28-ago

ROC da reunião anterior: https://i.imgur.com/VWg5Ydf.png (0.84 +- 0.14) (CNN-Z_2x3F_K5_256_False11)
    ROC do modelo atual: https://i.imgur.com/kGui4eb.png (0.88 +- 0.07) (CNN-Z_6x8F_K5_128_False5, 654KB)

teste com augmented data: https://i.imgur.com/hTkjsDg.png (0.84 +- 0.07)


valid melhores modelos
CNN-Z_12x8F_K5_128_True14.h5, Valid: 0.6833
CNN-Z_6x8F_K5_128_True18.h5,  Valid: 0.7069
CNN-Z_6x8F_K5_128_False5.h5,  Valid: 0.7090
CNN-Z_2x8F_K5_128_True40.h5,  Valid: 0.6787
CNN-Z_2x8F_K5_128_False15.h5, Valid: 0.7218
CNN-Z_12x8F_K3_128_True7.h5,  Valid: 0.7020
CNN-Z_6x8F_K3_128_True3.h5,   Valid: 0.7147
CNN-Z_6x8F_K3_128_False22.h5, Valid: 0.7276
CNN-Z_2x8F_K3_128_True1.h5,   Valid: 0.6930


TODO: analisar casos de erro
exemplo https://imgur.com/a/pa2r6Y7



data augmentation:

coord_offset = [-10, 10, -15, 15, -20, 20, -25, 25]
ou: [-12, 12, -24, 24]

exemplo - S2_U860_12.jpg,1478,906,2072,1500
pega as coordenadas e aplica o offset



distribuição total da amostra:

    Jersey_1 | Good: 337  | Bad: 165  | +-: 165  | 51% 25% 25% (337/667)
Jersey_1_aug | Good: 5729 | Bad: 2805 | +-: 2805 | 51% 25% 25% (5729/11339)
    Jersey_2 | Good: 116  | Bad: 122  | +-: 56   | 39% 41% 19% (116/294)
Jersey_2_aug | Good: 1856 | Bad: 1952 | +-: 896  | 39% 41% 19% (1856/4704)
    Jersey_3 | Good: 309  | Bad: 116  | +-: 73   | 62% 23% 15% (309/498)
Jersey_3_aug | Good: 5252 | Bad: 2000 | +-: 1262 | 62% 23% 15% (5252/8514)
       Total | Good: 13599 | Bad: 7160 | +-: 5257 | 52% 28% 20% (13599/26016)

      USP-P1 | Good: 238  | Bad: 45   | +-: 17   | 79% 15% 5.7% (238/300)
  USP-P1_aug | Good: 1904 | Bad: 360  | +-: 136  | 79% 15% 5.7% (1904/2400)
      USP-P2 | Good: 183  | Bad: 92   | +-: 57   | 55% 28% 17% (183/332)
  USP-P2_aug | Good: 1464 | Bad: 736  | +-: 456  | 55% 28% 17% (1464/2656)
      USP-P3 | Good: 274  | Bad: 36   | +-: 45   | 77% 10% 13% (274/355)
  USP-P3_aug | Good: 2192 | Bad: 288  | +-: 360  | 77% 10% 13% (2192/2840)
       Total | Good: 6255 | Bad: 1557 | +-: 1071 | 70% 18% 12% (6255/8883)

    Puruna_1 | Good: 54   | Bad: 776  | +-: 45   | 6.2% 89% 5.1% (54/875)
Puruna_1_aug | Good: 216  | Bad: 3104 | +-: 180  | 6.2% 89% 5.1% (216/3500)
    Puruna_2 | Good: 89   | Bad: 1132 | +-: 75   | 6.9% 87% 5.8% (89/1296)
Puruna_2_aug | Good: 356  | Bad: 4528 | +-: 300  | 6.9% 87% 5.8% (356/5184)
    Puruna_3 | Good: 70   | Bad: 1656 | +-: 67   | 3.9% 92% 3.7% (70/1793)
Puruna_3_aug | Good: 280  | Bad: 6624 | +-: 268  | 3.9% 92% 3.7% (280/7172)
       Total | Good: 1065 | Bad: 17820 | +-: 935  | 5.4% 90% 4.7% (1065/19820)

       Total | Good: 20919 | Bad: 26537 | +-: 7263 | 38% 48% 13% (20919/54719)




testes por animal (sample):

