StackGAN++: Realistic Image Synthesis with Stacked Generative Adversarial Networks
The following implementation is tested on
- Python 3.7.10
- Tensorflow 2.5.0
- Keras 2.5.0
It is recommended that the latest stable releases of the following libraries/scripting langauge be used for stable performance.
- Python
- Tensorflow
- Keras
- PIL
- pickle
- Numpy
- Pandas
- Matplotlib
In the Bash Shell, run:
$ python3 main.py --epoch1 'No of Epochs for stage 1' --epoch2 'No of Epochs for stage 2'NOTE: For running on Colab Notebook use the following command:
!git clone link-to-repo
%run main.py --epoch1 'No of Epochs for stage 1' --epoch2 'No of Epochs for stage 2Generative Adversarial Networks (GANs) have been there ever since Goodfellow et al introduced it in Generative Adversarial Networks in 2014. But their training had been a challenging task. The training process is usually unstable and sensitive to the choices of hyper-parameters. Moreover,a common failure phenomenon for GANs training is mode collapse, where many of the generated samples contain the same color or texture pattern. To stabilize the trainin of GANs and generate high resolution photo-realistic images, Han Zhang et al. introduced Stacked Generative Adversarial Networks in 2016. This work is an implementation of StackGAN-v1 (Version 1 of the StackGAN introduced by authors) in the research paper which aims to generate photo-realistic images from text description.
In the paper, a two-stage generative adversarial network, StackGAN-v1, to generate images from text descriptions Low-resolution is proposed. Images are first generated by Stage-I GAN. On top of this, Stage-II GAN is stacked to generate high- resolution (e.g., 256×256) images. By conditioning on the Stage-I result and the text again, Stage-II GAN learns to capture the text information that is omitted by Stage-I GAN and draws more details. Further, a novel Conditioning Augmentation (CA) technique to encourage smoothness in the latent conditioning manifold here is given. It allows small random perturbations in the conditioning manifold and increases the diversity of synthesized images.
The whole GAN is divided into two stacked models.
- Stage-I GAN: It sketches the primitive shape and basic colors of the object conditioned on the given text description, and draws the background layout from a random noise vector, yielding a low-resolution (64*64) image.
- Stage-II GAN: It corrects defects in the low-resolution image from Stage-I and completes details of the object by reading the text description again, producing a highresolution (256*256) photo-realistic image.
The authors trained the model for 600 epochs on stage 1 GAN and 600 epochs on stage 2 GAN and obtained the following results.
- Official Paper: https://arxiv.org/pdf/1612.03242.pdf
- Authors: Han Zhang,Tao Xu,Hongsheng Li,Shaoting Zhang,Xiaogang Wang,Xiaolei Huang,Dimitris Metaxas
Other References of work in this project are following:-
-
Official Github Code: (https://github.com/hanzhanggit/StackGAN)
-
Medium Blog: (https://medium.com/@mrgarg.rajat/implementing-stackgan-using-keras-a0a1b381125e)
Model: "dis_stage1"
__________________________________________________________________________________________________
Layer (type) Output Shape Param # Connected to
==================================================================================================
input_1 (InputLayer) [(None, 64, 64, 3)] 0
__________________________________________________________________________________________________
zero_padding2d (ZeroPadding2D) (None, 66, 66, 3) 0 input_1[0][0]
__________________________________________________________________________________________________
conv2d (Conv2D) (None, 32, 32, 64) 3136 zero_padding2d[0][0]
__________________________________________________________________________________________________
leaky_re_lu (LeakyReLU) (None, 32, 32, 64) 0 conv2d[0][0]
__________________________________________________________________________________________________
zero_padding2d_1 (ZeroPadding2D (None, 34, 34, 64) 0 leaky_re_lu[0][0]
__________________________________________________________________________________________________
conv2d_1 (Conv2D) (None, 16, 16, 128) 131200 zero_padding2d_1[0][0]
__________________________________________________________________________________________________
batch_normalization (BatchNorma (None, 16, 16, 128) 512 conv2d_1[0][0]
__________________________________________________________________________________________________
leaky_re_lu_1 (LeakyReLU) (None, 16, 16, 128) 0 batch_normalization[0][0]
