box_coder_utils.py

import torch


class ResidualCoder(object):
    def __init__(self, code_size=7, **kwargs):
        super().__init__()
        self.code_size = code_size
        assert code_size in [7, 9]

    @staticmethod
    def encode_torch(boxes, anchors):
        """
        Args:
            boxes: (N, 7 + C) [x, y, z, dx, dy, dz, heading, ...]
            anchors: (N, 7 + C) [x, y, z, dx, dy, dz, heading, ...]

        Returns:

        """
        anchors[:, 3:6] = torch.clamp_min(anchors[:, 3:6], min=1e-5)
        boxes[:, 3:6] = torch.clamp_min(boxes[:, 3:6], min=1e-5)

        xa, ya, za, dxa, dya, dza, ra, *cas = torch.split(anchors, 1, dim=-1)
        xg, yg, zg, dxg, dyg, dzg, rg, *cgs = torch.split(boxes, 1, dim=-1)

        diagonal = torch.sqrt(dxa ** 2 + dya ** 2)
        xt = (xg - xa) / diagonal
        yt = (yg - ya) / diagonal
        zt = (zg - za) / dza
        dxt = torch.log(dxg / dxa)
        dyt = torch.log(dyg / dya)
        dzt = torch.log(dzg / dza)
        rt = rg - ra

        cts = [g - a for g, a in zip(cgs, cas)]
        return torch.cat([xt, yt, zt, dxt, dyt, dzt, rt, *cts], dim=-1)

    @staticmethod
    def decode_torch(box_encodings, anchors):
        """
        Args:
            box_encodings: (B, N, 7 + C) or (N, 7 + C) [x, y, z, dx, dy, dz, heading, ...]
            anchors: (B, N, 7 + C) or (N, 7 + C) [x, y, z, dx, dy, dz, heading, ...]

        Returns:

        """
        xa, ya, za, dxa, dya, dza, ra, *cas = torch.split(anchors, 1, dim=-1)
        xt, yt, zt, dxt, dyt, dzt, rt, *cts = torch.split(box_encodings, 1, dim=-1)

        diagonal = torch.sqrt(dxa ** 2 + dya ** 2)
        xg = xt * diagonal + xa
        yg = yt * diagonal + ya
        zg = zt * dza + za

        dxg = torch.exp(dxt) * dxa
        dyg = torch.exp(dyt) * dya
        dzg = torch.exp(dzt) * dza
        rg = rt + ra

        cgs = [t + a for t, a in zip(cts, cas)]
        return torch.cat([xg, yg, zg, dxg, dyg, dzg, rg, *cgs], dim=-1)


class PreviousResidualDecoder(object):
    def __init__(self, code_size=7, **kwargs):
        super().__init__()
        self.code_size = code_size

    @staticmethod
    def decode_torch(box_encodings, anchors):
        """
        Args:
            box_encodings:  (B, N, 7 + ?) x, y, z, w, l, h, r, custom values
            anchors: (B, N, 7 + C) or (N, 7 + C) [x, y, z, dx, dy, dz, heading, ...]

        Returns:

        """
        xa, ya, za, dxa, dya, dza, ra, *cas = torch.split(anchors, 1, dim=-1)
        xt, yt, zt, wt, lt, ht, rt, *cts = torch.split(box_encodings, 1, dim=-1)

        diagonal = torch.sqrt(dxa ** 2 + dya ** 2)
        xg = xt * diagonal + xa
        yg = yt * diagonal + ya
        zg = zt * dza + za

        dxg = torch.exp(lt) * dxa
        dyg = torch.exp(wt) * dya
        dzg = torch.exp(ht) * dza
        rg = rt + ra

        cgs = [t + a for t, a in zip(cts, cas)]
        return torch.cat([xg, yg, zg, dxg, dyg, dzg, rg, *cgs], dim=-1)


class PreviousResidualRoIDecoder(object):
    def __init__(self, code_size=7, **kwargs):
        super().__init__()
        self.code_size = code_size

    @staticmethod
    def decode_torch(box_encodings, anchors):
        """
        Args:
            box_encodings:  (B, N, 7 + ?) x, y, z, w, l, h, r, custom values
            anchors: (B, N, 7 + C) or (N, 7 + C) [x, y, z, dx, dy, dz, heading, ...]

        Returns:

        """
        xa, ya, za, dxa, dya, dza, ra, *cas = torch.split(anchors, 1, dim=-1)
        xt, yt, zt, wt, lt, ht, rt, *cts = torch.split(box_encodings, 1, dim=-1)

        diagonal = torch.sqrt(dxa ** 2 + dya ** 2)
        xg = xt * diagonal + xa
        yg = yt * diagonal + ya
        zg = zt * dza + za

        dxg = torch.exp(lt) * dxa
        dyg = torch.exp(wt) * dya
        dzg = torch.exp(ht) * dza
        rg = ra - rt

        cgs = [t + a for t, a in zip(cts, cas)]
        return torch.cat([xg, yg, zg, dxg, dyg, dzg, rg, *cgs], dim=-1)