 : loss =  5.3503776 ( 0.11776971817016602 sec/step)
global step 30000 : loss =  5.217521 ( 0.12575197219848633 sec/step)
global step 40000 : loss =  5.3631873 ( 0.12670445442199707 sec/step)
global step 50000 : loss =  5.1465087 ( 0.12709450721740723 sec/step)
global step 60000 : loss =  5.193937 ( 0.12746381759643555 sec/step)
global step 70000 : loss =  5.206148 ( 0.11620140075683594 sec/step)
global step 80000 : loss =  5.1347294 ( 0.1254732608795166 sec/step)
global step 90000 : loss =  5.1385555 ( 0.11708283424377441 sec/step)
global step 100000 : loss =  4.945298 ( 0.12205672264099121 sec/step)
global step 110000 : loss =  5.1626496 ( 0.12178158760070801 sec/step)
global step 120000 : loss =  5.1032085 ( 0.12511992454528809 sec/step)
global step 130000 : loss =  5.016778 ( 0.1231393814086914 sec/step)
global step 140000 : loss =  5.065547 ( 0.1253063678741455 sec/step)
global step 150000 : loss =  5.0726404 ( 0.1224985122680664 sec/step)
global step 287000 : loss =  4.9665976 ( 0.1260669231414795 sec/step)
global step 288000 : loss =  4.980357 ( 0.12383174896240234 sec/step)
global step 290000 : loss =  5.0085516 ( 0.13026738166809082 sec/step)
global step 294000 : loss =  4.959462 ( 0.12028694152832031 sec/step)
global step 295000 : loss =  4.9149723 ( 0.12395691871643066 sec/step)
global step 300000 : loss =  5.061846 ( 0.1275169849395752 sec/step)
global step 301000 : loss =  4.9229937 ( 0.13216423988342285 sec/step)
global step 303000 : loss =  4.935828 ( 0.11971735954284668 sec/step)
global step 305000 : loss =  4.9140754 ( 0.12793803215026855 sec/step)
global step 310000 : loss =  4.8928385 ( 0.1263582706451416 sec/step)
global step 313000 : loss =  4.955532 ( 0.1397106647491455 sec/step)
global step 318000 : loss =  4.916205 ( 0.12558603286743164 sec/step)
global step 320000 : loss =  4.924187 ( 0.12659668922424316 sec/step)
global step 323000 : loss =  4.9170866 ( 0.12105250358581543 sec/step)
global step 330000 : loss =  5.043866 ( 0.11905956268310547 sec/step)
global step 333000 : loss =  4.920663 ( 0.12226295471191406 sec/step)
global step 336000 : loss =  4.9310083 ( 0.13434505462646484 sec/step)
global step 338000 : loss =  4.897778 ( 0.11655092239379883 sec/step)
global step 340000 : loss =  4.936147 ( 0.12731051445007324 sec/step)
global step 346000 : loss =  4.8914347 ( 0.13010644912719727 sec/step)
global step 350000 : loss =  5.0154543 ( 0.12137627601623535 sec/step)
  • single query: mAP = 0.535280, r1 precision = 0.753860, r5 precision = 0.891627(98843)
  • single query: mAP = 0.548744, r1 precision = 0.766033, r5 precision = 0.894596(148703)
  • single query: mAP = 0.544599, r1 precision = 0.757126, r5 precision = 0.889549(150000)
  • single query: mAP = 0.552475, r1 precision = 0.766924, r5 precision = 0.897268(250000)
  • single query: mAP = 0.555462, r1 precision = 0.764252, r5 precision = 0.897862(294778)
  • single query: mAP = 0.555270, r1 precision = 0.771971, r5 precision = 0.899941(299736)
  • single query: mAP = 0.564761, r1 precision = 0.774941, r5 precision = 0.902910(304690)
  • single query: mAP = 0.561744, r1 precision = 0.772565, r5 precision = 0.901722(350000)

