def load_demo_images(paths):
paths = [os.path.join(paths, x) for x in os.listdir(paths)]
img_h = cfg.CONST.IMG_H
img_w = cfg.CONST.IMG_W
model = torchvision.models.detection.maskrcnn_resnet50_fpn(pretrained=True)
"""
imgs = []
for path in paths:
img = Image.open(path)
#img = segment(path)
#img = transforms.ToPILImage()(img)
img = img.resize((img_h, img_w), Image.ANTIALIAS)
img = preprocess_img(img, train=False)
imgs.append([np.array(img).transpose( \
(2, 0, 1)).astype(np.float32)])
"""
imgs = segment(paths, model)
result = []
for img in imgs:
img = transforms.ToPILImage()(img)
img = img.resize((img_h, img_w), Image.ANTIALIAS)
img = preprocess_img(img, train=False)
result.append([np.array(img).transpose( \
(2, 0, 1)).astype(np.float32)])
ims_np = np.array(result).astype(np.float32)
return torch.from_numpy(ims_np)
def load_demo_images():
ims = []
for i in range(3):
im = Image.open('imgs/%02d.png' % i)
im = preprocess_img(im, train=False)[0]
ims.append([np.array(im).transpose((2, 0, 1)).astype(np.float32)])
return np.array(ims)
def load_demo_images():
ims = []
for i in range(3):
im = preprocess_img(Image.open('imgs/%d.jpg' % i).resize((127, 127)),
train=False)
ims.append([np.array(im).transpose((2, 0, 1)).astype(np.float32)])
plt.imshow(im)
plt.show()
return np.array(ims)
def load_demo_images(imgs='./imgs/', maxrange=3):
global demo_imgs
ims = []
# print("maxrange", maxrange)
for i in range(maxrange):
im = preprocess_img(
# Image.open(imgs + '%d.jpg' % #进来的时候是127*127*3
Image.open(imgs + '0%d.png' % #进来的时候是127*127*3
i).resize((127, 127)),
train=False)
ims.append([np.array(im).transpose((2, 0, 1)).astype(np.float32)])
# return np.array(ims)
demo_imgs = np.array(ims)
def load_demo_images():
img_h = cfg.CONST.IMG_H
img_w = cfg.CONST.IMG_W
imgs = []
for i in range(3):
img = Image.open('imgs/%d.png' % i)
img = img.resize((img_h, img_w), Image.ANTIALIAS)
img = preprocess_img(img, train=False)
imgs.append([np.array(img).transpose( \
(2, 0, 1)).astype(np.float32)])
ims_np = np.array(imgs).astype(np.float32)
return torch.from_numpy(ims_np)
def load_imgs(taken_time):
assert (os.path.isdir(taken_time))
img_h = cfg.CONST.IMG_H
img_w = cfg.CONST.IMG_W
imgs = []
for file in os.listdir(taken_time):
if file[-4:] in [".JPG", ".png"]:
img = Image.open(os.path.join(taken_time, file))
img = img.resize((img_h, img_w), Image.ANTIALIAS)
img = preprocess_img(img, train=False)
imgs.append([np.array(img).transpose( \
(2, 0, 1)).astype(np.float32)])
plt.imshow(img)
plt.show()
return np.array(imgs)
def load_img(self, category, model_id, image_id):
image_fn = get_rendering_file(category, model_id, image_id)
im = Image.open(image_fn)
t_im = preprocess_img(im, self.train)
return t_im
def load_img(category, model_id, img_id, train=False):
img_fn = get_rendering_file(category, model_id, img_id)
im = Image.open(img_fn)
t_im = preprocess_img(im, train=train)
return t_im
