import tensorflow as tf
import numpy as np
import argparse
import logdir_helpers
import val_images
import constants
import autoencoder
import probclass
import glob
import bits
import imageio
import ms_ssim_np
from codec_distance import CodecDistance, CodecDistanceReadException
from fjcommon import tf_helpers, config_parser
from images_iterator import ImagesIterator
from saver import Saver
import os
from os import path
import skimage.measure
from collections import defaultdict
from collections import namedtuple
from val_files import ValidationDirs, MeasuresWriter
import bpp_helpers
_VALIDATION_INFO_STR = """
- VALIDATION -------------------------------------------------------------------"""
_CKPT_ITR_INFO_STR = """- Validating ckpt {} ----------"""
OutputFlags = namedtuple('OutputFlags', ['save_ours', 'ckpt_step', 'real_bpp'])
def validate(val_dirs: ValidationDirs, images_iterator: ImagesIterator, flags: OutputFlags):
"""
Saves in val_dirs.log_dir/val/dataset_name/measures.csv:
- `img_name,bpp,psnr,ms-ssim forall img_name`
"""
print(_VALIDATION_INFO_STR)
validated_checkpoints = val_dirs.get_validated_checkpoints() # :: [10000, 18000, ..., 256000], ie, [int]
all_ckpts = Saver.all_ckpts_with_iterations(val_dirs.ckpt_dir)
if len(all_ckpts) == 0:
print('No checkpoints found in {}'.format(val_dirs.ckpt_dir))
return
# if ckpt_step is -1, then all_ckpt[:-1:flags.ckpt_step] === [] because of how strides work
ckpt_to_check = all_ckpts[:-1:flags.ckpt_step] + [all_ckpts[-1]] # every ckpt_step-th checkpoint plus the last one
if flags.ckpt_step == -1:
assert len(ckpt_to_check) == 1
print('Validating {}/{} checkpoints (--ckpt_step {})...'.format(
len(ckpt_to_check), len(all_ckpts), flags.ckpt_step))
missing_checkpoints = [(ckpt_itr, ckpt_path)
for ckpt_itr, ckpt_path in ckpt_to_check
if ckpt_itr not in validated_checkpoints]
if len(missing_checkpoints) == 0:
print('All checkpoints validated, stopping...')
return
# ---
# create networks
autoencoder_config_path, probclass_config_path = logdir_helpers.config_paths_from_log_dir(
val_dirs.log_dir, base_dirs=[constants.CONFIG_BASE_AE, constants.CONFIG_BASE_PC])
ae_config, ae_config_rel_path = config_parser.parse(autoencoder_config_path)
pc_config, pc_config_rel_path = config_parser.parse(probclass_config_path)
ae_cls = autoencoder.get_network_cls(ae_config)
pc_cls = probclass.get_network_cls(pc_config)
# Instantiate autoencoder and probability classifier
ae = ae_cls(ae_config)
pc = pc_cls(pc_config, num_centers=ae_config.num_centers)
x_val_ph = tf.placeholder(tf.uint8, (3, None, None), name='x_val_ph')
x_val_uint8 = tf.expand_dims(x_val_ph, 0, name='batch')
x_val = tf.to_float(x_val_uint8, name='x_val')
enc_out_val = ae.encode(x_val, is_training=False)
x_out_val = ae.decode(enc_out_val.qhard, is_training=False)
bc_val = pc.bitcost(enc_out_val.qbar, enc_out_val.symbols, is_training=False, pad_value=pc.auto_pad_value(ae))
bpp_val = bits.bitcost_to_bpp(bc_val, x_val)
x_out_val_uint8 = tf.cast(x_out_val, tf.uint8, name='x_out_val_uint8')
# Using numpy implementation due to dynamic shapes
msssim_val = ms_ssim_np.tf_msssim_np(x_val_uint8, x_out_val_uint8, data_format='NCHW')
psnr_val = psnr_np(x_val_uint8, x_out_val_uint8)
restorer = Saver(val_dirs.ckpt_dir, var_list=Saver.get_var_list_of_ckpt_dir(val_dirs.ckpt_dir))
# create fetch_dict
fetch_dict = {
'bpp': bpp_val,
'ms-ssim': msssim_val,
'psnr': psnr_val,
}
if flags.real_bpp:
fetch_dict['sym'] = enc_out_val.symbols # NCHW
if flags.save_ours:
fetch_dict['img_out'] = x_out_val_uint8
# ---
fw = tf.summary.FileWriter(val_dirs.out_dir, graph=tf.get_default_graph())
def full_summary_tag(summary_name):
return '/'.join(['val', images_iterator.dataset_name, summary_name])
# Distance
try:
codec_distance_ms_ssim = CodecDistance(images_iterator.dataset_name, codec='bpg', metric='ms-ssim')
codec_distance_psnr = CodecDistance(images_iterator.dataset_name, codec='bpg', metric='psnr')
except CodecDistanceReadException as e: # no codec distance values stored for the current setup
print('*** Distance to BPG not available for {}:\n{}'.format(images_iterator.dataset_name, str(e)))