model_2

global step 1000 : loss =  7.168634 ( 0.13558173179626465 sec/step)
global step 10000 : loss =  5.528709 ( 0.13562488555908203 sec/step)
global step 20000 : loss =  5.414777 ( 0.1252753734588623 sec/step)
global step 30000 : loss =  5.2316475 ( 0.13341617584228516 sec/step)
global step 40000 : loss =  5.1650457 ( 0.12183976173400879 sec/step)
global step 50000 : loss =  5.2512503 ( 0.1226351261138916 sec/step)
global step 60000 : loss =  5.1913385 ( 0.12853550910949707 sec/step)
global step 70000 : loss =  5.021592 ( 0.12180423736572266 sec/step)
global step 80000 : loss =  5.152728 ( 0.11937189102172852 sec/step)
global step 90000 : loss =  5.162696 ( 0.12498116493225098 sec/step)
global step 100000 : loss =  5.135518 ( 0.13256406784057617 sec/step)
global step 110000 : loss =  5.097643 ( 0.12162947654724121 sec/step)
global step 120000 : loss =  5.1232347 ( 0.1366441249847412 sec/step)
global step 130000 : loss =  4.968847 ( 0.12698817253112793 sec/step)
global step 140000 : loss =  5.092561 ( 0.12272047996520996 sec/step)
global step 150000 : loss =  5.06645 ( 0.13303208351135254 sec/step)
global step 160000 : loss =  4.9991093 ( 0.12385416030883789 sec/step)
global step 170000 : loss =  4.979229 ( 0.1217641830444336 sec/step)
global step 180000 : loss =  5.0365458 ( 0.12199831008911133 sec/step)
global step 190000 : loss =  5.1220293 ( 0.11827874183654785 sec/step)
global step 200000 : loss =  4.9968386 ( 0.1278698444366455 sec/step)
global step 210000 : loss =  5.062701 ( 0.1260511875152588 sec/step)
global step 220000 : loss =  4.9834375 ( 0.12623143196105957 sec/step)
global step 230000 : loss =  4.938182 ( 0.12402153015136719 sec/step)
global step 240000 : loss =  4.976327 ( 0.1277775764465332 sec/step)
global step 250000 : loss =  4.957697 ( 0.11791801452636719 sec/step)
global step 260000 : loss =  4.965317 ( 0.11903858184814453 sec/step)
global step 270000 : loss =  5.0312824 ( 0.12018823623657227 sec/step)
global step 280000 : loss =  5.021184 ( 0.12349724769592285 sec/step)
global step 290000 : loss =  4.9589043 ( 0.1296100616455078 sec/step)
global step 300000 : loss =  4.9984503 ( 0.1328732967376709 sec/step)
global step 310000 : loss =  4.9793367 ( 0.12455058097839355 sec/step)
global step 320000 : loss =  5.003694 ( 0.12512612342834473 sec/step)
global step 330000 : loss =  4.966714 ( 0.1330428123474121 sec/step)
global step 340000 : loss =  4.99354 ( 0.1234135627746582 sec/step)
global step 350000 : loss =  4.9873514 ( 0.12108373641967773 sec/step)
  • single query: mAP = 0.545830, r1 precision = 0.761876, r5 precision = 0.898753(250978)
  • single query: mAP = 0.559310, r1 precision = 0.773159, r5 precision = 0.904691(290375)
  • single query: mAP = 0.551953, r1 precision = 0.769299, r5 precision = 0.897268 (300229)
  • single query: mAP = 0.555358, r1 precision = 0.768112, r5 precision = 0.898753(329784)
  • single query: mAP = 0.552775, r1 precision = 0.762173, r5 precision = 0.895190 (350000)

