Datasets:

Languages:
English
ArXiv:
License:
OCTScenes / preprocess.py
Yinxuan's picture
Create preprocess.py
05fa938
import numpy as np
from PIL import Image
def crop_and_resize(image_path, bbox_path, crop_path, resize_path):
img = Image.open(image_path)
width, height = img.size
with open(bbox_path, 'r') as f:
_, x_center, y_center, _, _ = map(float, f.readline().split())
x_center, y_center = int(x_center * width), int(y_center * height)
x1, x2 = max(0, x_center - 128), min(width, x_center + 128)
y1, y2 = max(0, y_center - 128), min(height, y_center + 128)
crop = img.crop((x1, y1, x2, y2))
assert crop.size == (256, 256)
crop.save(crop_path)
resize = crop.resize(size=(128, 128), resample=Image.BICUBIC)
assert resize.size == (128, 128)
resize.save(resize_path)
def transform_intrinsic(intrinsic_path, bbox_path, crop_path, resize_path):
intrinsic = np.loadtxt(intrinsic_path)
fx, fy = intrinsic[0, 0], intrinsic[1, 1]
cx, cy = intrinsic[0, 2], intrinsic[1, 2]
width, height = cx * 2, cy * 2
with open(bbox_path, 'r') as f:
_, x_center, y_center, _, _ = map(float, f.readline().split())
x_center, y_center = int(x_center * width), int(y_center * height)
x1, y1 = max(0, x_center - 128), max(0, y_center - 128)
K = np.array([[fx, 0, cx - x1],
[0, fy, cy - y1],
[0, 0, 1]])
np.savetxt(crop_path, K, fmt="%.5f", delimiter=" ")
K[:2] /= 2
np.savetxt(resize_path, K, fmt="%.5f", delimiter=" ")
if __name__ == "__main__":
# crop and resize image
image_path = '640x480/image/0000_00.png'
bbox_path = '640x480/bbox/0000_00.txt'
crop_path = '256x256/image/0000_00.png'
resize_path = '128x128/image/0000_00.png'
crop_and_resize(image_path, bbox_path, crop_path, resize_path)
# transform intrinsic matrix
intrinsic_path = '640x480/intrinsic.txt'
bbox_path = '640x480/bbox/0000_00.txt'
crop_path = '256x256/intrinsic/0000_00.txt'
resize_path = '128x128/intrinsic/0000_00.txt'
transform_intrinsic(intrinsic_path, bbox_path, crop_path, resize_path)