File size: 7,638 Bytes
7062b81 2bbc3ee |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 |
'''
This code is partially borrowed from IFRNet (https://github.com/ltkong218/IFRNet).
'''
import re
import sys
import torch
import random
import numpy as np
from PIL import ImageFile
import torch.nn.functional as F
from imageio import imread, imwrite
ImageFile.LOAD_TRUNCATED_IMAGES = True
class InputPadder:
""" Pads images such that dimensions are divisible by divisor """
def __init__(self, dims, divisor=16):
self.ht, self.wd = dims[-2:]
pad_ht = (((self.ht // divisor) + 1) * divisor - self.ht) % divisor
pad_wd = (((self.wd // divisor) + 1) * divisor - self.wd) % divisor
self._pad = [pad_wd//2, pad_wd - pad_wd//2, pad_ht//2, pad_ht - pad_ht//2]
def pad(self, *inputs):
if len(inputs) == 1:
return F.pad(inputs[0], self._pad, mode='replicate')
else:
return [F.pad(x, self._pad, mode='replicate') for x in inputs]
def unpad(self, *inputs):
if len(inputs) == 1:
return self._unpad(inputs[0])
else:
return [self._unpad(x) for x in inputs]
def _unpad(self, x):
ht, wd = x.shape[-2:]
c = [self._pad[2], ht-self._pad[3], self._pad[0], wd-self._pad[1]]
return x[..., c[0]:c[1], c[2]:c[3]]
def img2tensor(img):
return torch.tensor(img).permute(2, 0, 1).unsqueeze(0) / 255.0
def tensor2img(img_t):
return (img_t * 255.).detach(
).squeeze(0).permute(1, 2, 0).cpu().numpy(
).clip(0, 255).astype(np.uint8)
def read(file):
if file.endswith('.float3'): return readFloat(file)
elif file.endswith('.flo'): return readFlow(file)
elif file.endswith('.ppm'): return readImage(file)
elif file.endswith('.pgm'): return readImage(file)
elif file.endswith('.png'): return readImage(file)
elif file.endswith('.jpg'): return readImage(file)
elif file.endswith('.pfm'): return readPFM(file)[0]
else: raise Exception('don\'t know how to read %s' % file)
def write(file, data):
if file.endswith('.float3'): return writeFloat(file, data)
elif file.endswith('.flo'): return writeFlow(file, data)
elif file.endswith('.ppm'): return writeImage(file, data)
elif file.endswith('.pgm'): return writeImage(file, data)
elif file.endswith('.png'): return writeImage(file, data)
elif file.endswith('.jpg'): return writeImage(file, data)
elif file.endswith('.pfm'): return writePFM(file, data)
else: raise Exception('don\'t know how to write %s' % file)
def readPFM(file):
file = open(file, 'rb')
color = None
width = None
height = None
scale = None
endian = None
header = file.readline().rstrip()
if header.decode("ascii") == 'PF':
color = True
elif header.decode("ascii") == 'Pf':
color = False
else:
raise Exception('Not a PFM file.')
dim_match = re.match(r'^(\d+)\s(\d+)\s$', file.readline().decode("ascii"))
if dim_match:
width, height = list(map(int, dim_match.groups()))
else:
raise Exception('Malformed PFM header.')
scale = float(file.readline().decode("ascii").rstrip())
if scale < 0:
endian = '<'
scale = -scale
else:
endian = '>'
data = np.fromfile(file, endian + 'f')
shape = (height, width, 3) if color else (height, width)
data = np.reshape(data, shape)
data = np.flipud(data)
return data, scale
def writePFM(file, image, scale=1):
file = open(file, 'wb')
color = None
if image.dtype.name != 'float32':
raise Exception('Image dtype must be float32.')
image = np.flipud(image)
if len(image.shape) == 3 and image.shape[2] == 3:
color = True
elif len(image.shape) == 2 or len(image.shape) == 3 and image.shape[2] == 1:
color = False
else:
raise Exception('Image must have H x W x 3, H x W x 1 or H x W dimensions.')
