Spaces:
Running
Running
File size: 9,624 Bytes
fcc02a2 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 |
import os
import requests
import tqdm
from typing import List, Optional, TYPE_CHECKING
def img_root_path(img_id: str):
return os.path.dirname(os.path.dirname(img_id))
if TYPE_CHECKING:
from .dataset_tools_config_modules import DatasetSyncCollectionConfig
img_exts = ['.jpg', '.jpeg', '.webp', '.png']
class Photo:
def __init__(
self,
id,
host,
width,
height,
url,
filename
):
self.id = str(id)
self.host = host
self.width = width
self.height = height
self.url = url
self.filename = filename
def get_desired_size(img_width: int, img_height: int, min_width: int, min_height: int):
if img_width > img_height:
scale = min_height / img_height
else:
scale = min_width / img_width
new_width = int(img_width * scale)
new_height = int(img_height * scale)
return new_width, new_height
def get_pexels_images(config: 'DatasetSyncCollectionConfig') -> List[Photo]:
all_images = []
next_page = f"https://api.pexels.com/v1/collections/{config.collection_id}?page=1&per_page=80&type=photos"
while True:
response = requests.get(next_page, headers={
"Authorization": f"{config.api_key}"
})
response.raise_for_status()
data = response.json()
all_images.extend(data['media'])
if 'next_page' in data and data['next_page']:
next_page = data['next_page']
else:
break
photos = []
for image in all_images:
new_width, new_height = get_desired_size(image['width'], image['height'], config.min_width, config.min_height)
url = f"{image['src']['original']}?auto=compress&cs=tinysrgb&h={new_height}&w={new_width}"
filename = os.path.basename(image['src']['original'])
photos.append(Photo(
id=image['id'],
host="pexels",
width=image['width'],
height=image['height'],
url=url,
filename=filename
))
return photos
def get_unsplash_images(config: 'DatasetSyncCollectionConfig') -> List[Photo]:
headers = {
# "Authorization": f"Client-ID {UNSPLASH_ACCESS_KEY}"
"Authorization": f"Client-ID {config.api_key}"
}
# headers['Authorization'] = f"Bearer {token}"
url = f"https://api.unsplash.com/collections/{config.collection_id}/photos?page=1&per_page=30"
response = requests.get(url, headers=headers)
response.raise_for_status()
res_headers = response.headers
# parse the link header to get the next page
# 'Link': '<https://api.unsplash.com/collections/mIPWwLdfct8/photos?page=82>; rel="last", <https://api.unsplash.com/collections/mIPWwLdfct8/photos?page=2>; rel="next"'
has_next_page = False
if 'Link' in res_headers:
has_next_page = True
link_header = res_headers['Link']
link_header = link_header.split(',')
link_header = [link.strip() for link in link_header]
link_header = [link.split(';') for link in link_header]
link_header = [[link[0].strip('<>'), link[1].strip().strip('"')] for link in link_header]
link_header = {link[1]: link[0] for link in link_header}
# get page number from last url
last_page = link_header['rel="last']
last_page = last_page.split('?')[1]
last_page = last_page.split('&')
last_page = [param.split('=') for param in last_page]
last_page = {param[0]: param[1] for param in last_page}
last_page = int(last_page['page'])
all_images = response.json()
if has_next_page:
# assume we start on page 1, so we don't need to get it again
for page in tqdm.tqdm(range(2, last_page + 1)):
url = f"https://api.unsplash.com/collections/{config.collection_id}/photos?page={page}&per_page=30"
response = requests.get(url, headers=headers)
response.raise_for_status()
all_images.extend(response.json())
photos = []
for image in all_images:
new_width, new_height = get_desired_size(image['width'], image['height'], config.min_width, config.min_height)
url = f"{image['urls']['raw']}&w={new_width}"
filename = f"{image['id']}.jpg"
photos.append(Photo(
id=image['id'],
host="unsplash",
width=image['width'],
height=image['height'],
url=url,
filename=filename
))
return photos
def get_img_paths(dir_path: str):
os.makedirs(dir_path, exist_ok=True)
local_files = os.listdir(dir_path)
# remove non image files
local_files = [file for file in local_files if os.path.splitext(file)[1].lower() in img_exts]
# make full path
local_files = [os.path.join(dir_path, file) for file in local_files]
return local_files
