def proc_label():
label_list = file_io.get_listfile(ucsd_label_path, label_ext)
for l_name in label_list:
mat = sio.loadmat(l_name)
desmap = mat["gtDensities"]
desmap = desmap.astype(np.float32)
new_label_name = l_name.replace(label_ext, new_label_ext)
desmap.tofile(new_label_name)
def sample(avid_dir_list):
avi_dir_list = file_io.get_dir_list(avid_dir_list)
for avi_dir in avi_dir_list:
avi_file_list = file_io.get_listfile(avi_dir, ".avi")
avi_file_list.sort()
for avi in avi_file_list:
image_dir = avi.replace(".avi", "")
command = "ffmpeg -i " + avi + " " + image_dir + "/%06d.jpg"
os.system(command)
if __name__ == "__main__":
if not len(sys.argv) == 3:
print("Usage: label_preprocess.py data_dir size_len")
exit(1)
data_dir = sys.argv[1]
size_len = int(sys.argv[2])
global dsize
dsize = (size_len, size_len)
mask_dir_list = file_io.get_dir_list(data_dir)
for mask_dir in mask_dir_list:
mask_list = file_io.get_listfile(mask_dir, ".msk")
for mask in mask_list:
image_dir_name = mask.replace(".msk", "")
image_list = file_io.get_listfile(image_dir_name, "jpg")
mask_name = mask.replace(".msk", ".png")
mask_bbox = load_mask(mask_name)
# if mask == "../data/data_new/Training_Data/Cam181/01.msk":
# mask_bin = cv2.imread(mask_dir + "/01_mask.jpg")
# mask_bin = mask_bin[:,:,1]
# mask_bin /= 255
# else:
# try:
# mask_pts = load_mask(mask)
# mask_bin = gen_mask_image(mask_pts)
# except:
# mask_pts = load_mask(mask)
im_color = cv2.applyColorMap(desmap, cv2.COLORMAP_JET)
return im_color
def cen_crop(input_image, dsize):
h, w, c = input_image.shape
offset_h = int((h - dsize[0]) / 2)
offset_w = int((w - dsize[1]) / 2)
input_image = input_image[offset_h:offset_h + dsize[0],
offset_w:offset_w + dsize[1], :]
return input_image
ucsd_dir = "/media/dog/data/UCSD/images/"
desmap_list = file_io.get_listfile(ucsd_dir, "infer_desmap")
mask_name = "/media/dog/data/UCSD/Mask/mask227.npy"
mask = np.fromfile(mask_name, np.float32)
mask = np.expand_dims(np.reshape(mask, (227, 227)), 2)
mask = cen_crop(mask, (224, 224))
mask = np.squeeze(mask)
for des_name in desmap_list:
desmap = opencv_plot(des_name, mask)
img_name = des_name.replace(".infer_desmap", ".jpg")
img = cv2.imread(img_name)
img = cen_crop(img, (224, 224))
combine_img = np.hstack((img, desmap))
save_name = img_name.replace(".jpg", "_combine.png")
cv2.imwrite(save_name, combine_img)
def cen_crop(input_image, dsize):
h, w, c = input_image.shape
offset_h = int((h - dsize[0])/2)
offset_w = int((w - dsize[1])/2)
input_image = input_image[offset_h:offset_h + dsize[0], offset_w:offset_w+dsize[1], :]
return input_image
desmap_dir = "/media/dog/data/WebCamT_60000"
cam_list = file_io.get_dir_list(desmap_dir)
for cam_dir in cam_list:
#cam_dir = "/Users/Geoff/Documents/my_git/data/desmap/253/"
video_list = file_io.get_dir_list(cam_dir)
for video in video_list:
img_list = file_io.get_listfile(video, "_256.jpg")
mask_name = video+ "_msk_256.npy"
if not os.path.exists(mask_name):
continue
mask = np.fromfile(mask_name, np.float32)
mask = np.expand_dims(np.reshape(mask, (256, 256)), 2)
mask = cen_crop(mask, (227,227))
mask = np.squeeze(mask)
#mask = np.tile(mask, (1,1,3))
for img_name in img_list:
desmap_name = img_name.replace(".jpg", ".infer_desmap")
if not os.path.exists(desmap_name):
continue
desmap_name = img_name.replace(".jpg", ".desmap")
desmap = opencv_plot(desmap_name, mask)
def proc_image():
image_list = file_io.get_listfile(ucsd_image_path, image_ext)
for img_name in image_list:
img = cv2.imread(img_name)
new_image_name = img_name.replace(image_ext, ".jpg")
cv2.imwrite(new_image_name, img)
print("Usage: label_preprocess.py data_dir size_len")
exit(1)
data_dir = sys.argv[1]
size_len = int(sys.argv[2])
dsize = (size_len, size_len)
data_ext = "_%d.jpg" % size_len
label_ext = "_%d.desmap" % size_len
seg_ext = "_%d.segmap" % size_len
cam_dir_list = file_io.get_dir_list(data_dir)
train_list = list()
test_list = list()
full_file_list = list()
for cam_dir in cam_dir_list:
video_list = file_io.get_listfile(cam_dir, ".avi")
for file_name in video_list:
data_dir_name = file_name.replace(".avi", "")