file.write('PF\n' if color else 'Pf\n'.encode())
file.write('%d %d\n'.encode() % (image.shape[1], image.shape[0]))
endian = image.dtype.byteorder
if endian == '<' or endian == '=' and sys.byteorder == 'little':
scale = -scale
file.write('%f\n'.encode() % scale)
image.tofile(file)
def readFlow(name):
if name.endswith('.pfm') or name.endswith('.PFM'):
return readPFM(name)[0][:,:,0:2]
f = open(name, 'rb')
header = f.read(4)
if header.decode("utf-8") != 'PIEH':
raise Exception('Flow file header does not contain PIEH')
width = np.fromfile(f, np.int32, 1).squeeze()
height = np.fromfile(f, np.int32, 1).squeeze()
flow = np.fromfile(f, np.float32, width * height * 2).reshape((height, width, 2))
return flow.astype(np.float32)
def readImage(name):
if name.endswith('.pfm') or name.endswith('.PFM'):
data = readPFM(name)[0]
if len(data.shape)==3:
return data[:,:,0:3]
else:
return data
return imread(name)
def writeImage(name, data):
if name.endswith('.pfm') or name.endswith('.PFM'):
return writePFM(name, data, 1)
return imwrite(name, data)
def writeFlow(name, flow):
f = open(name, 'wb')
f.write('PIEH'.encode('utf-8'))
np.array([flow.shape[1], flow.shape[0]], dtype=np.int32).tofile(f)
flow = flow.astype(np.float32)
flow.tofile(f)
def readFloat(name):
f = open(name, 'rb')
if(f.readline().decode("utf-8")) != 'float\n':
raise Exception('float file %s did not contain <float> keyword' % name)
dim = int(f.readline())
dims = []
count = 1
for i in range(0, dim):
d = int(f.readline())
dims.append(d)
count *= d
dims = list(reversed(dims))
data = np.fromfile(f, np.float32, count).reshape(dims)
if dim > 2:
data = np.transpose(data, (2, 1, 0))
data = np.transpose(data, (1, 0, 2))
return data
def writeFloat(name, data):
f = open(name, 'wb')
dim=len(data.shape)
if dim>3:
raise Exception('bad float file dimension: %d' % dim)
f.write(('float\n').encode('ascii'))
f.write(('%d\n' % dim).encode('ascii'))
if dim == 1:
f.write(('%d\n' % data.shape[0]).encode('ascii'))
else:
f.write(('%d\n' % data.shape[1]).encode('ascii'))
f.write(('%d\n' % data.shape[0]).encode('ascii'))
for i in range(2, dim):
f.write(('%d\n' % data.shape[i]).encode('ascii'))
data = data.astype(np.float32)
if dim==2:
data.tofile(f)
else:
np.transpose(data, (2, 0, 1)).tofile(f)
def warp(img, flow):
B, _, H, W = flow.shape
xx = torch.linspace(-1.0, 1.0, W).view(1, 1, 1, W).expand(B, -1, H, -1)
yy = torch.linspace(-1.0, 1.0, H).view(1, 1, H, 1).expand(B, -1, -1, W)
grid = torch.cat([xx, yy], 1).to(img)
flow_ = torch.cat([flow[:, 0:1, :, :] / ((W - 1.0) / 2.0), flow[:, 1:2, :, :] / ((H - 1.0) / 2.0)], 1)
grid_ = (grid + flow_).permute(0, 2, 3, 1)
output = F.grid_sample(input=img, grid=grid_, mode='bilinear', padding_mode='border', align_corners=True)
return output
def check_dim_and_resize(tensor_list):
shape_list = []
for t in tensor_list:
shape_list.append(t.shape[2:])
if len(set(shape_list)) > 1:
desired_shape = shape_list[0]
print(f'Inconsistent size of input video frames. All frames will be resized to {desired_shape}')
resize_tensor_list = []
for t in tensor_list:
resize_tensor_list.append(torch.nn.functional.interpolate(t, size=tuple(desired_shape), mode='bilinear'))
tensor_list = resize_tensor_list
return tensor_list
|