Spaces:
No application file
No application file
File size: 4,689 Bytes
430de99 |
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 |
# Copyright (c) Facebook, Inc. and its affiliates.
from torch import nn
from torchvision.ops import roi_align as tv_roi_align
try:
from torchvision import __version__
version = tuple(int(x) for x in __version__.split(".")[:2])
USE_TORCHVISION = version >= (0, 7) # https://github.com/pytorch/vision/pull/2438
except ImportError: # only open source torchvision has __version__
USE_TORCHVISION = True
if USE_TORCHVISION:
roi_align = tv_roi_align
else:
from torch.nn.modules.utils import _pair
from torch.autograd import Function
from torch.autograd.function import once_differentiable
from detectron2 import _C
class _ROIAlign(Function):
@staticmethod
def forward(ctx, input, roi, output_size, spatial_scale, sampling_ratio, aligned):
ctx.save_for_backward(roi)
ctx.output_size = _pair(output_size)
ctx.spatial_scale = spatial_scale
ctx.sampling_ratio = sampling_ratio
ctx.input_shape = input.size()
ctx.aligned = aligned
output = _C.roi_align_forward(
input, roi, spatial_scale, output_size[0], output_size[1], sampling_ratio, aligned
)
return output
@staticmethod
@once_differentiable
def backward(ctx, grad_output):
(rois,) = ctx.saved_tensors
output_size = ctx.output_size
spatial_scale = ctx.spatial_scale
sampling_ratio = ctx.sampling_ratio
bs, ch, h, w = ctx.input_shape
grad_input = _C.roi_align_backward(
grad_output,
rois,
spatial_scale,
output_size[0],
output_size[1],
bs,
ch,
h,
w,
sampling_ratio,
ctx.aligned,
)
return grad_input, None, None, None, None, None
roi_align = _ROIAlign.apply
# NOTE: torchvision's RoIAlign has a different default aligned=False
class ROIAlign(nn.Module):
def __init__(self, output_size, spatial_scale, sampling_ratio, aligned=True):
"""
Args:
output_size (tuple): h, w
spatial_scale (float): scale the input boxes by this number
sampling_ratio (int): number of inputs samples to take for each output
sample. 0 to take samples densely.
aligned (bool): if False, use the legacy implementation in
Detectron. If True, align the results more perfectly.
Note:
The meaning of aligned=True:
Given a continuous coordinate c, its two neighboring pixel indices (in our
pixel model) are computed by floor(c - 0.5) and ceil(c - 0.5). For example,
c=1.3 has pixel neighbors with discrete indices [0] and [1] (which are sampled
from the underlying signal at continuous coordinates 0.5 and 1.5). But the original
roi_align (aligned=False) does not subtract the 0.5 when computing neighboring
pixel indices and therefore it uses pixels with a slightly incorrect alignment
(relative to our pixel model) when performing bilinear interpolation.
With `aligned=True`,
we first appropriately scale the ROI and then shift it by -0.5
prior to calling roi_align. This produces the correct neighbors; see
detectron2/tests/test_roi_align.py for verification.
The difference does not make a difference to the model's performance if
ROIAlign is used together with conv layers.
"""
super(ROIAlign, self).__init__()
self.output_size = output_size
self.spatial_scale = spatial_scale
self.sampling_ratio = sampling_ratio
self.aligned = aligned
def forward(self, input, rois):
"""
Args:
input: NCHW images
rois: Bx5 boxes. First column is the index into N. The other 4 columns are xyxy.
"""
assert rois.dim() == 2 and rois.size(1) == 5
return roi_align(
input,
rois.to(dtype=input.dtype),
self.output_size,
self.spatial_scale,
self.sampling_ratio,
self.aligned,
)
def __repr__(self):
tmpstr = self.__class__.__name__ + "("
tmpstr += "output_size=" + str(self.output_size)
tmpstr += ", spatial_scale=" + str(self.spatial_scale)
tmpstr += ", sampling_ratio=" + str(self.sampling_ratio)
tmpstr += ", aligned=" + str(self.aligned)
tmpstr += ")"
return tmpstr
|