curr_data_list = file_io.get_listfile(data_dir_name, label_ext)
for desmap_filename in curr_data_list:
desmap = np.fromfile(desmap_filename, np.float32)
desmap = np.reshape(desmap, dsize)
segmap = desmap > 0
segmap = segmap.astype(np.float32)
segmap_filename = desmap_filename.replace(label_ext, seg_ext)
segmap.tofile(segmap_filename)
#cv2.imshow("image", segmap)
#cv2.waitKey(0)
import numpy as np
import cv2
from TensorflowToolbox.utility import file_io
desmap_dir = "/media/dog/data/UCSD/gtDensities/"
desmap_list = file_io.get_listfile(desmap_dir, "desmap")
new_list = list()
for f in desmap_list:
num = np.sum(np.fromfile(f, np.float32))
format_num = "%.2f" % num
new_list.append(f + " " + format_num)
new_list.sort()
file_io.save_file(new_list, "ucsd_file_label_num.txt")
return reverse_image_list
if __name__ == "__main__":
image_ext = ".jpg"
desmap_ext = ".desmap"
mask_ext = "_mask.npy"
file_list_dir = "../file_list/"
unroll_num = 5
mul_mask_name = " ".join([mask_file_name] * 5)
save_train_file_name = "ucsd_train_list1.txt"
save_test_file_name = "ucsd_test_list1.txt"
image_list = file_io.get_listfile(ucsd_image_path, image_ext)
image_list.sort()
image_list = file_list_to_list(image_list, unroll_num)
reverse_list = reverse_copy_list(image_list)
file_list = file_list_to_train_list(image_list)
reverse_file_list = file_list_to_train_list(reverse_list)
#train_list = reverse_file_list[600:1400]
train_list = file_list[600:1400] #+ reverse_file_list[600:1400]
test_list = file_list[0:600] + file_list[1400:]
file_io.save_file(train_list, file_list_dir + save_train_file_name, True)
img = cv2.fillPoly(img, [pts], (255, 255, 255))
# cv2.imshow("image", img / 255)
# cv2.waitKey(0)
cv2.imwrite(pic_name, img)
if __name__ == "__main__":
if len(sys.argv) > 1:
data_dir = sys.argv[1]
else:
data_dir = "../data"
cam_list = file_io.get_dir_list(data_dir)
for cam in cam_list:
print(cam)
msk_list = file_io.get_listfile(cam, "msk")
for msk in msk_list:
pic_name = msk.replace(".msk", ".png")
f = open(msk, "r")
lines = f.readlines()
f.close()
points = []
for line in lines:
if "[" in line and "," in line and "]" in line:
a = line.replace("[", "").replace("]", "").replace(
"\n", "").replace("\r", "").split(",")
x = int(a[0])
y = int(a[1])
if x <= 2:
x = 0
if x >= 349:
if __name__ == "__main__":
if not len(sys.argv) == 3:
print("Usage: msk_resize.py data_dir size_len")
exit(1)
data_dir = sys.argv[1]
size_len = int(sys.argv[2])
dsize = (size_len, size_len)
#if len(sys.argv) > 1:
# data_dir = sys.argv[1]
#else:
# data_dir = "../data"
mask_dir_list = file_io.get_dir_list(data_dir)
for mask_dir in mask_dir_list:
mask_list = file_io.get_listfile(mask_dir, ".msk")
for mask in mask_list:
mask_img_name = mask.replace("msk", "png")
mask_img = cv2.imread(mask_img_name, 0)
bbox = image_utility_func.get_bbox(mask_img, 127)
mask_img = mask_img[bbox[1]: bbox[1] + bbox[3],
bbox[0]: bbox[0] + bbox[2]]
mask_img = cv2.resize(mask_img, dsize)
mask_img = mask_img / 255
mask_img = mask_img.astype(np.float32)
mask_img[mask_img > 0] = 1.0
np_name = mask_img_name.replace(".png", "_msk_%d.npy"%size_len)
mask_img.tofile(np_name)
label_path = "/media/dog/data/WorldExpo/train_label/"
mask_dir = "/media/dog/data/WorldExpo/mask/"
desmap_dir = "/media/dog/data/WorldExpo/gtDensity/gtDensity/"
img_dir = "/media/dog/data/WorldExpo/train_frame/"
else:
label_path = "/media/dog/data/WorldExpo/test_label/"
mask_dir = "/media/dog/data/WorldExpo/test_mask/"
desmap_dir = "/media/dog/data/WorldExpo/test_gtDensity/test_gtDensity/"
img_dir = "/media/dog/data/WorldExpo/test_frame/"
if i == 0:
file_list = train_list
else:
file_list = test_list
img_list = file_io.get_listfile(img_dir, ".jpg")
scale_str = str(new_dsize[0])
for img_name in img_list:
if not img_name.endswith("_" + scale_str + ".jpg"):
continue
label_name = img_name.replace(img_dir,
desmap_dir).replace(".jpg", ".npy")
mask_name = mask_dir + img_name.split("/")[-1][:6] \
+ "_mask_" + scale_str + ".npy"
if not os.path.exists(label_name):
print(label_name, "is not exist")
exit(1)
if not os.path.exists(mask_name):