__________________________________________________________________________________________________
zero_padding2d_2 (ZeroPadding2D (None, 18, 18, 128) 0 leaky_re_lu_1[0][0]
__________________________________________________________________________________________________
conv2d_2 (Conv2D) (None, 8, 8, 256) 524544 zero_padding2d_2[0][0]
__________________________________________________________________________________________________
batch_normalization_1 (BatchNor (None, 8, 8, 256) 1024 conv2d_2[0][0]
__________________________________________________________________________________________________
leaky_re_lu_2 (LeakyReLU) (None, 8, 8, 256) 0 batch_normalization_1[0][0]
__________________________________________________________________________________________________
input_2 (InputLayer) [(None, 1024)] 0
__________________________________________________________________________________________________
zero_padding2d_3 (ZeroPadding2D (None, 10, 10, 256) 0 leaky_re_lu_2[0][0]
__________________________________________________________________________________________________
dense (Dense) (None, 128) 131200 input_2[0][0]
__________________________________________________________________________________________________
conv2d_3 (Conv2D) (None, 4, 4, 512) 2097664 zero_padding2d_3[0][0]
__________________________________________________________________________________________________
re_lu (ReLU) (None, 128) 0 dense[0][0]
__________________________________________________________________________________________________
batch_normalization_2 (BatchNor (None, 4, 4, 512) 2048 conv2d_3[0][0]
__________________________________________________________________________________________________
tf.reshape (TFOpLambda) (None, 1, 1, 128) 0 re_lu[0][0]
__________________________________________________________________________________________________
leaky_re_lu_3 (LeakyReLU) (None, 4, 4, 512) 0 batch_normalization_2[0][0]
__________________________________________________________________________________________________
tf.tile (TFOpLambda) (None, 4, 4, 128) 0 tf.reshape[0][0]
__________________________________________________________________________________________________
concatenate (Concatenate) (None, 4, 4, 640) 0 leaky_re_lu_3[0][0]
tf.tile[0][0]
__________________________________________________________________________________________________
conv2d_4 (Conv2D) (None, 4, 4, 512) 328192 concatenate[0][0]
__________________________________________________________________________________________________
batch_normalization_3 (BatchNor (None, 4, 4, 512) 2048 conv2d_4[0][0]
__________________________________________________________________________________________________
leaky_re_lu_4 (LeakyReLU) (None, 4, 4, 512) 0 batch_normalization_3[0][0]
__________________________________________________________________________________________________
flatten (Flatten) (None, 8192) 0 leaky_re_lu_4[0][0]
__________________________________________________________________________________________________
dense_1 (Dense) (None, 1) 8193 flatten[0][0]
__________________________________________________________________________________________________
activation (Activation) (None, 1) 0 dense_1[0][0]
==================================================================================================
Total params: 3,229,761
Trainable params: 0
Non-trainable params: 3,229,761
__________________________________________________________________________________________________
Model: "gen_stage1"
__________________________________________________________________________________________________
Layer (type) Output Shape Param # Connected to
==================================================================================================
input_3 (InputLayer) [(None, 1024)] 0
__________________________________________________________________________________________________
dense_2 (Dense) (None, 256) 262400 input_3[0][0]
__________________________________________________________________________________________________
leaky_re_lu_5 (LeakyReLU) (None, 256) 0 dense_2[0][0]
__________________________________________________________________________________________________
lambda (Lambda) (None, 128) 0 leaky_re_lu_5[0][0]
__________________________________________________________________________________________________
input_4 (InputLayer) [(None, 100)] 0
__________________________________________________________________________________________________
concatenate_1 (Concatenate) (None, 228) 0 lambda[0][0]
