SRDenseNet-Tensorflow

Tensorflow implemetation of SRDensenet

Prerequisites

python 3.x
Tensorflow > 1.5
Scipy version > 0.18 ('mode' option from scipy.misc.imread function)
matplotlib
argparse

Properties (what's different from reference code)

This code requires Tensorflow. This code was fully implemented based on Python 3
This code supports only RGB color images (demo type) and Ychannel of YCbCr (eval type)
This code supports data augmentation (rotation and mirror flip)
This code supports custom dataset

Usage

usage: main.py [-h] [--exp_tag EXP_TAG] [--gpu GPU] [--epoch EPOCH]
               [--batch_size BATCH_SIZE] [--patch_size PATCH_SIZE]
               [--base_lr BASE_LR] [--lr_min LR_MIN]
               [--num_denseblock NUM_DENSEBLOCK]
               [--num_denselayer NUM_DENSELAYER] [--growth_rate GROWTH_RATE]
               [--lr_decay_rate LR_DECAY_RATE] [--lr_step_size LR_STEP_SIZE]
               [--scale SCALE] [--checkpoint_dir CHECKPOINT_DIR]
               [--cpkt_itr CPKT_ITR] [--save_period SAVE_PERIOD]
               [--train_subdir TRAIN_SUBDIR] [--test_subdir TEST_SUBDIR]
               [--infer_subdir INFER_SUBDIR] [--infer_imgpath INFER_IMGPATH]
               [--type {eval,demo}] [--c_dim C_DIM]
               [--mode {train,test,inference,test_plot}]
               [--result_dir RESULT_DIR] [--save_extension {jpg,png}]

Namespace(base_lr=0.0001, batch_size=32, c_dim=3, checkpoint_dir='checkpoint', cpkt_itr=0, epoch=80, exp_tag='SRDenseNet tensorflow. Implemented by Dohyun Kim', gpu=1, growth_rate=16, infer_imgpath='monarch.bmp', infer_subdir='Custom', lr_decay_rate=0.1, lr_min=1e-06, lr_step_size=30, mode='train', num_denseblock=8, num_denselayer=8, patch_size=33, result_dir='result', save_extension='.jpg', save_period=1, scale=3, test_subdir='Set5', train_subdir='291', type='demo')

For training, python3 main.py --mode train --type demo --check_itr 0 [set 0 for training from scratch, -1 for latest]
For testing, python 3main.py --mode test --type demo
For inference with cumstom dataset, python3 main.py --mode inference --infer_imgpath 3.bmp [result will be generated in ./result/inference]
For running tensorboard, tensorboard --logdir=./board then access localhost:6006 with your browser