def done(self):
Path(self.directory).mkdir(parents=True, exist_ok=True)
stream = torch.from_numpy(np.stack(self.t))
for image in stream:
write_png(image.permute(2, 0, 1),
f'{self.directory}/{self.image_id}.png')
self.image_id += 1
def example_load_frame():
v, a, info = read_video("/home/tt/Videos/VID_20201202_133703_090.mp4",
pts_unit='sec')
print(v.shape) # torch.Size([467, 1080, 1920, 3])
# write a frame
single_frame = v[100]
print(single_frame.shape) # torch.Size([1080, 1920, 3])
single_frame = single_frame.permute(2, 0, 1) # to CHW
print(single_frame.shape)
file_out = os.path.join(out_path, "single_frame.png")
write_png(single_frame, file_out)
print("done write to ", file_out)
def generate_image(feature, save_as_numpy=False):
save_as_numpy = feature.get("save_as_numpy", save_as_numpy)
try:
from torchvision.io import write_png
except ImportError:
logger.error(" torchvision is not installed. "
"In order to install all image feature dependencies run "
"pip install ludwig[image]")
sys.exit(-1)
# Read num_channels, width, height
destination_folder = feature.get("destination_folder", "image_files")
if PREPROCESSING in feature:
height = feature[PREPROCESSING].get("height", 28)
width = feature[PREPROCESSING].get("width", 28)
num_channels = feature[PREPROCESSING].get("num_channels", 1)
else:
height = feature.get("height", 28)
width = feature.get("width", 28)
num_channels = feature.get("num_channels", 1)
if width <= 0 or height <= 0 or num_channels < 1:
raise ValueError("Invalid arguments for generating images")
# Create a Random Image
img = torch.randint(0,
255, (num_channels, width, height),
dtype=torch.uint8)
# Generate a unique random filename
image_filename = uuid.uuid4().hex[:10].upper() + ".png"
# Save the image to disk either in a specified location/new folder
try:
if not os.path.exists(destination_folder):
os.makedirs(destination_folder)
image_dest_path = os.path.join(destination_folder, image_filename)
# save_image(torch.from_numpy(img.astype("uint8")), image_dest_path)
if save_as_numpy:
with open(image_dest_path, "wb") as f:
np.save(f, img.detach().cpu().numpy())
else:
write_png(img, image_dest_path)
except OSError as e:
raise OSError(
"Unable to create a folder for images/save image to disk."
"{}".format(e))
return image_dest_path
# Blog: https://blog.csdn.net/fengbingchun/article/details/121191194
# 下载MNIST数据集: torchvision.datasets包
test = datasets.MNIST("../../data", train=False, download=True)
train = datasets.MNIST("../../data", train=True, download=False)
print(f"raw_folder: test: {test.raw_folder}, train: {train.raw_folder}")
print(
f"processed_folder: test: {test.processed_folder}, train: {train.processed_folder}"
)
print(f"extra_repr:\ntest: {test.extra_repr}\ntrain: {train.extra_repr}")
print(f"class to index: {test.class_to_idx}")
# 读写图像: torchvision.io包
tensor = io.read_image("../../data/image/1.jpg")
print("tensor shape:", tensor.shape)
io.write_png(tensor, "../../data/image/result.png")
tensor = io.read_image("../../data/image/lena.png")
print("tensor shape:", tensor.shape)
io.write_jpeg(tensor, "../../data/image/result.jpg")
# 下载pre-trained AlexNet模型: torchvision.models包
net = models.alexnet(pretrained=True)
# 计算机视觉操作: torchvision.ops包
boxes = torch.tensor([[1, 1, 101, 101], [3, 5, 13, 15], [2, 4, 22, 44]])
area = ops.box_area(boxes)
print(f"area: {area}")
index = ops.remove_small_boxes(boxes, min_size=20)
print(f"index: {index}")
print("density_map_tensor.shape",
density_map_tensor.shape) # torch.Size([1, 1, 46, 82])
# module = nn.UpsamplingBilinear2d(scale_factor=8)
# upsampling_density_map_tensor = module(density_map_tensor)
upsampling_density_map_tensor = nn.functional.interpolate(density_map_tensor,
scale_factor=8) / 64
print(upsampling_density_map_tensor.sum())
print(upsampling_density_map_tensor.shape)
pad_density_map_tensor = torch.zeros(
(1, 3, img_tensor.shape[1], img_tensor.shape[2]))
pad_density_map_tensor[:, 0, :upsampling_density_map_tensor.
shape[2], :upsampling_density_map_tensor.
shape[3]] = upsampling_density_map_tensor
print(pad_density_map_tensor.shape)
pad_density_map_tensor = (pad_density_map_tensor.squeeze(dim=0) /
pad_density_map_tensor.max() * 255)
# pad_density_map_tensor = pad_density_map_tensor.squeeze(dim=0)
print(img_tensor.dtype)
print(pad_density_map_tensor.dtype)
overlay_density_map = img_tensor.detach().clone()
overlay_density_map[0] = torch.clamp_max(img_tensor[0] +
pad_density_map_tensor[0] * 2,
max=255)
write_png(overlay_density_map.type(torch.uint8),
"../visualize/pic/overlay.png")
write_png(pad_density_map_tensor.type(torch.uint8), "../visualize/pic/pad.png")