def get_local_image_ids(dir_path: str):
os.makedirs(dir_path, exist_ok=True)
local_files = get_img_paths(dir_path)
# assuming local files are named after Unsplash IDs, e.g., 'abc123.jpg'
return set([os.path.basename(file).split('.')[0] for file in local_files])
def get_local_image_file_names(dir_path: str):
os.makedirs(dir_path, exist_ok=True)
local_files = get_img_paths(dir_path)
# assuming local files are named after Unsplash IDs, e.g., 'abc123.jpg'
return set([os.path.basename(file) for file in local_files])
def download_image(photo: Photo, dir_path: str, min_width: int = 1024, min_height: int = 1024):
img_width = photo.width
img_height = photo.height
if img_width < min_width or img_height < min_height:
raise ValueError(f"Skipping {photo.id} because it is too small: {img_width}x{img_height}")
img_response = requests.get(photo.url)
img_response.raise_for_status()
os.makedirs(dir_path, exist_ok=True)
filename = os.path.join(dir_path, photo.filename)
with open(filename, 'wb') as file:
file.write(img_response.content)
def update_caption(img_path: str):
# if the caption is a txt file, convert it to a json file
filename_no_ext = os.path.splitext(os.path.basename(img_path))[0]
# see if it exists
if os.path.exists(os.path.join(os.path.dirname(img_path), f"{filename_no_ext}.json")):
# todo add poi and what not
return # we have a json file
caption = ""
# see if txt file exists
if os.path.exists(os.path.join(os.path.dirname(img_path), f"{filename_no_ext}.txt")):
# read it
with open(os.path.join(os.path.dirname(img_path), f"{filename_no_ext}.txt"), 'r') as file:
caption = file.read()
# write json file
with open(os.path.join(os.path.dirname(img_path), f"{filename_no_ext}.json"), 'w') as file:
file.write(f'{{"caption": "{caption}"}}')
# delete txt file
os.remove(os.path.join(os.path.dirname(img_path), f"{filename_no_ext}.txt"))
# def equalize_img(img_path: str):
# input_path = img_path
# output_path = os.path.join(img_root_path(img_path), COLOR_CORRECTED_DIR, os.path.basename(img_path))
# os.makedirs(os.path.dirname(output_path), exist_ok=True)
# process_img(
# img_path=input_path,
# output_path=output_path,
# equalize=True,
# max_size=2056,
# white_balance=False,
# gamma_correction=False,
# strength=0.6,
# )
# def annotate_depth(img_path: str):
# # make fake args
# args = argparse.Namespace()
# args.annotator = "midas"
# args.res = 1024
#
# img = cv2.imread(img_path)
# img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
#
# output = annotate(img, args)
#
# output = output.astype('uint8')
# output = cv2.cvtColor(output, cv2.COLOR_RGB2BGR)
#
# os.makedirs(os.path.dirname(img_path), exist_ok=True)
# output_path = os.path.join(img_root_path(img_path), DEPTH_DIR, os.path.basename(img_path))
#
# cv2.imwrite(output_path, output)
# def invert_depth(img_path: str):
# img = cv2.imread(img_path)
# img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
# # invert the colors
# img = cv2.bitwise_not(img)
#
# os.makedirs(os.path.dirname(img_path), exist_ok=True)
# output_path = os.path.join(img_root_path(img_path), INVERTED_DEPTH_DIR, os.path.basename(img_path))
# cv2.imwrite(output_path, img)
#
# # update our list of raw images
# raw_images = get_img_paths(raw_dir)
#
# # update raw captions
# for image_id in tqdm.tqdm(raw_images, desc="Updating raw captions"):
# update_caption(image_id)
#
# # equalize images
# for img_path in tqdm.tqdm(raw_images, desc="Equalizing images"):
# if img_path not in eq_images:
# equalize_img(img_path)
#
# # update our list of eq images
# eq_images = get_img_paths(eq_dir)
# # update eq captions
# for image_id in tqdm.tqdm(eq_images, desc="Updating eq captions"):
# update_caption(image_id)
#
# # annotate depth
# depth_dir = os.path.join(root_dir, DEPTH_DIR)
# depth_images = get_img_paths(depth_dir)
# for img_path in tqdm.tqdm(eq_images, desc="Annotating depth"):
# if img_path not in depth_images:
# annotate_depth(img_path)
#
# depth_images = get_img_paths(depth_dir)
#
# # invert depth
# inv_depth_dir = os.path.join(root_dir, INVERTED_DEPTH_DIR)
# inv_depth_images = get_img_paths(inv_depth_dir)
# for img_path in tqdm.tqdm(depth_images, desc="Inverting depth"):
# if img_path not in inv_depth_images:
# invert_depth(img_path)
|