input_4[0][0]
__________________________________________________________________________________________________
dense_3 (Dense) (None, 16384) 3751936 concatenate_1[0][0]
__________________________________________________________________________________________________
re_lu_1 (ReLU) (None, 16384) 0 dense_3[0][0]
__________________________________________________________________________________________________
reshape (Reshape) (None, 4, 4, 1024) 0 re_lu_1[0][0]
__________________________________________________________________________________________________
up_sampling2d (UpSampling2D) (None, 8, 8, 1024) 0 reshape[0][0]
__________________________________________________________________________________________________
conv2d_5 (Conv2D) (None, 8, 8, 512) 4719104 up_sampling2d[0][0]
__________________________________________________________________________________________________
batch_normalization_4 (BatchNor (None, 8, 8, 512) 2048 conv2d_5[0][0]
__________________________________________________________________________________________________
re_lu_2 (ReLU) (None, 8, 8, 512) 0 batch_normalization_4[0][0]
__________________________________________________________________________________________________
up_sampling2d_1 (UpSampling2D) (None, 16, 16, 512) 0 re_lu_2[0][0]
__________________________________________________________________________________________________
conv2d_6 (Conv2D) (None, 16, 16, 256) 1179904 up_sampling2d_1[0][0]
__________________________________________________________________________________________________
batch_normalization_5 (BatchNor (None, 16, 16, 256) 1024 conv2d_6[0][0]
__________________________________________________________________________________________________
re_lu_3 (ReLU) (None, 16, 16, 256) 0 batch_normalization_5[0][0]
__________________________________________________________________________________________________
up_sampling2d_2 (UpSampling2D) (None, 32, 32, 256) 0 re_lu_3[0][0]
__________________________________________________________________________________________________
conv2d_7 (Conv2D) (None, 32, 32, 128) 295040 up_sampling2d_2[0][0]
__________________________________________________________________________________________________
batch_normalization_6 (BatchNor (None, 32, 32, 128) 512 conv2d_7[0][0]
__________________________________________________________________________________________________
re_lu_4 (ReLU) (None, 32, 32, 128) 0 batch_normalization_6[0][0]
__________________________________________________________________________________________________
up_sampling2d_3 (UpSampling2D) (None, 64, 64, 128) 0 re_lu_4[0][0]
__________________________________________________________________________________________________
conv2d_8 (Conv2D) (None, 64, 64, 64) 73792 up_sampling2d_3[0][0]
__________________________________________________________________________________________________
batch_normalization_7 (BatchNor (None, 64, 64, 64) 256 conv2d_8[0][0]
__________________________________________________________________________________________________
re_lu_5 (ReLU) (None, 64, 64, 64) 0 batch_normalization_7[0][0]
__________________________________________________________________________________________________
conv2d_9 (Conv2D) (None, 64, 64, 3) 1731 re_lu_5[0][0]
__________________________________________________________________________________________________
activation_1 (Activation) (None, 64, 64, 3) 0 conv2d_9[0][0]
==================================================================================================
Total params: 10,287,747
Trainable params: 10,285,827
Non-trainable params: 1,920
__________________________________________________________________________________________________
Model: "dis_stage2"
__________________________________________________________________________________________________
Layer (type) Output Shape Param # Connected to
==================================================================================================
input_8 (InputLayer) [(None, 256, 256, 3) 0
__________________________________________________________________________________________________
zero_padding2d_4 (ZeroPadding2D (None, 258, 258, 3) 0 input_8[0][0]
__________________________________________________________________________________________________
conv2d_10 (Conv2D) (None, 128, 128, 64) 3136 zero_padding2d_4[0][0]
__________________________________________________________________________________________________
leaky_re_lu_6 (LeakyReLU) (None, 128, 128, 64) 0 conv2d_10[0][0]
