DL-Art-School/codes/test.py

152 lines
5.9 KiB
Python
Raw Normal View History

2019-08-23 13:42:47 +00:00
import os.path as osp
import logging
import time
import argparse
from collections import OrderedDict
2020-08-12 14:46:15 +00:00
import os
2019-08-23 13:42:47 +00:00
import options.options as option
import utils.util as util
from data.util import bgr2ycbcr
2020-07-24 18:26:44 +00:00
import models.archs.SwitchedResidualGenerator_arch as srg
2020-08-12 14:46:15 +00:00
from switched_conv_util import save_attention_to_image, save_attention_to_image_rgb
from switched_conv import compute_attention_specificity
2019-08-23 13:42:47 +00:00
from data import create_dataset, create_dataloader
from models import create_model
2020-04-24 05:59:09 +00:00
from tqdm import tqdm
import torch
2020-07-24 18:26:44 +00:00
import models.networks as networks
# Concepts: Swap transformations around. Normalize attention. Disable individual switches, both randomly and one at
# a time, starting at the last switch. Pick random regions in an image and print out the full attention vector for
# each switch. Yield an output directory name for each alteration and None when last alteration is completed.
def alter_srg(srg: srg.ConfigurableSwitchedResidualGenerator2):
# First alteration, strip off switches one at a time.
yield "naked"
2020-08-12 14:46:15 +00:00
'''
2020-07-24 18:26:44 +00:00
for i in range(1, len(srg.switches)):
srg.switches = srg.switches[:-i]
yield "stripped-%i" % (i,)
2020-08-12 14:46:15 +00:00
'''
for sw in srg.switches:
sw.set_temperature(.001)
yield "specific"
for sw in srg.switches:
sw.set_temperature(1000)
yield "normalized"
for sw in srg.switches:
sw.set_temperature(1)
sw.switch.attention_norm = None
yield "no_anorm"
2020-07-24 18:26:44 +00:00
return None
def analyze_srg(srg: srg.ConfigurableSwitchedResidualGenerator2, path, alteration_suffix):
2020-08-12 14:46:15 +00:00
mean_hists = [compute_attention_specificity(att, 2) for att in srg.attentions]
means = [i[0] for i in mean_hists]
hists = [torch.histc(i[1].clone().detach().cpu().flatten().float(), bins=srg.transformation_counts) for i in mean_hists]
hists = [h / torch.sum(h) for h in hists]
for i in range(len(means)):
print("%s - switch_%i_specificity" % (alteration_suffix, i), means[i])
print("%s - switch_%i_histogram" % (alteration_suffix, i), hists[i])
[save_attention_to_image_rgb(path, srg.attentions[i], srg.transformation_counts, alteration_suffix, i) for i in range(len(srg.attentions))]
2020-07-24 18:26:44 +00:00
def forward_pass(model, output_dir, alteration_suffix=''):
model.feed_data(data, need_GT=need_GT)
model.test()
visuals = model.get_current_visuals(need_GT)['rlt'].cpu()
fea_loss = 0
2020-07-24 18:26:44 +00:00
for i in range(visuals.shape[0]):
img_path = data['GT_path'][i] if need_GT else data['LQ_path'][i]
img_name = osp.splitext(osp.basename(img_path))[0]
sr_img = util.tensor2img(visuals[i]) # uint8
# save images
suffix = alteration_suffix
if suffix:
save_img_path = osp.join(output_dir, img_name + suffix + '.png')
else:
save_img_path = osp.join(output_dir, img_name + '.png')
if need_GT:
fea_loss += model.compute_fea_loss(visuals[i], data['GT'][i])
2020-07-24 18:26:44 +00:00
util.save_img(sr_img, save_img_path)