Animal: 946 | [38/80 (0.47)
Animal: F522 | [14/90 (0.16)
Animal: W303 | [23/80 (0.29)
Animal: A1115 | [19/40 (0.47)
Animal: A1080 | [48/75 (0.64)
Animal: U889 | [12/35 (0.34)
Animal: L310 | [0/40 (0.00)
Animal: 1016 | [21/30 (0.70)
Animal: F493 | [58/80 (0.72)
Animal: M796 | [20/50 (0.40)
Animal: F445 | [30/50 (0.60)
Animal: PR29 | [72/120 (0.60)
Animal: O201 | [56/75 (0.75)
Animal: 920 | [32/65 (0.49)
Animal: W915 | [46/65 (0.71)
Animal: A51 | [44/70 (0.63)
Animal: E31 | [15/15 (1.00)
Animal: O335 | [9/10 (0.90)
Animal: K237 | [10/10 (1.00)
Animal: M789 | [39/50 (0.78)
Animal: A740 | [44/50 (0.88)
Animal: P1089 | [5/5 (1.00)
Animal: MP26 | [65/75 (0.87)
Animal: W345 | [29/30 (0.97)
Animal: R526 | [21/30 (0.70)
Animal: H600 | [57/70 (0.81)
Animal: M719 | [40/40 (1.00)
Animal: P31 | [72/80 (0.90)
Animal: P649 | [59/65 (0.91)
Animal: V35 | [30/45 (0.67)
Animal: P761 | [57/85 (0.67)
Animal: E390 | [29/30 (0.97)
Animal: N764 | [78/95 (0.82)
Animal: G497 | [19/20 (0.95)
Animal: R434 | [33/35 (0.94)
Animal: N269 | [82/85 (0.96)
Animal: O411 | [37/45 (0.82)
Animal: N261 | [60/70 (0.86)
Animal: A1063 | [35/35 (1.00)
Animal: W360 | [5/5 (1.00)
Animal: H609 | [43/50 (0.86)
Animal: I238 | [21/30 (0.70)
Animal: P930 | [19/20 (0.95)
Animal: O254 | [70/80 (0.88)
Animal: K80 | [13/15 (0.87)
Animal: P819 | [34/35 (0.97)
Animal: P956 | [67/110 (0.61)
Animal: 916 | [90/100 (0.90)
Animal: M364 | [20/20 (1.00)
Animal: H550 | [15/20 (0.75)
Animal: M782 | [103/115 (0.90)
Animal: NRR79 | [39/40 (0.97)
Animal: G428 | [12/15 (0.80)
Animal: P1097 | [81/85 (0.95)
Animal: F495 | [23/25 (0.92)
Animal: N957 | [9/15 (0.60)
Animal: P689 | [49/75 (0.65)
Animal: N371 | [76/85 (0.89)
Animal: MP6 | [104/105 (0.99)
Animal: U757 | [37/40 (0.93)
Animal: A22 | [44/50 (0.88)
Animal: P703 | [30/30 (1.00)
Animal: A741 | [43/45 (0.96)
Animal: P29 | [32/70 (0.46)
Animal: W359 | [48/50 (0.96)
Animal: P1108 | [25/25 (1.00)
Animal: U767 | [23/25 (0.92)
Animal: A56 | [29/30 (0.97)
Animal: A7010 | [29/30 (0.97)
Animal: W358 | [26/30 (0.87)
Animal: O395 | [40/40 (1.00)
Animal: O428 | [42/45 (0.93)
Animal: O331 | [62/80 (0.78)
Animal: U878 | [82/85 (0.96)
Animal: H560 | [14/15 (0.93)
Animal: N979 | [40/45 (0.89)
Animal: P1105 | [19/20 (0.95)
Animal: N421 | [40/40 (1.00)
Animal: M791 | [37/40 (0.93)
Animal: O226 | [4/5 (0.80)
Animal: A1116 | [10/20 (0.50)
Animal: N1010 | [23/25 (0.92)
Animal: H558 | [24/25 (0.96)
Animal: H348 | [18/20 (0.90)
Animal: W937 | [21/25 (0.84)
Animal: M709 | [34/50 (0.68)
Animal: P954 | [10/15 (0.67)
Animal: P1113 | [4/5 (0.80)
Animal: U632 | [10/10 (1.00)
Animal: N805 | [23/25 (0.92)
Animal: N909 | [15/15 (1.00)
Animal: G501 | [15/15 (1.00)
Animal: C624 | [10/10 (1.00)
Animal: I164 | [9/10 (0.90)
Animal: P959 | [31/35 (0.89)
Animal: C641 | [5/5 (1.00)
Animal: A726 | [5/5 (1.00)
Animal: A466 | [5/5 (1.00)
Animal: N751 | [21/30 (0.70)
Animal: A1094 | [13/15 (0.87)
Animal: U884 | [5/5 (1.00)
Animal: 102 | [200/289 (0.69)
Animal: 106 | [212/272 (0.78)
Animal: 98 | [74/187 (0.40)
Animal: 92 | [136/306 (0.44)
Animal: 114 | [39/187 (0.21)
Animal: 11 | [304/391 (0.78)
Animal: 351 | [110/255 (0.43)
Animal: 99 | [114/289 (0.39)
Animal: 70 | [110/289 (0.38)
Animal: 91 | [106/357 (0.30)
Animal: 101 | [102/238 (0.43)
Animal: 55 | [44/323 (0.14)
Animal: 97 | [92/340 (0.27)
Animal: 111 | [128/238 (0.54)
Animal: 41 | [177/272 (0.65)
Animal: 107 | [66/204 (0.32)
Animal: 110 | [313/476 (0.66)
Animal: 90 | [181/238 (0.76)
Animal: 16 | [61/144 (0.42)
Animal: 5 | [99/207 (0.48)
Animal: 6 | [134/153 (0.88)
Animal: 9 | [76/189 (0.40)
Animal: 12 | [141/225 (0.63)
Animal: 4 | [160/180 (0.89)
Animal: 10 | [66/144 (0.46)
Animal: 14 | [121/153 (0.79)
Animal: 15 | [89/144 (0.62)
Animal: 8 | [65/171 (0.38)
Animal: 2 | [65/144 (0.45)
Animal: 3 | [47/99 (0.47)
Animal: 13 | [96/144 (0.67)
Animal: 17 | [36/153 (0.24)
Animal: 7 | [56/153 (0.37)
Animal: 1 | [52/144 (0.36)
EndAcc: 0.670 | EndPre: 0.567 | Sample: 12073