__________________________________________________________________________________________________
zero_padding2d_5 (ZeroPadding2D (None, 130, 130, 64) 0 leaky_re_lu_6[0][0]
__________________________________________________________________________________________________
conv2d_11 (Conv2D) (None, 64, 64, 128) 131200 zero_padding2d_5[0][0]
__________________________________________________________________________________________________
batch_normalization_8 (BatchNor (None, 64, 64, 128) 512 conv2d_11[0][0]
__________________________________________________________________________________________________
leaky_re_lu_7 (LeakyReLU) (None, 64, 64, 128) 0 batch_normalization_8[0][0]
__________________________________________________________________________________________________
zero_padding2d_6 (ZeroPadding2D (None, 66, 66, 128) 0 leaky_re_lu_7[0][0]
__________________________________________________________________________________________________
conv2d_12 (Conv2D) (None, 32, 32, 256) 524544 zero_padding2d_6[0][0]
__________________________________________________________________________________________________
batch_normalization_9 (BatchNor (None, 32, 32, 256) 1024 conv2d_12[0][0]
__________________________________________________________________________________________________
leaky_re_lu_8 (LeakyReLU) (None, 32, 32, 256) 0 batch_normalization_9[0][0]
__________________________________________________________________________________________________
zero_padding2d_7 (ZeroPadding2D (None, 34, 34, 256) 0 leaky_re_lu_8[0][0]
__________________________________________________________________________________________________
conv2d_13 (Conv2D) (None, 16, 16, 512) 2097664 zero_padding2d_7[0][0]
__________________________________________________________________________________________________
batch_normalization_10 (BatchNo (None, 16, 16, 512) 2048 conv2d_13[0][0]
__________________________________________________________________________________________________
leaky_re_lu_9 (LeakyReLU) (None, 16, 16, 512) 0 batch_normalization_10[0][0]
__________________________________________________________________________________________________
zero_padding2d_8 (ZeroPadding2D (None, 18, 18, 512) 0 leaky_re_lu_9[0][0]
__________________________________________________________________________________________________
conv2d_14 (Conv2D) (None, 8, 8, 1024) 8389632 zero_padding2d_8[0][0]
__________________________________________________________________________________________________
batch_normalization_11 (BatchNo (None, 8, 8, 1024) 4096 conv2d_14[0][0]
__________________________________________________________________________________________________
leaky_re_lu_10 (LeakyReLU) (None, 8, 8, 1024) 0 batch_normalization_11[0][0]
__________________________________________________________________________________________________
zero_padding2d_9 (ZeroPadding2D (None, 10, 10, 1024) 0 leaky_re_lu_10[0][0]
__________________________________________________________________________________________________
conv2d_15 (Conv2D) (None, 4, 4, 2048) 33556480 zero_padding2d_9[0][0]
__________________________________________________________________________________________________
batch_normalization_12 (BatchNo (None, 4, 4, 2048) 8192 conv2d_15[0][0]
__________________________________________________________________________________________________
leaky_re_lu_11 (LeakyReLU) (None, 4, 4, 2048) 0 batch_normalization_12[0][0]
__________________________________________________________________________________________________
conv2d_16 (Conv2D) (None, 4, 4, 1024) 2098176 leaky_re_lu_11[0][0]
__________________________________________________________________________________________________
batch_normalization_13 (BatchNo (None, 4, 4, 1024) 4096 conv2d_16[0][0]
__________________________________________________________________________________________________
input_9 (InputLayer) [(None, 1024)] 0
__________________________________________________________________________________________________
leaky_re_lu_12 (LeakyReLU) (None, 4, 4, 1024) 0 batch_normalization_13[0][0]
__________________________________________________________________________________________________
dense_4 (Dense) (None, 128) 131200 input_9[0][0]
__________________________________________________________________________________________________
conv2d_17 (Conv2D) (None, 4, 4, 512) 524800 leaky_re_lu_12[0][0]
__________________________________________________________________________________________________
re_lu_6 (ReLU) (None, 128) 0 dense_4[0][0]