codec_distance_ms_ssim = None
codec_distance_psnr = None
# Note that for each checkpoint, the structure of the network will be the same. Thus the pad depending image
# loading can be cached.
# create session
with tf_helpers.create_session() as sess:
if flags.real_bpp:
pred = probclass.PredictionNetwork(pc, pc_config, ae.get_centers_variable(), sess)
checker = probclass.ProbclassNetworkTesting(pc, ae, sess)
bpp_fetcher = bpp_helpers.BppFetcher(pred, checker)
fetcher = sess.make_callable(fetch_dict, feed_list=[x_val_ph])
last_ckpt_itr = missing_checkpoints[-1][0]
for ckpt_itr, ckpt_path in missing_checkpoints:
if not ckpt_still_exists(ckpt_path):
# May happen if job is still training
print('Checkpoint disappeared: {}'.format(ckpt_path))
continue
print(_CKPT_ITR_INFO_STR.format(ckpt_itr))
restorer.restore_ckpt(sess, ckpt_path)
values_aggregator = ValuesAggregator('bpp', 'ms-ssim', 'psnr')
# truncates the previous measures.csv file! This way, only the last valid checkpoint is saved.
measures_writer = MeasuresWriter(val_dirs.out_dir)
# ----------------------------------------
# iterate over images
# images are padded to work with current auto encoder
for img_i, (img_name, img_content) in enumerate(images_iterator.iter_imgs(pad=ae.get_subsampling_factor())):
otp = fetcher(img_content)
measures_writer.append(img_name, otp)
if flags.real_bpp:
# Calculate
bpp_real, bpp_theory = bpp_fetcher.get_bpp(
otp['sym'], bpp_helpers.num_pixels_in_image(img_content))
# Logging
bpp_loss = otp['bpp']
diff_percent_tr = (bpp_theory/bpp_real) * 100
diff_percent_lt = (bpp_loss/bpp_theory) * 100
print('BPP: Real {:.5f}\n'
' Theoretical: {:.5f} [{:5.1f}% of real]\n'
' Loss: {:.5f} [{:5.1f}% of real]'.format(
bpp_real, bpp_theory, diff_percent_tr, bpp_loss, diff_percent_lt))
assert abs(bpp_theory - bpp_loss) < 1e-3, 'Expected bpp_theory to match loss! Got {} and {}'.format(
bpp_theory, bpp_loss)
if flags.save_ours and ckpt_itr == last_ckpt_itr:
save_img(img_name, otp['img_out'], val_dirs)
values_aggregator.update(otp)
print('{: 10d} {img_name} | Mean: {avgs}'.format(
img_i, img_name=img_name, avgs=values_aggregator.averages_str()),
end=('\r' if not flags.real_bpp else '\n'), flush=True)
measures_writer.close()
print() # add newline
avgs = values_aggregator.averages()
avg_bpp, avg_ms_ssim, avg_psnr = avgs['bpp'], avgs['ms-ssim'], avgs['psnr']
tf_helpers.log_values(fw,
[(full_summary_tag('avg_bpp'), avg_bpp),
(full_summary_tag('avg_ms_ssim'), avg_ms_ssim),
(full_summary_tag('avg_psnr'), avg_psnr)],
iteration=ckpt_itr)
if codec_distance_ms_ssim and codec_distance_psnr:
try:
d_ms_ssim = codec_distance_ms_ssim.distance(avg_bpp, avg_ms_ssim)
d_pnsr = codec_distance_psnr.distance(avg_bpp, avg_psnr)
print('Distance to BPG: {:.3f} ms-ssim // {:.3f} psnr'.format(d_ms_ssim, d_pnsr))
tf_helpers.log_values(fw,
[(full_summary_tag('distance_BPG_MS-SSIM'), d_ms_ssim),
(full_summary_tag('distance_BPG_PSNR'), d_pnsr)],
iteration=ckpt_itr)
except ValueError as e: # out of range errors from distance calls
print(e)
val_dirs.add_validated_checkpoint(ckpt_itr)
print('Validation completed {}'.format(val_dirs))
def save_img(img_name, img_out, val_dirs):
assert img_name.endswith('.png')
assert img_out.ndim == 4 and img_out.shape[1] == 3, 'Expected NCHW, got {}'.format(img_out)
img_dir = path.join(val_dirs.out_dir, 'imgs')
os.makedirs(img_dir, exist_ok=True)
