Re-implement CycleGAN in Tensorlayer

Last update: Aug 15, 2022

Overview

CycleGAN_Tensorlayer

Re-implement CycleGAN in TensorLayer

Original CycleGAN
Improved CycleGAN with resize-convolution

Prerequisites:

TensorLayer
TensorFlow
Python

Run:

CUDA_VISIBLE_DEVICES=0 python main.py

(if datasets are collected by yourself, you can use dataset_clean.py or dataset_crop.py to pre-process images)

Theory:

The generator process:

The discriminator process:

Result Improvement

Data augmentation
Resize convolution[4]
Instance normalization[5]

data augmentation:

Instance normalization（comparision by original paper https://arxiv.org/abs/1607.08022）:

Resize convolution (Remove Checkerboard Artifacts):

Final Results:

Reference:

[1] Original Paper: https://arxiv.org/pdf/1703.10593.pdf
[2] Original implement in Torch: https://github.com/junyanz/CycleGAN/
[3] TensorLayer by HaoDong: https://github.com/zsdonghao/tensorlayer
[4] Resize Convolution: https://distill.pub/2016/deconv-checkerboard/
[5] Instance Normalization: https://arxiv.org/abs/1607.08022

Comments

Difference from original code
HI very nice implemented cyclegan I have a few questions...

What does "Resize Convolution" mean?

I wonder what is different from the original code of the author.
opened by taki0112 7
Color inversion, black image and nan in loss after ~20 epochs

I've tried to train the model on original summer2winter_yosemite dataset. After ~20 epochs all sample images turned completely black, and all all loss parameters turned to nan. However, the model continued to run for 30 more epochs regularly saving checkpoints until I stopped it.

I've also used another, my own dataset, and it ran correctly for 70 epochs at least, unfortunately the only result I had was color inversion of images. Any advice on changing training parameters (I used default)?

opened by victor-felicitas 0
How to change test output size?

Hi! It is a great implementation of Cyclegan, providing excellent results on Hiptensorflow and ROCm. However, I could not use it to generate test images of different from 256x256 sizes. How can I change that?

For now, I have trained the model on 256x256 images and try to test it on bigger ones. I tried adding two more flags to main.py: flags.DEFINE_integer("image_width", 420, "The size of image to use (will be center cropped) [256]") flags.DEFINE_integer("image_height", 420, "The size of image to use (will be center cropped) [256]")

Which I use later in Test section: test_A = tf.placeholder(tf.float32, [FLAGS.batch_size, FLAGS.image_height, FLAGS.image_width, FLAGS.c_dim], name='test_x') test_B = tf.placeholder(tf.float32, [FLAGS.batch_size, FLAGS.image_height, FLAGS.image_width, FLAGS.c_dim], name='test_y')

However, I always get error: Invalid argument: Conv2DSlowBackpropInput: Size of out_backprop doesn't match computed: actual = 105, computed = 64 Traceback (most recent call last): File "main.py", line 285, in tf.app.run() File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/platform/app.py", line 44, in run _sys.exit(main(_sys.argv[:1] + flags_passthrough)) File "main.py", line 281, in main test_cyclegan() File "main.py", line 262, in test_cyclegan fake_img = sess.run(net_g_logits, feed_dict={in_var: sample_image}) File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.py", line 767, in run run_metadata_ptr) File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.py", line 965, in _run feed_dict_string, options, run_metadata) File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.py", line 1015, in _do_run target_list, options, run_metadata) File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.py", line 1035, in _do_call raise type(e)(node_def, op, message) tensorflow.python.framework.errors_impl.InvalidArgumentError: Conv2DSlowBackpropInput: Size of out_backprop doesn't match computed: actual = 105, computed = 64 [[Node: gen_A2B/u64/conv2d_transpose = Conv2DBackpropInput[T=DT_FLOAT, data_format="NHWC", padding="SAME", strides=[1, 2, 2, 1], use_cudnn_on_gpu=true, _device="/job:localhost/replica:0/task:0/gpu:0"](gen_A2B/u64/conv2d_transpose/output_shape, gen_A2B/u64/W_deconv2d/read, gen_A2B/b_residual_add/8)]]

Is there any way to choose output image size? Original Cyclegan has special option to choose it - how can i implement it? resize_or_crop = 'resize_and_crop', -- resizing/cropping strategy: resize_and_crop | crop | scale_width | scale_height

Any help would be appreciated!

opened by victor-felicitas 0
About the imagepool.

I noticed in https://github.com/luoxier/CycleGAN_Tensorlayer/blob/master/main.py#L88 you obtain the logit of image sampled from imagepool but do not use it, is that for some reason or just do not intend to implement it?

opened by Zardinality 0
Error in main.py?

