File size: 9,570 Bytes
0f52c9d |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 |
import os
import ntpath
import time
from . import util
from . import html
import numpy as np
from PIL import Image as PILImage
import torch
from collections import OrderedDict
try:
from StringIO import StringIO
except ImportError:
from io import BytesIO
class Visualizer():
def __init__(self, opt):
self.opt = opt
self.tf_log = opt.isTrain and opt.tf_log
self.use_html = opt.isTrain and not opt.no_html
self.win_size = opt.display_winsize
self.name = opt.name
if self.tf_log:
import tensorflow as tf
self.tf = tf
self.log_dir = os.path.join(opt.checkpoints_dir, opt.name, 'logs')
self.writer = tf.summary.FileWriter(self.log_dir)
if self.use_html:
self.web_dir = os.path.join(opt.checkpoints_dir, opt.name, 'web')
self.img_dir = os.path.join(self.web_dir, 'images')
print('create web directory %s...' % self.web_dir)
util.mkdirs([self.web_dir, self.img_dir])
if opt.isTrain:
self.log_name = os.path.join(opt.checkpoints_dir, opt.name, 'loss_log.txt')
with open(self.log_name, "a") as log_file:
now = time.strftime("%c")
log_file.write('================ Training Loss (%s) ================\n' % now)
def convert_map_to_numpy(self, data_map):
if data_map is None or not isinstance(data_map, torch.Tensor):
return None
if data_map.dim() == 4:
data_map = data_map[0]
if data_map.size(0) > 1:
data_map = data_map[0, :, :].unsqueeze(0)
map_numpy = data_map.cpu().float().numpy()
min_val, max_val = np.min(map_numpy), np.max(map_numpy)
if max_val - min_val > 1e-6:
map_numpy = (map_numpy - min_val) / (max_val - min_val)
else:
map_numpy = np.zeros_like(map_numpy)
map_numpy = (map_numpy * 255.0).astype(np.uint8)
if map_numpy.shape[0] == 1:
map_numpy = np.transpose(map_numpy, (1, 2, 0))
map_numpy = np.repeat(map_numpy, 3, axis=2)
else:
map_numpy = np.stack((map_numpy,) * 3, axis=-1)
return map_numpy
def display_current_results(self, visuals, epoch, step):
visuals_np = OrderedDict()
for label, image in visuals.items():
if image is None:
continue
if 'light_map' in label:
image_numpy = self.convert_map_to_numpy(image)
elif 'input_label' in label:
image_numpy = util.tensor2label(image, self.opt.label_nc, tile=False)
else:
image_numpy = util.tensor2im(image, tile=False)
if image_numpy.ndim == 4:
image_numpy = image_numpy[0]
visuals_np[label] = image_numpy
if self.tf_log:
img_summaries = []
for label, image_numpy in visuals_np.items():
if image_numpy is None: continue
try:
s = BytesIO()
pil_img = PILImage.fromarray(image_numpy)
pil_img.save(s, format="jpeg")
img_sum = self.tf.Summary.Image(encoded_image_string=s.getvalue(), height=image_numpy.shape[0],
width=image_numpy.shape[1])
img_summaries.append(self.tf.Summary.Value(tag=f'epoch_{epoch}/{label}', image=img_sum))
except Exception as e:
print(f"Could not write image {label} to TF logs: {e}")
if img_summaries:
summary = self.tf.Summary(value=img_summaries)
self.writer.add_summary(summary, step)
if self.use_html:
webpage = html.HTML(self.web_dir, 'Experiment name = %s' % self.name, refresh=0)
webpage.add_header('Epoch [%d] Iteration [%d]' % (epoch, step))
visuals_for_html = []
labels_for_html = []
standard_height = self.opt.crop_size
for label, image_numpy in visuals_np.items():
if image_numpy is None: continue
pil_img = PILImage.fromarray(image_numpy)
if pil_img.height != standard_height:
aspect_ratio = pil_img.width / pil_img.height
new_width = int(standard_height * aspect_ratio)
pil_img = pil_img.resize((new_width, standard_height), PILImage.LANCZOS)
visuals_for_html.append(np.array(pil_img))
labels_for_html.append(label)
if not visuals_for_html:
return
try:
concatenated_image = np.concatenate(visuals_for_html, axis=1)
image_name = 'epoch%.3d_iter%.7d_combined.png' % (epoch, step)
save_path = os.path.join(self.img_dir, image_name)
util.save_image(concatenated_image, save_path)
webpage.add_images([image_name], [' | '.join(labels_for_html)], [image_name],
width=self.win_size * len(visuals_for_html))
webpage.save()
except ValueError as e:
print(f"Error during HTML image concatenation for step {step}: {e}")
print("Skipping HTML log for this step. Image shapes might be incompatible even after resizing.")
