def _group_channels(url_list, order):
"""
given a list of img urls, this will group them into the same well and
site, per plate
Arguments:
-----------
order : boolean
sort channel numbers into numerical order
"""
grouped_list = []
urls = [parse.img_filename(i) for i in url_list]
tmp_df = pd.DataFrame(list(url_list), columns=["img_url"])
tmp_df["plate_name"] = [parse.plate_name(i) for i in url_list]
tmp_df["plate_num"] = [parse.plate_num(i) for i in url_list]
# get_well and get_site use the image URL rather than the full path
tmp_df["well"] = [parse.img_well(i) for i in urls]
tmp_df["site"] = [parse.img_site(i) for i in urls]
grouped_df = tmp_df.groupby(
["plate_name", "plate_num", "well", "site"])
if order is True:
# order by channel
for _, group in grouped_df:
grouped = list(group["img_url"])
channel_nums = [parse.img_channel(i) for i in grouped]
# create tuple(path, channel_number) and sort by channel_number
sort_im = sorted(zip(grouped, channel_nums),
key=lambda x: x[1])
# return only the file-paths back from the list of tuples
grouped_list.append([i[0] for i in sort_im])
elif order is False:
for _, group in grouped_df:
grouped_list.append(list(group["img_url"]))
else:
raise ValueError("order needs to be a boolean")
return grouped_list
def keep_channels(img_list, channels):
"""
given a list of image paths, this will keep specified channel numbers,
and remove all others.
Parameters:
-----------
img_list : list
list of image URLs
channels : list of integers
list of channel numbers to keep
Returns:
--------
list of image URLs
"""
# find if img_urls are full paths or just filenames
if utils.is_full_path(img_list[0]):
just_file_path = [parse.img_filename(i) for i in img_list]
else:
just_file_path = img_list
channel_nums = [parse.img_channel(i) for i in just_file_path]
# make sure we zip the original img_list, *not* just_file_path
ch_img_tup = zip(channel_nums, img_list)
filtered_tup = [i for i in ch_img_tup if i[0] in channels]
_, img_urls = zip(*filtered_tup)
return img_urls
def test_ImageDict_keep_channels():
channels_to_keep = [1, 2, 3]
ImgDict = image_prep.ImageDict()
ans = ImgDict.keep_channels(IMG_URLS, channels_to_keep)
# parse channel numbers out of ans
img_names = [parse.img_filename(f) for f in ans]
img_channels = [parse.img_channel(name) for name in img_names]
for channel in img_channels:
assert channel in channels_to_keep
def test_ImageDict_remove_channels():
channels_to_remove = [4, 5]
ImgDict = image_prep.ImageDict()
ans = ImgDict.remove_channels(IMG_URLS, channels_to_remove)
# parse channel numbers out of ans
img_names = [parse.img_filename(f) for f in ans]
img_channels = [parse.img_channel(name) for name in img_names]
for channel in img_channels:
assert channel not in channels_to_remove
def _group_images(df_img):
"""group images by well and site"""
grouped_list = []
for _, group in df_img.groupby(["Metadata_well", "Metadata_site"]):
grouped = list(group["img_paths"])
channel_nums = [_parse.img_channel(i) for i in grouped]
# create tuple (path, channel_number) and sort by channel number
sort_im = sorted(list(zip(grouped, channel_nums)), key=lambda x: x[1])
# return on the file-paths back from the list of tuples
grouped_list.append([i[0] for i in sort_im])
return grouped_list
def test_ImageDict_sort_channels():
ImgDict = image_prep.ImageDict()
# un-sorted channels
# reverse channels as already sorted
rev_img_urls = IMG_URLS[::-1]
ImgDict.add_class("foo", rev_img_urls)
ImgDict.group_image_channels(order=False)
order_false_dict = ImgDict.parent_dict
order_false_vals = order_false_dict["foo"][0]
order_false_chnnls = [parse.img_channel(val) for val in order_false_vals]
assert sorted(order_false_chnnls) != order_false_chnnls
# sort channels
# need to create new ImageDict class otherwise we get a warning due to
# adding a new class to already grouped data
ImgDict2 = image_prep.ImageDict()
ImgDict2.add_class("bar", IMG_URLS)
ImgDict2.group_image_channels(order=True)
order_true_dict = ImgDict2.parent_dict
order_true_vals = order_true_dict["bar"][0]
order_true_chnnls = [parse.img_channel(val) for val in order_true_vals]
assert sorted(order_true_chnnls) == order_true_chnnls
def create_long_loaddata(img_list):
"""
create a dataframe of image paths with metadata columns
"""
just_filenames = [_parse.img_filename(i) for i in img_list]
df_img = _pd.DataFrame({
"URL":
just_filenames,
"path": [_parse.path(i) for i in img_list],
"Metadata_platename": [_parse.plate_name(i) for i in img_list],
"Metadata_well": [_parse.img_well(i) for i in just_filenames],
"Metadata_site": [_parse.img_site(i) for i in just_filenames],
"Metadata_channel": [_parse.img_channel(i) for i in just_filenames],
"Metadata_platenum": [_parse.plate_num(i) for i in img_list]
})
return df_img
def test_img_channel():
filename = parse.img_filename(EXAMPLE_PATH)
new_filename = parse.img_filename(NEW_EXAMPLE_PATH)
assert parse.img_channel(filename) == 1
assert parse.img_channel(new_filename) == 1