Hi @zsdonghao @luoxier , Is there an error in your main.py: _, errGB2A = sess.run([g_b2a_optim, g_b2a_loss], feed_dict={real_A: batch_imgB, real_B: batch_imgB}) Does it should be: _, errGB2A = sess.run([g_b2a_optim, g_b2a_loss], feed_dict={real_A: batch_imgA, real_B: batch_imgB}) Could you please check it and let me know, thanks.

opened by yongqiangzhang1 2
Where are datasets shown in readme?

There are sunflower2daisy and leopard2tiger results shown in readme, but I don't find any clue about where to download them in code. In https://github.com/luoxier/CycleGAN_Tensorlayer/blob/master/main.py#L32 an optional value for dataset_dir is sunflower2daisy, where can I get it? The author of original paper doesn't seem to provide it.

opened by Zardinality 7

Releases(0.1)

0.1(Sep 30, 2017)
TensorFlow 1.3

TensorLayer (self-contained)

Source code(tar.gz)
Source code(zip)

Owner

GitHub Repository

Code for our method RePRI for Few-Shot Segmentation. Paper at http://arxiv.org/abs/2012.06166

Region Proportion Regularized Inference (RePRI) for Few-Shot Segmentation In this repo, we provide the code for our paper : "Few-Shot Segmentation Wit

138 Dec 12, 2022

Decision Transformer: A brand new Offline RL Pattern

DecisionTransformer_StepbyStep Intro Decision Transformer: A brand new Offline RL Pattern. 这是关于NeurIPS 2021 热门论文Decision Transformer的复现。 👍 原文地址: Deci

14 Nov 22, 2022

Official pytorch code for SSAT: A Symmetric Semantic-Aware Transformer Network for Makeup Transfer and Removal

SSAT: A Symmetric Semantic-Aware Transformer Network for Makeup Transfer and Removal This is the official pytorch code for SSAT: A Symmetric Semantic-

57 Dec 13, 2022

Open-source code for Generic Grouping Network (GGN, CVPR 2022)

Open-World Instance Segmentation: Exploiting Pseudo Ground Truth From Learned Pairwise Affinity Pytorch implementation for "Open-World Instance Segmen

99 Dec 06, 2022

salabim - discrete event simulation in Python

Object oriented discrete event simulation and animation in Python. Includes process control features, resources, queues, monitors. statistical distrib

181 Dec 21, 2022

The official code of Anisotropic Stroke Control for Multiple Artists Style Transfer

ASMA-GAN Anisotropic Stroke Control for Multiple Artists Style Transfer Proceedings of the 28th ACM International Conference on Multimedia The officia

146 Nov 21, 2022

Deep Learning as a Cloud API Service.

Deep API Deep Learning as Cloud APIs. This project provides pre-trained deep learning models as a cloud API service. A web interface is available as w

4 Jan 06, 2023

Repo for the ACMMM20 submission: "Personalized breath based biometric authentication with wearable multimodality".

personalized-breath Repo for the ACMMM20 submission: "Personalized breath based biometric authentication with wearable multimodality". Guideline To ex

2 Nov 15, 2021

My solutions for Stanford University course CS224W: Machine Learning with Graphs Fall 2021 colabs (GNN, GAT, GraphSAGE, GCN)

machine-learning-with-graphs My solutions for Stanford University course CS224W: Machine Learning with Graphs Fall 2021 colabs Course materials can be

7 Dec 14, 2022

PyTorch implementation and pretrained models for XCiT models. See XCiT: Cross-Covariance Image Transformer

Official code Cross-Covariance Image Transformer (XCiT)

605 Jan 02, 2023

This computer program provides a reference implementation of Lagrangian Monte Carlo in metric induced by the Monge patch

This computer program provides a reference implementation of Lagrangian Monte Carlo in metric induced by the Monge patch. The code was prepared to the final version of the accepted manuscript in AIST

2 May 06, 2022

Circuit Training: An open-source framework for generating chip floor plans with distributed deep reinforcement learning

Circuit Training: An open-source framework for generating chip floor plans with distributed deep reinforcement learning. Circuit Training is an open-s

479 Dec 25, 2022

Datasets, tools, and benchmarks for representation learning of code.

The CodeSearchNet challenge has been concluded We would like to thank all participants for their submissions and we hope that this challenge provided

1.8k Dec 25, 2022

PyTorch implementation of hand mesh reconstruction described in CMR and MobRecon.

Hand Mesh Reconstruction Introduction This repo is the PyTorch implementation of hand mesh reconstruction described in CMR and MobRecon. Update 2021-1

236 Dec 29, 2022

[CVPR 2022] Official code for the paper: "A Stitch in Time Saves Nine: A Train-Time Regularizing Loss for Improved Neural Network Calibration"

MDCA Calibration This is the official PyTorch implementation for the paper: "A Stitch in Time Saves Nine: A Train-Time Regularizing Loss for Improved

21 Dec 22, 2022

Re-implement CycleGAN in Tensorlayer

Related tags