(pequena cópia de trecho)
Path: Nelore~USP-P1_aug/0/USP_1_2yb.jpg, Pred: 1, Prob: 1.00 | Correct: False
Path: Nelore~USP-P1_aug/0/USP_1_0yg.jpg, Pred: 1, Prob: 1.00 | Correct: False
Path: Nelore~USP-P1_aug/0/USP_7_11yi.jpg, Pred: 1, Prob: 1.00 | Correct: False
Path: Nelore~USP-P1_aug/0/USP_1_3ye.jpg, Pred: 1, Prob: 1.00 | Correct: False
Path: Nelore~USP-P1_aug/0/USP_7_12ye.jpg, Pred: 1, Prob: 1.00 | Correct: False
Path: Nelore~USP-P1_aug/0/USP_1_15yc.jpg, Pred: 1, Prob: 1.00 | Correct: False
Path: Nelore~USP-P1_aug/0/USP_1_13yc.jpg, Pred: 1, Prob: 1.00 | Correct: False
Path: Nelore~USP-P1_aug/0/USP_1_8ye.jpg, Pred: 1, Prob: 1.00 | Correct: False
Path: Nelore~USP-P1_aug/0/USP_16_3yh.jpg, Pred: 1, Prob: 1.00 | Correct: False
Path: Nelore~USP-P1_aug/0/USP_7_6yc.jpg, Pred: 1, Prob: 0.98 | Correct: False
Path: Nelore~USP-P1_aug/0/USP_16_3yi.jpg, Pred: 1, Prob: 1.00 | Correct: False
Path: Nelore~USP-P1_aug/0/USP_14_15yd.jpg, Pred: 1, Prob: 0.99 | Correct: False
Path: Nelore~USP-P1_aug/0/USP_17_10yb.jpg, Pred: 1, Prob: 0.99 | Correct: False
Path: Nelore~USP-P1_aug/0/USP_1_13ye.jpg, Pred: 1, Prob: 1.00 | Correct: False
Path: Nelore~USP-P1_aug/0/USP_7_15yd.jpg, Pred: 1, Prob: 0.98 | Correct: False
Path: Nelore~USP-P1_aug/0/USP_1_11yf.jpg, Pred: 1, Prob: 1.00 | Correct: False