def plot_current_errors(self, errors, step):
if self.tf_log:
for tag, value in errors.items():
if isinstance(value, torch.Tensor):
value_to_log = value.mean().float().item()
elif isinstance(value, (float, int)):
value_to_log = float(value)
else:
continue
summary = self.tf.Summary(value=[self.tf.Summary.Value(tag=tag, simple_value=value_to_log)])
self.writer.add_summary(summary, step)
def print_current_errors(self, epoch, i, errors, t):
message = '(epoch: %d, iters: %d, time: %.3f) ' % (epoch, i, t)
for k, v_orig in errors.items():
v_to_print = v_orig
if isinstance(v_orig, torch.Tensor):
if v_orig.numel() > 0:
v_to_print = v_orig.mean().item()
else:
v_to_print = 0.0
elif not isinstance(v_orig, (float, int)):
continue
message += '%s: %.3f ' % (k, float(v_to_print))
print(message)
with open(self.log_name, "a") as log_file:
log_file.write('%s\n' % message)
def save_images(self, webpage, visuals, image_path_list, alpha=1.0):
visuals_np = OrderedDict()
for label, image in visuals.items():
if 'light_map' in label:
visuals_np[label] = self.convert_map_to_numpy(image)
else:
visuals_np[label] = util.tensor2im(image)
base_image_dir = webpage.get_image_dir()
image_path_str = image_path_list[0] if isinstance(image_path_list, (list, tuple)) else image_path_list
short_path = ntpath.basename(image_path_str)
name_prefix = os.path.splitext(short_path)[0]
current_alpha_float = alpha
if isinstance(current_alpha_float, torch.Tensor):
current_alpha_float = current_alpha_float.mean().item()
elif not isinstance(current_alpha_float, (float, int)):
try:
current_alpha_float = float(current_alpha_float)
except ValueError:
current_alpha_float = 1.0
alpha_folder_name = "alpha_{:.3f}".format(current_alpha_float).replace('.', '_')
specific_alpha_image_dir = os.path.join(base_image_dir, alpha_folder_name)
util.mkdirs(specific_alpha_image_dir)
image_name_final = '%s.png' % (name_prefix)
save_path = os.path.join(specific_alpha_image_dir, image_name_final)
images_to_concatenate = []
for label, image_numpy in visuals_np.items():
img_to_add = image_numpy
if image_numpy.ndim == 4 and image_numpy.shape[0] == 1:
img_to_add = image_numpy.squeeze(0)
elif image_numpy.ndim != 2 and image_numpy.ndim != 3:
continue
if img_to_add.ndim == 2:
img_to_add = np.stack((img_to_add,) * 3, axis=-1)
if img_to_add.ndim == 3 and img_to_add.shape[2] == 1:
img_to_add = np.concatenate([img_to_add] * 3, axis=2)
if img_to_add.shape[2] == 3:
images_to_concatenate.append(img_to_add)
if not images_to_concatenate:
return
try:
image_concatenated_horizontally = np.concatenate(images_to_concatenate, axis=1)
util.save_image(image_concatenated_horizontally, save_path, create_dir=True)
except ValueError as e:
print(f"Error concatenating images for {save_path}: {e}")
print("Concatenated images list content (shapes):")
for idx, vis_np_item in enumerate(images_to_concatenate):
print(f" Visual {idx}: shape {vis_np_item.shape if hasattr(vis_np_item, 'shape') else 'N/A'}")
relative_image_path_for_html = os.path.join(alpha_folder_name, image_name_final)
webpage.add_images([relative_image_path_for_html], [f"{name_prefix}_alpha_{current_alpha_float:.3f}"],
[relative_image_path_for_html], width=self.win_size * len(images_to_concatenate)) |