import numpy as np

from skimage import img_as_ubyte

from skimage.transform import resize

def bb_intersection_over_union(boxA, boxB):

xA = max(boxA[0], boxB[0])

yA = max(boxA[1], boxB[1])

xB = min(boxA[2], boxB[2])

yB = min(boxA[3], boxB[3])

interArea = max(0, xB - xA + 1) * max(0, yB - yA + 1)

boxAArea = (boxA[2] - boxA[0] + 1) * (boxA[3] - boxA[1] + 1)

boxBArea = (boxB[2] - boxB[0] + 1) * (boxB[3] - boxB[1] + 1)

iou = interArea / float(boxAArea + boxBArea - interArea)

return iou

def one_box_inside_other(boxA, boxB):

xA = boxA[0] <= boxB[0]

yA = boxA[1] <= boxB[1]

xB = boxA[2] >= boxB[2]

yB = boxA[3] >= boxB[3]

return xA and yA and xB and yB

def join(tube_bbox, bbox):

xA = min(tube_bbox[0], bbox[0])

yA = min(tube_bbox[1], bbox[1])

xB = max(tube_bbox[2], bbox[2])

yB = max(tube_bbox[3], bbox[3])

return (xA, yA, xB, yB)

def compute_aspect_preserved_bbox(bbox, increase_area):

left, top, right, bot = bbox

width = right - left

height = bot - top

width_increase = max(increase_area, ((1 + 2 * increase_area) * height - width) / (2 * width))

height_increase = max(increase_area, ((1 + 2 * increase_area) * width - height) / (2 * height))

left = int(left - width_increase * width)

top = int(top - height_increase * height)

right = int(right + width_increase * width)

bot = int(bot + height_increase * height)

return (left, top, right, bot)

def crop_bbox_from_frames(frame_list, tube_bbox, min_frames=16, image_shape=(256, 256), min_size=200,

increase_area=0.1, max_pad=10):

frame_shape = frame_list[0].shape

# Filter short sequences

if len(frame_list) < min_frames:

return None, None

left, top, right, bot = tube_bbox

width = right - left

height = bot - top

# Filter if it is too small

if max(width, height) < min_size:

return None, None

left, top, right, bot = compute_aspect_preserved_bbox(tube_bbox, increase_area)

# If something is out of bounds, pad with white

left_oob = -min(0, left)

right_oob = right - min(right, frame_shape[1])

top_oob = -min(0, top)

bot_oob = bot - min(bot, frame_shape[0])

left += left_oob

right += left_oob

top += top_oob

bot += top_oob

#Not use near the border

if max(left_oob / float(width), right_oob / float(width), top_oob / float(height), bot_oob / float(height)) > max_pad:

return None, None

selected = [frame[top:bot, left:right] for frame in frame_list]

out = [img_as_ubyte(resize(frame, image_shape, anti_aliasing=True)) for frame in selected]

return out, [left, top, right, bot]

from multiprocessing import Pool

from itertools import cycle

from tqdm import tqdm

def scheduler(data_list, fn, args):

device_ids = args.device_ids.split(",")

pool = Pool(processes=args.workers)

args_list = cycle([args])

f = open(args.chunks_metadata, 'w')

line = "{video_id},{start},{end},{bbox},{fps},{width},{height}"

print (line.replace('{', '').replace('}', ''), file=f)

for chunks_data in tqdm(pool.imap_unordered(fn, zip(data_list, cycle(device_ids), args_list))):

for data in chunks_data:

print (line.format(**data), file=f)

f.flush()

f.close()