__________________________________________________________________________________________________
batch_normalization_14 (BatchNo (None, 4, 4, 512) 2048 conv2d_17[0][0]
__________________________________________________________________________________________________
tf.reshape_1 (TFOpLambda) (None, 1, 1, 128) 0 re_lu_6[0][0]
__________________________________________________________________________________________________
leaky_re_lu_13 (LeakyReLU) (None, 4, 4, 512) 0 batch_normalization_14[0][0]
__________________________________________________________________________________________________
tf.tile_1 (TFOpLambda) (None, 4, 4, 128) 0 tf.reshape_1[0][0]
__________________________________________________________________________________________________
concatenate_2 (Concatenate) (None, 4, 4, 640) 0 leaky_re_lu_13[0][0]
tf.tile_1[0][0]
__________________________________________________________________________________________________
conv2d_18 (Conv2D) (None, 4, 4, 512) 328192 concatenate_2[0][0]
__________________________________________________________________________________________________
batch_normalization_15 (BatchNo (None, 4, 4, 512) 2048 conv2d_18[0][0]
__________________________________________________________________________________________________
leaky_re_lu_14 (LeakyReLU) (None, 4, 4, 512) 0 batch_normalization_15[0][0]
__________________________________________________________________________________________________
flatten_1 (Flatten) (None, 8192) 0 leaky_re_lu_14[0][0]
__________________________________________________________________________________________________
dense_5 (Dense) (None, 1) 8193 flatten_1[0][0]
__________________________________________________________________________________________________
activation_2 (Activation) (None, 1) 0 dense_5[0][0]
==================================================================================================
Total params: 47,817,281
Trainable params: 0
Non-trainable params: 47,817,281
__________________________________________________________________________________________________
Model: "gen_stage2"
__________________________________________________________________________________________________
Layer (type) Output Shape Param # Connected to
==================================================================================================
input_11 (InputLayer) [(None, 64, 64, 3)] 0
__________________________________________________________________________________________________
zero_padding2d_10 (ZeroPadding2 (None, 66, 66, 3) 0 input_11[0][0]
__________________________________________________________________________________________________
conv2d_19 (Conv2D) (None, 64, 64, 128) 3584 zero_padding2d_10[0][0]
__________________________________________________________________________________________________
re_lu_7 (ReLU) (None, 64, 64, 128) 0 conv2d_19[0][0]
__________________________________________________________________________________________________
zero_padding2d_11 (ZeroPadding2 (None, 66, 66, 128) 0 re_lu_7[0][0]
__________________________________________________________________________________________________
conv2d_20 (Conv2D) (None, 32, 32, 256) 524544 zero_padding2d_11[0][0]
__________________________________________________________________________________________________
batch_normalization_16 (BatchNo (None, 32, 32, 256) 1024 conv2d_20[0][0]
__________________________________________________________________________________________________
re_lu_8 (ReLU) (None, 32, 32, 256) 0 batch_normalization_16[0][0]
__________________________________________________________________________________________________
input_10 (InputLayer) [(None, 1024)] 0
__________________________________________________________________________________________________
zero_padding2d_12 (ZeroPadding2 (None, 34, 34, 256) 0 re_lu_8[0][0]
__________________________________________________________________________________________________
dense_6 (Dense) (None, 256) 262400 input_10[0][0]
__________________________________________________________________________________________________
conv2d_21 (Conv2D) (None, 16, 16, 512) 2097664 zero_padding2d_12[0][0]
__________________________________________________________________________________________________
leaky_re_lu_15 (LeakyReLU) (None, 256) 0 dense_6[0][0]
__________________________________________________________________________________________________
batch_normalization_17 (BatchNo (None, 16, 16, 512) 2048 conv2d_21[0][0]
__________________________________________________________________________________________________
lambda_1 (Lambda) (None, 128) 0 leaky_re_lu_15[0][0]
