TF之pix2pix：基于TF利用Facades數(shù)據(jù)集訓(xùn)練pix2pix模型、測試并進(jìn)行生成過程全記錄

?

?

目錄

TB監(jiān)控

1、SCALARS

2、IMAGES

3、GRAPHS

4、DISTRIBUTIONS

輸出結(jié)果

訓(xùn)練代碼運(yùn)行過程全記錄

測試代碼全記錄

---

?

?

?

TB監(jiān)控

1、SCALARS

2、IMAGES

inputs_summary		outputs_summary
predict_fake_summary		predict_real_summary
targets_summary		?	?

?

3、GRAPHS

4、DISTRIBUTIONS

?

?

?

輸出結(jié)果

?

?

訓(xùn)練代碼運(yùn)行過程全記錄

2 epoch：2126~2218

開始訓(xùn)練 aspect_ratio = 1.0 batch_size = 1 beta1 = 0.5 checkpoint = None display_freq = 0 flip = True gan_weight = 1.0 input_dir = facades/train l1_weight = 100.0 lab_colorization = False lr = 0.0002 max_epochs = 200 max_steps = None mode = train ndf = 64 ngf = 64 output_dir = facades_train output_filetype = png progress_freq = 50 save_freq = 5000 scale_size = 286 seed = 407313043 summary_freq = 100 trace_freq = 0 which_direction = BtoA2018-10-07 21:26:49.558601: parameter_count = 57183616 progress epoch 1 step 50 image/sec 0.4 remaining 35m discrim_loss 0.59409106 gen_loss_GAN 0.3667453 gen_loss_L1 0.14627346 recording summary progress epoch 1 step 100 image/sec 0.3 remaining 33m discrim_loss 0.77973217 gen_loss_GAN 0.74193096 gen_loss_L1 0.21620709 progress epoch 1 step 150 image/sec 0.4 remaining 30m discrim_loss 0.7097481 gen_loss_GAN 1.2259353 gen_loss_L1 0.27181715 recording summary progress epoch 1 step 200 image/sec 0.4 remaining 28m discrim_loss 0.6670909 gen_loss_GAN 1.5955695 gen_loss_L1 0.30515844 progress epoch 1 step 250 image/sec 0.4 remaining 25m discrim_loss 0.5505945 gen_loss_GAN 1.897398 gen_loss_L1 0.3195855 recording summary progress epoch 1 step 300 image/sec 0.4 remaining 23m discrim_loss 0.54358536 gen_loss_GAN 2.1270702 gen_loss_L1 0.33635142 progress epoch 1 step 350 image/sec 0.4 remaining 21m discrim_loss 0.53915083 gen_loss_GAN 2.234504 gen_loss_L1 0.33556485 recording summary progress epoch 1 step 400 image/sec 0.4 remaining 18m discrim_loss 0.5494336 gen_loss_GAN 2.349324 gen_loss_L1 0.33941123 progress epoch 2 step 50 image/sec 0.4 remaining 16m discrim_loss 0.5763757 gen_loss_GAN 2.3618762 gen_loss_L1 0.3413253 recording summary progress epoch 2 step 100 image/sec 0.4 remaining 14m discrim_loss 0.63876843 gen_loss_GAN 2.2650375 gen_loss_L1 0.3409966 progress epoch 2 step 150 image/sec 0.4 remaining 11m discrim_loss 0.6011929 gen_loss_GAN 2.264903 gen_loss_L1 0.34726414 recording summary progress epoch 2 step 200 image/sec 0.4 remaining 9m discrim_loss 0.59052 gen_loss_GAN 2.302569 gen_loss_L1 0.3522855 progress epoch 2 step 250 image/sec 0.3 remaining 7m discrim_loss 0.57109314 gen_loss_GAN 2.324084 gen_loss_L1 0.35149702 recording summary progress epoch 2 step 300 image/sec 0.3 remaining 4m discrim_loss 0.4946928 gen_loss_GAN 2.5188313 gen_loss_L1 0.3564302 progress epoch 2 step 350 image/sec 0.3 remaining 2m discrim_loss 0.5365153 gen_loss_GAN 2.5414586 gen_loss_L1 0.35425124 recording summary progress epoch 2 step 400 image/sec 0.3 remaining 0m discrim_loss 0.56210524 gen_loss_GAN 2.5018184 gen_loss_L1 0.35051015 saving model

?

測試代碼全記錄

開始測試 loaded lab_colorization = False loaded ndf = 64 loaded ngf = 64 loaded which_direction = BtoA aspect_ratio = 1.0 batch_size = 1 beta1 = 0.5 checkpoint = facades_train display_freq = 0 flip = False gan_weight = 1.0 input_dir = facades/val l1_weight = 100.0 lab_colorization = False lr = 0.0002 max_epochs = None max_steps = None mode = test ndf = 64 ngf = 64 output_dir = facades_test output_filetype = png progress_freq = 50 save_freq = 5000 scale_size = 256 seed = 651085994 summary_freq = 100 trace_freq = 0 which_direction = BtoAexamples count = 100 parameter_count = 57183616 loading model from checkpoint evaluated image 1 evaluated image 2 evaluated image 3 evaluated image 4 evaluated image 5 evaluated image 6 evaluated image 7 evaluated image 8 evaluated image 9 evaluated image 10 evaluated image 11 evaluated image 12 evaluated image 13 evaluated image 14 evaluated image 15 evaluated image 16 evaluated image 17 evaluated image 18 evaluated image 19 evaluated image 20 evaluated image 21 evaluated image 22 evaluated image 23 evaluated image 24 evaluated image 25 evaluated image 26 evaluated image 27 evaluated image 28 evaluated image 29 evaluated image 30 evaluated image 31 evaluated image 32 evaluated image 33 evaluated image 34 evaluated image 35 evaluated image 36 evaluated image 37 evaluated image 38 evaluated image 39 evaluated image 40 evaluated image 41 evaluated image 42 evaluated image 43 evaluated image 44 evaluated image 45 evaluated image 46 evaluated image 47 evaluated image 48 evaluated image 49 evaluated image 50 evaluated image 51 evaluated image 52 evaluated image 53 evaluated image 54 evaluated image 55 evaluated image 56 evaluated image 57 evaluated image 58 evaluated image 59 evaluated image 60 evaluated image 61 evaluated image 62 evaluated image 63 evaluated image 64 evaluated image 65 evaluated image 66 evaluated image 67 evaluated image 68 evaluated image 69 evaluated image 70 evaluated image 71 evaluated image 72 evaluated image 73 evaluated image 74 evaluated image 75 evaluated image 76 evaluated image 77 evaluated image 78 evaluated image 79 evaluated image 80 evaluated image 81 evaluated image 82 evaluated image 83 evaluated image 84 evaluated image 85 evaluated image 86 evaluated image 87 evaluated image 88 evaluated image 89 evaluated image 90 evaluated image 91 evaluated image 92 evaluated image 93 evaluated image 94 evaluated image 95 evaluated image 96 evaluated image 97 evaluated image 98 evaluated image 99 evaluated image 100 wrote index at facades_test\index.html

?

?

?

?

?

?

?

?

?

總結(jié)

以上是生活随笔為你收集整理的TF之pix2pix：基于TF利用Facades数据集训练pix2pix模型、测试并进行生成过程全记录的全部內(nèi)容，希望文章能夠幫你解決所遇到的問題。

如果覺得生活随笔網(wǎng)站內(nèi)容還不錯(cuò)，歡迎將生活随笔推薦給好友。