return fea_loss
2020-07-24 18:26:44 +00:00
2019-08-23 13:42:47 +00:00
2020-04-24 05:59:09 +00:00
if __name__ == "__main__":
#### options
torch.backends.cudnn.benchmark = True
2020-08-12 14:46:15 +00:00
srg_analyze = False
2020-04-24 05:59:09 +00:00
parser = argparse.ArgumentParser()
2020-09-06 23:27:08 +00:00
parser.add_argument('-opt', type=str, help='Path to options YAML file.', default='../options/srgan_compute_feature.yml')
2020-04-24 05:59:09 +00:00
opt = option.parse(parser.parse_args().opt, is_train=False)
opt = option.dict_to_nonedict(opt)
2019-08-23 13:42:47 +00:00
2020-04-24 05:59:09 +00:00
util.mkdirs(
(path for key, path in opt['path'].items()
if not key == 'experiments_root' and 'pretrain_model' not in key and 'resume' not in key))
util.setup_logger('base', opt['path']['log'], 'test_' + opt['name'], level=logging.INFO,
screen=True, tofile=True)
logger = logging.getLogger('base')
logger.info(option.dict2str(opt))
2019-08-23 13:42:47 +00:00
2020-04-24 05:59:09 +00:00
#### Create test dataset and dataloader
test_loaders = []
for phase, dataset_opt in sorted(opt['datasets'].items()):
test_set = create_dataset(dataset_opt)
test_loader = create_dataloader(test_set, dataset_opt)
logger.info('Number of test images in [{:s}]: {:d}'.format(dataset_opt['name'], len(test_set)))
test_loaders.append(test_loader)
2019-08-23 13:42:47 +00:00
2020-04-24 05:59:09 +00:00
model = create_model(opt)
fea_loss = 0
2020-04-24 05:59:09 +00:00
for test_loader in test_loaders:
test_set_name = test_loader.dataset.opt['name']
logger.info('\nTesting [{:s}]...'.format(test_set_name))
test_start_time = time.time()
dataset_dir = osp.join(opt['path']['results_root'], test_set_name)
util.mkdir(dataset_dir)
2019-08-23 13:42:47 +00:00
2020-04-24 05:59:09 +00:00
test_results = OrderedDict()
test_results['psnr'] = []
test_results['ssim'] = []
test_results['psnr_y'] = []
test_results['ssim_y'] = []
2019-08-23 13:42:47 +00:00
2020-04-24 05:59:09 +00:00
tq = tqdm(test_loader)
for data in tq:
need_GT = False if test_loader.dataset.opt['dataroot_GT'] is None else True
2019-08-23 13:42:47 +00:00
2020-07-24 18:26:44 +00:00
if srg_analyze:
orig_model = model.netG
model_copy = networks.define_G(opt).to(model.device)
model_copy.load_state_dict(orig_model.state_dict())
model.netG = model_copy
for alteration_suffix in alter_srg(model_copy):
2020-08-12 14:46:15 +00:00
alt_path = osp.join(dataset_dir, alteration_suffix)
2020-07-24 18:26:44 +00:00
img_path = data['GT_path'][0] if need_GT else data['LQ_path'][0]
2020-08-12 14:46:15 +00:00
img_name = osp.splitext(osp.basename(img_path))[0] + opt['name']
2020-07-24 18:26:44 +00:00
alteration_suffix += img_name
2020-08-12 14:46:15 +00:00
os.makedirs(alt_path, exist_ok=True)
2020-07-24 18:26:44 +00:00
forward_pass(model, dataset_dir, alteration_suffix)
2020-08-12 14:46:15 +00:00
analyze_srg(model_copy, alt_path, alteration_suffix)
2020-07-24 18:26:44 +00:00
# Reset model and do next alteration.
model_copy = networks.define_G(opt).to(model.device)
model_copy.load_state_dict(orig_model.state_dict())
model.netG = model_copy
else:
fea_loss += forward_pass(model, dataset_dir, opt['name'])
# log
logger.info('# Validation # Fea: {:.4e}'.format(fea_loss / len(test_loader)))