__________________________________________________________________________________________________
re_lu_9 (ReLU) (None, 16, 16, 512) 0 batch_normalization_17[0][0]
__________________________________________________________________________________________________
lambda_2 (Lambda) (None, 16, 16, 640) 0 lambda_1[0][0]
re_lu_9[0][0]
__________________________________________________________________________________________________
zero_padding2d_13 (ZeroPadding2 (None, 18, 18, 640) 0 lambda_2[0][0]
__________________________________________________________________________________________________
conv2d_22 (Conv2D) (None, 16, 16, 512) 2949632 zero_padding2d_13[0][0]
__________________________________________________________________________________________________
batch_normalization_18 (BatchNo (None, 16, 16, 512) 2048 conv2d_22[0][0]
__________________________________________________________________________________________________
re_lu_10 (ReLU) (None, 16, 16, 512) 0 batch_normalization_18[0][0]
__________________________________________________________________________________________________
conv2d_23 (Conv2D) (None, 16, 16, 512) 2359808 re_lu_10[0][0]
__________________________________________________________________________________________________
batch_normalization_19 (BatchNo (None, 16, 16, 512) 2048 conv2d_23[0][0]
__________________________________________________________________________________________________
re_lu_11 (ReLU) (None, 16, 16, 512) 0 batch_normalization_19[0][0]
__________________________________________________________________________________________________
conv2d_24 (Conv2D) (None, 16, 16, 512) 2359808 re_lu_11[0][0]
__________________________________________________________________________________________________
batch_normalization_20 (BatchNo (None, 16, 16, 512) 2048 conv2d_24[0][0]
__________________________________________________________________________________________________
add (Add) (None, 16, 16, 512) 0 batch_normalization_20[0][0]
re_lu_10[0][0]
__________________________________________________________________________________________________
re_lu_12 (ReLU) (None, 16, 16, 512) 0 add[0][0]
__________________________________________________________________________________________________
conv2d_25 (Conv2D) (None, 16, 16, 512) 2359808 re_lu_12[0][0]
__________________________________________________________________________________________________
batch_normalization_21 (BatchNo (None, 16, 16, 512) 2048 conv2d_25[0][0]
__________________________________________________________________________________________________
re_lu_13 (ReLU) (None, 16, 16, 512) 0 batch_normalization_21[0][0]
__________________________________________________________________________________________________
conv2d_26 (Conv2D) (None, 16, 16, 512) 2359808 re_lu_13[0][0]
__________________________________________________________________________________________________
batch_normalization_22 (BatchNo (None, 16, 16, 512) 2048 conv2d_26[0][0]
__________________________________________________________________________________________________
add_1 (Add) (None, 16, 16, 512) 0 batch_normalization_22[0][0]
re_lu_12[0][0]
__________________________________________________________________________________________________
re_lu_14 (ReLU) (None, 16, 16, 512) 0 add_1[0][0]
__________________________________________________________________________________________________
conv2d_27 (Conv2D) (None, 16, 16, 512) 2359808 re_lu_14[0][0]
__________________________________________________________________________________________________
batch_normalization_23 (BatchNo (None, 16, 16, 512) 2048 conv2d_27[0][0]
__________________________________________________________________________________________________
re_lu_15 (ReLU) (None, 16, 16, 512) 0 batch_normalization_23[0][0]
__________________________________________________________________________________________________
conv2d_28 (Conv2D) (None, 16, 16, 512) 2359808 re_lu_15[0][0]
__________________________________________________________________________________________________
batch_normalization_24 (BatchNo (None, 16, 16, 512) 2048 conv2d_28[0][0]
__________________________________________________________________________________________________
add_2 (Add) (None, 16, 16, 512) 0 batch_normalization_24[0][0]
re_lu_14[0][0]
__________________________________________________________________________________________________
re_lu_16 (ReLU) (None, 16, 16, 512) 0 add_2[0][0]