model_3

global step 1000 : loss =  7.04678 ( 0.2009446620941162 sec/step)
global step 10000 : loss =  5.3856635 ( 0.18761372566223145 sec/step)
global step 20000 : loss =  5.22773 ( 0.19115328788757324 sec/step)
global step 30000 : loss =  5.063499 ( 0.19239163398742676 sec/step)
global step 40000 : loss =  5.0455146 ( 0.19614768028259277 sec/step)
global step 50000 : loss =  5.045067 ( 0.19063067436218262 sec/step)
global step 60000 : loss =  5.046233 ( 0.19148945808410645 sec/step)
global step 70000 : loss =  5.086785 ( 0.18630337715148926 sec/step)
global step 80000 : loss =  5.026585 ( 0.1829357147216797 sec/step)
global step 90000 : loss =  4.9325647 ( 0.19374656677246094 sec/step)
global step 100000 : loss =  4.9528813 ( 0.19474244117736816 sec/step)
global step 110000 : loss =  4.92362 ( 0.18029260635375977 sec/step)
global step 120000 : loss =  5.008074 ( 0.20298123359680176 sec/step)
global step 130000 : loss =  4.9537654 ( 0.18552637100219727 sec/step)
global step 140000 : loss =  4.953067 ( 0.19675636291503906 sec/step)
global step 150000 : loss =  4.915715 ( 0.18825840950012207 sec/step)
global step 160000 : loss =  4.847957 ( 0.19228315353393555 sec/step)
global step 170000 : loss =  4.9900312 ( 0.19379043579101562 sec/step)
global step 180000 : loss =  4.9385934 ( 0.1805553436279297 sec/step)
global step 190000 : loss =  4.887682 ( 0.19129061698913574 sec/step)
global step 191000 : loss =  4.85817 ( 0.19269442558288574 sec/step)
global step 197000 : loss =  4.854458 ( 0.19942831993103027 sec/step)
global step 199000 : loss =  4.869013 ( 0.19484639167785645 sec/step)
global step 200000 : loss =  4.923344 ( 0.18692803382873535 sec/step)
  • single query: mAP = 0.537164, r1 precision = 0.763361, r5 precision = 0.896378(109060)
  • single query: mAP = 0.549250, r1 precision = 0.769002, r5 precision = 0.902613(131521)
  • single query: mAP = 0.547300, r1 precision = 0.770487, r5 precision = 0.902613(150839)
  • single query: mAP = 0.554844, r1 precision = 0.776425, r5 precision = 0.903504(154056)
  • single query: mAP = 0.551156, r1 precision = 0.769002, r5 precision = 0.894596(157261)
  • single query: mAP = 0.549954, r1 precision = 0.766924, r5 precision = 0.901128(160473)
  • single query: mAP = 0.551575, r1 precision = 0.767518, r5 precision = 0.901722(166887)
  • single query: mAP = 0.551306, r1 precision = 0.769893, r5 precision = 0.899050(170092)
  • single query: mAP = 0.555924, r1 precision = 0.771081, r5 precision = 0.904394(173298)
  • single query: mAP = 0.559138, r1 precision = 0.776425, r5 precision = 0.904394(176504)
  • single query: mAP = 0.553942, r1 precision = 0.773753, r5 precision = 0.907957(179717)
  • single query: mAP = 0.554061, r1 precision = 0.769002, r5 precision = 0.900831(182921)
  • single query: mAP = 0.552965, r1 precision = 0.769299, r5 precision = 0.902910(189335)
  • single query: mAP = 0.557376, r1 precision = 0.772268, r5 precision = 0.903800(192548)
  • single query: mAP = 0.554273, r1 precision = 0.774941, r5 precision = 0.899644(195760)
  • single query: mAP = 0.558506, r1 precision = 0.774941, r5 precision = 0.902316(198973)
  • single query: mAP = 0.556326, r1 precision = 0.777613, r5 precision = 0.905582(200000)

