roi_align.py 3.24 KB
Newer Older
1
import torch
2
from torch import nn, Tensor
3
4

from torch.nn.modules.utils import _pair
5
from torch.jit.annotations import BroadcastingList2
6

7
from torchvision.extension import _assert_has_ops
8
from ._utils import convert_boxes_to_roi_format, check_roi_boxes_shape
9
10


11
12
13
14
15
16
17
18
def roi_align(
    input: Tensor,
    boxes: Tensor,
    output_size: BroadcastingList2[int],
    spatial_scale: float = 1.0,
    sampling_ratio: int = -1,
    aligned: bool = False,
) -> Tensor:
19
20
21
    """
    Performs Region of Interest (RoI) Align operator described in Mask R-CNN

22
    Args:
23
        input (Tensor[N, C, H, W]): input tensor
24
        boxes (Tensor[K, 5] or List[Tensor[L, 4]]): the box coordinates in (x1, y1, x2, y2)
25
26
27
28
29
30
31
32
33
34
35
36
37
            format where the regions will be taken from. If a single Tensor is passed,
            then the first column should contain the batch index. If a list of Tensors
            is passed, then each Tensor will correspond to the boxes for an element i
            in a batch
        output_size (int or Tuple[int, int]): the size of the output after the cropping
            is performed, as (height, width)
        spatial_scale (float): a scaling factor that maps the input coordinates to
            the box coordinates. Default: 1.0
        sampling_ratio (int): number of sampling points in the interpolation grid
            used to compute the output value of each pooled output bin. If > 0,
            then exactly sampling_ratio x sampling_ratio grid points are used. If
            <= 0, then an adaptive number of grid points are used (computed as
            ceil(roi_width / pooled_w), and likewise for height). Default: -1
AhnDW's avatar
AhnDW committed
38
39
40
        aligned (bool): If False, use the legacy implementation.
            If True, pixel shift it by -0.5 for align more perfectly about two neighboring pixel indices.
            This version in Detectron2
41
42
43
44

    Returns:
        output (Tensor[K, C, output_size[0], output_size[1]])
    """
45
    _assert_has_ops()
46
    check_roi_boxes_shape(boxes)
47
    rois = boxes
48
    output_size = _pair(output_size)
49
50
    if not isinstance(rois, torch.Tensor):
        rois = convert_boxes_to_roi_format(rois)
51
52
    return torch.ops.torchvision.roi_align(input, rois, spatial_scale,
                                           output_size[0], output_size[1],
AhnDW's avatar
AhnDW committed
53
                                           sampling_ratio, aligned)
54
55
56
57
58
59


class RoIAlign(nn.Module):
    """
    See roi_align
    """
60
61
62
63
64
65
66
    def __init__(
        self,
        output_size: BroadcastingList2[int],
        spatial_scale: float,
        sampling_ratio: int,
        aligned: bool = False,
    ):
67
68
69
70
        super(RoIAlign, self).__init__()
        self.output_size = output_size
        self.spatial_scale = spatial_scale
        self.sampling_ratio = sampling_ratio
AhnDW's avatar
AhnDW committed
71
        self.aligned = aligned
72

73
    def forward(self, input: Tensor, rois: Tensor) -> Tensor:
AhnDW's avatar
AhnDW committed
74
        return roi_align(input, rois, self.output_size, self.spatial_scale, self.sampling_ratio, self.aligned)
75

76
    def __repr__(self) -> str:
77
78
79
80
        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)
AhnDW's avatar
AhnDW committed
81
        tmpstr += ', aligned=' + str(self.aligned)
82
83
        tmpstr += ')'
        return tmpstr