__________________________________________________________________________________________________
conv2d_29 (Conv2D) (None, 16, 16, 512) 2359808 re_lu_16[0][0]
__________________________________________________________________________________________________
batch_normalization_25 (BatchNo (None, 16, 16, 512) 2048 conv2d_29[0][0]
__________________________________________________________________________________________________
re_lu_17 (ReLU) (None, 16, 16, 512) 0 batch_normalization_25[0][0]
__________________________________________________________________________________________________
conv2d_30 (Conv2D) (None, 16, 16, 512) 2359808 re_lu_17[0][0]
__________________________________________________________________________________________________
batch_normalization_26 (BatchNo (None, 16, 16, 512) 2048 conv2d_30[0][0]
__________________________________________________________________________________________________
add_3 (Add) (None, 16, 16, 512) 0 batch_normalization_26[0][0]
re_lu_16[0][0]
__________________________________________________________________________________________________
re_lu_18 (ReLU) (None, 16, 16, 512) 0 add_3[0][0]
__________________________________________________________________________________________________
up_sampling2d_4 (UpSampling2D) (None, 32, 32, 512) 0 re_lu_18[0][0]
__________________________________________________________________________________________________
conv2d_31 (Conv2D) (None, 32, 32, 512) 2359808 up_sampling2d_4[0][0]
__________________________________________________________________________________________________
batch_normalization_27 (BatchNo (None, 32, 32, 512) 2048 conv2d_31[0][0]
__________________________________________________________________________________________________
re_lu_19 (ReLU) (None, 32, 32, 512) 0 batch_normalization_27[0][0]
__________________________________________________________________________________________________
up_sampling2d_5 (UpSampling2D) (None, 64, 64, 512) 0 re_lu_19[0][0]
__________________________________________________________________________________________________
conv2d_32 (Conv2D) (None, 64, 64, 256) 1179904 up_sampling2d_5[0][0]
__________________________________________________________________________________________________
batch_normalization_28 (BatchNo (None, 64, 64, 256) 1024 conv2d_32[0][0]
__________________________________________________________________________________________________
re_lu_20 (ReLU) (None, 64, 64, 256) 0 batch_normalization_28[0][0]
__________________________________________________________________________________________________
up_sampling2d_6 (UpSampling2D) (None, 128, 128, 256 0 re_lu_20[0][0]
__________________________________________________________________________________________________
conv2d_33 (Conv2D) (None, 128, 128, 128 295040 up_sampling2d_6[0][0]
__________________________________________________________________________________________________
batch_normalization_29 (BatchNo (None, 128, 128, 128 512 conv2d_33[0][0]
__________________________________________________________________________________________________
re_lu_21 (ReLU) (None, 128, 128, 128 0 batch_normalization_29[0][0]
__________________________________________________________________________________________________
up_sampling2d_7 (UpSampling2D) (None, 256, 256, 128 0 re_lu_21[0][0]
__________________________________________________________________________________________________
conv2d_34 (Conv2D) (None, 256, 256, 64) 73792 up_sampling2d_7[0][0]
__________________________________________________________________________________________________
batch_normalization_30 (BatchNo (None, 256, 256, 64) 256 conv2d_34[0][0]
__________________________________________________________________________________________________
re_lu_22 (ReLU) (None, 256, 256, 64) 0 batch_normalization_30[0][0]
__________________________________________________________________________________________________
conv2d_35 (Conv2D) (None, 256, 256, 3) 1731 re_lu_22[0][0]
__________________________________________________________________________________________________
activation_3 (Activation) (None, 256, 256, 3) 0 conv2d_35[0][0]
==================================================================================================
Total params: 28,651,907
Trainable params: 28,639,235
Non-trainable params: 12,672
__________________________________________________________________________________________________
 |
|---|
| Smoothened Stage 1 Generator Loss |
 |
|---|
| Smoothened Stage 1 Discrminator Loss |
 |
|---|
| Stage 2 Discrminator Loss |
 |
|---|
| Stage 2 Generator Loss |