img_out = np.transpose(img_out[0, :, :, :], (1, 2, 0)) # Make HWC
img_out_p = path.join(img_dir, img_name)
print('Saving {}...'.format(img_out_p))
imageio.imsave(img_out_p, img_out)
def psnr_np(img1, img2):
assert tf.uint8.is_compatible_with(img1.dtype), 'Expected uint8 intput'
assert tf.uint8.is_compatible_with(img2.dtype), 'Expected uint8 intput'
def _psnr(_img1, _img2):
return np.float32(skimage.measure.compare_psnr(_img1, _img2))
with tf.name_scope('psnr_np'):
v = tf.py_func(_psnr, [img1, img2], tf.float32, stateful=False, name='PSNR')
v.set_shape(())
return v
class ValuesAggregator(object):
def __init__(self, *tags_to_agregate):
self._tags_to_values = defaultdict(list) # log tag -> [log value]
self.tags_to_agregate = tags_to_agregate
def update(self, fetch_dict_out):
for tag, value in fetch_dict_out.items():
if tag in self.tags_to_agregate:
assert not np.isnan(value), 'nan encountered in {}'.format(fetch_dict_out)
self._tags_to_values[tag].append(value)
def averages(self):
return {tag: np.mean(values) for tag, values in self._tags_to_values.items()}
def averages_str(self, joiner=', '):
mean_values = self.averages()
avergaes_sorted = tuple((tag, mean_values[tag]) for tag in self.tags_to_agregate) # sort by tags_to_agregate
return joiner.join('{}: {:.3f}'.format(tag, value) for tag, value in avergaes_sorted)
def ckpt_still_exists(ckpt_path):
ckpt_files = glob.glob(ckpt_path + '*')
return len(ckpt_files) > 0
def main():
p = argparse.ArgumentParser()
p.add_argument('log_dir_root', help='Path to dir containing log_dirs.')
p.add_argument('job_ids', help='Comma separated list of job_ids.')
p.add_argument('images')
p.add_argument('--save_ours', '-o', action='store_const', const=True,
help='If given, store output images in VAL_OUT/imgs.')
p.add_argument('--how_many', type=int, help='Number of images to output')
p.add_argument('--image_cache_max', '-cache', type=int, default=500, help='Cache max in [MB]. Set to 0 to disable.')
p.add_argument('--restore_itr', '-i', type=int)
p.add_argument('--ckpt_step', '-s', type=int, default=2,
help='Every CKPT_STEP-th checkpoint will be validated. Set to 1 to validate all of them. '
'Last checkpoint will always be validated. Set to -1 to only validate last.')
p.add_argument('--reset', action='store_const', const=True, help='Remove previous output')
p.add_argument('--real_bpp', action='store_const', const=True,
help='If given, calculate real bpp using arithmetic encoding. Note: in our experiments, '
'this matches the theoretical bpp up to 1% precision. Note: this is very slow.')
flags, unknown_flags = p.parse_known_args()
if unknown_flags:
print('Unknown flags: {}'.format(unknown_flags))
image_paths, dataset_name = val_images.get_image_paths(flags.images)
images_iterator = ImagesIterator(image_paths[:flags.how_many], dataset_name, flags.image_cache_max)
val_flags = OutputFlags(flags.save_ours, flags.ckpt_step, flags.real_bpp)
for ckpt_dir in logdir_helpers.iter_ckpt_dirs(flags.log_dir_root, flags.job_ids):
try:
validate(ValidationDirs(ckpt_dir, flags.log_dir_root, dataset_name, flags.reset),
images_iterator,
val_flags)
except tf.errors.NotFoundError as e:
# happens if ckpt was deleted while validation
print('*** Caught {}'.format(e))
continue
tf.reset_default_graph()
print('*** All given job_ids validated.')
if __name__ == '__main__':
main()