model_4

  • decrease cuhk03 train image, see cuhk03-replace2
global step 1000 : loss =  6.868814 ( 0.1978294849395752 sec/step)
global step 10000 : loss =  5.195675 ( 0.1843864917755127 sec/step)
global step 20000 : loss =  4.985582 ( 0.19842743873596191 sec/step)
global step 30000 : loss =  4.934865 ( 0.18244123458862305 sec/step)
global step 40000 : loss =  4.9312634 ( 0.1845855712890625 sec/step)
global step 50000 : loss =  4.958419 ( 0.20286035537719727 sec/step)
global step 60000 : loss =  4.8272038 ( 0.20019769668579102 sec/step)
global step 70000 : loss =  4.9248657 ( 0.19191217422485352 sec/step)
global step 71000 : loss =  4.9062996 ( 0.1883537769317627 sec/step)
global step 72000 : loss =  4.815995 ( 0.19318318367004395 sec/step)
global step 73000 : loss =  4.8316536 ( 0.1977691650390625 sec/step)
global step 74000 : loss =  4.8287992 ( 0.19582104682922363 sec/step)
global step 75000 : loss =  4.83317 ( 0.2025148868560791 sec/step)
global step 76000 : loss =  4.849081 ( 0.18788599967956543 sec/step)
global step 77000 : loss =  4.7891703 ( 0.1795940399169922 sec/step)
global step 78000 : loss =  4.8238544 ( 0.18093132972717285 sec/step)
global step 79000 : loss =  4.8750515 ( 0.18540596961975098 sec/step)
global step 80000 : loss =  4.861011 ( 0.18457961082458496 sec/step)
global step 81000 : loss =  4.7869496 ( 0.19111156463623047 sec/step)
global step 82000 : loss =  4.8510237 ( 0.19002652168273926 sec/step)
global step 83000 : loss =  4.828524 ( 0.18643689155578613 sec/step)
global step 84000 : loss =  4.778386 ( 0.19179940223693848 sec/step)
global step 85000 : loss =  4.815672 ( 0.20000672340393066 sec/step)
global step 86000 : loss =  4.809256 ( 0.18834304809570312 sec/step)
global step 87000 : loss =  4.78335 ( 0.19159603118896484 sec/step)
global step 88000 : loss =  4.806515 ( 0.17940425872802734 sec/step)
global step 89000 : loss =  4.856204 ( 0.1884472370147705 sec/step)
global step 90000 : loss =  4.7324634 ( 0.19449877738952637 sec/step)
global step 100000 : loss =  4.866296 ( 0.19835376739501953 sec/step)
global step 110000 : loss =  4.8441086 ( 0.18153643608093262 sec/step)
global step 120000 : loss =  4.7972007 ( 0.1905515193939209 sec/step)
global step 130000 : loss =  4.682397 ( 0.20338201522827148 sec/step)
global step 140000 : loss =  4.7430344 ( 0.19087910652160645 sec/step)
global step 150000 : loss =  4.771241 ( 0.19053220748901367 sec/step)
global step 160000 : loss =  4.757398 ( 0.19220209121704102 sec/step)
global step 170000 : loss =  4.78076 ( 0.21085572242736816 sec/step)
global step 180000 : loss =  4.7323074 ( 1.0468299388885498 sec/step)
global step 190000 : loss =  4.779081 ( 0.19121479988098145 sec/step)
global step 200000 : loss =  4.7093287 ( 0.20177030563354492 sec/step)
global step 247000 : loss =  4.668269 ( 0.19473552703857422 sec/step)
global step 250000 : loss =  4.760028 ( 0.20276761054992676 sec/step)
  • single query: mAP = 0.531866, r1 precision = 0.752375, r5 precision = 0.895784(51130)
  • single query: mAP = 0.529515, r1 precision = 0.754751, r5 precision = 0.894299(60769)
  • single query: mAP = 0.549877, r1 precision = 0.771081, r5 precision = 0.903207(70423)
  • single query: mAP = 0.552362, r1 precision = 0.773159, r5 precision = 0.903504(80070)
  • single query: mAP = 0.556278, r1 precision = 0.778504, r5 precision = 0.901425(92856)
  • single query: mAP = 0.555275, r1 precision = 0.774644, r5 precision = 0.905582(102412)
  • single query: mAP = 0.551166, r1 precision = 0.767518, r5 precision = 0.899050(112049)
  • single query: mAP = 0.555813, r1 precision = 0.779691, r5 precision = 0.899941(115268)
  • single query: mAP = 0.554484, r1 precision = 0.782957, r5 precision = 0.900831(118481)
  • single query: mAP = 0.563170, r1 precision = 0.784145, r5 precision = 0.905582(121706)
  • single query: mAP = 0.558799, r1 precision = 0.779097, r5 precision = 0.905285(124922)
  • single query: mAP = 0.555233, r1 precision = 0.772565, r5 precision = 0.896971(128143)
  • single query: mAP = 0.560554, r1 precision = 0.774347, r5 precision = 0.906176(131359)
  • single query: mAP = 0.558923, r1 precision = 0.777316, r5 precision = 0.903504(141037)
  • single query: mAP = 0.559966, r1 precision = 0.776128, r5 precision = 0.905582(150000)
  • single query: mAP = 0.559821, r1 precision = 0.778504, r5 precision = 0.904394(182150)
  • single query: mAP = 0.559085, r1 precision = 0.774050, r5 precision = 0.903504(201513)
  • single query: mAP = 0.564209, r1 precision = 0.779691, r5 precision = 0.905285(237119)
  • single query: mAP = 0.564584, r1 precision = 0.784442, r5 precision = 0.906176(240353)
  • single query: mAP = 0.565926, r1 precision = 0.780285, r5 precision = 0.903207(246821)
  • single query: mAP = 0.563388, r1 precision = 0.779394, r5 precision = 0.907660(250000)

model_5

  • single query: mAP = 0.557474, r1 precision = 0.771675, r5 precision = 0.901722(250000)

model_6

  • single query: mAP = 0.558671, r1 precision = 0.777910, r5 precision = 0.900831(250000)

model_7

  • single query: mAP = 0.561001, r1 precision = 0.780582, r5 precision = 0.904691(250000)

model_8

  • single query: mAP = 0.560639, r1 precision = 0.774347, r5 precision = 0.900534(250000)

model_9

  • single query: mAP = 0.561676, r1 precision = 0.782363, r5 precision = 0.906473(179301)
  • single query: mAP = 0.567241, r1 precision = 0.782957, r5 precision = 0.904394(239088)
  • single query: mAP = 0.564340, r1 precision = 0.781473, r5 precision = 0.904394(244065)
  • single query: mAP = 0.571256, r1 precision = 0.788302, r5 precision = 0.913302(250000)

model_11

  • cuhk-replace3
  • single query: mAP = 0.555267, r1 precision = 0.775238, r5 precision = 0.905285(251050)
  • single query: mAP = 0.562045, r1 precision = 0.781176, r5 precision = 0.904691(300000)