代码解析随笔
以下说的下采样之后的图像像素实际上是特征点
以下代码是计算下采样之后的图像像素在原始图像上的位置，同时和离散深度进行组合
这一步是因为下采样之后的图像想投影到ego car坐标系，需要先知道像素在原始图像上的位置，才能利用内外参转换。
def create_frustum(self):# make grid in image planeogfH, ogfW = self.data_aug_conf['final_dim']fH, fW = ogfH // self.downsample, ogfW // self.downsampleds = torch.arange(*self.grid_conf['dbound'], dtype=torch.float).view(-1, 1, 1).expand(-1, fH, fW)D, _, _ = ds.shapexs = torch.linspace(0, ogfW - 1, fW, dtype=torch.float).view(1, 1, fW).expand(D, fH, fW)ys = torch.linspace(0, ogfH - 1, fH, dtype=torch.float).view(1, fH, 1).expand(D, fH, fW)# D x H x W x 3frustum = torch.stack((xs, ys, ds), -1)return nn.Parameter(frustum, requires_grad=False)
以下代码是将下采样之后图像上的点通过内外参（还需要撤销增强变换）转换到ego car坐标系
def get_geometry(self, rots, trans, intrins, post_rots, post_trans):"""Determine the (x,y,z) locations (in the ego frame)of the points in the point cloud.Returns B x N x D x H/downsample x W/downsample x 3"""B, N, _ = trans.shape# undo post-transformation# B x N x D x H x W x 3points = self.frustum - post_trans.view(B, N, 1, 1, 1, 3)points = torch.inverse(post_rots).view(B, N, 1, 1, 1, 3, 3).matmul(points.unsqueeze(-1))# cam_to_egopoints = torch.cat((points[:, :, :, :, :, :2] * points[:, :, :, :, :, 2:3],points[:, :, :, :, :, 2:3]), 5)combine = rots.matmul(torch.inverse(intrins))points = combine.view(B, N, 1, 1, 1, 3, 3).matmul(points).squeeze(-1)points += trans.view(B, N, 1, 1, 1, 3)return points
def voxel_pooling(self, geom_feats, x):B, N, D, H, W, C = x.shapeNprime = B*N*D*H*W# flatten xx = x.reshape(Nprime, C)# 以下需要将连续的egocar坐标系下的坐标转为离散的bev下的坐标，z轴的一个网格高度是20米# 是下采样图像上每个像素点在离散的bev网格中的位置坐标# flatten indicesgeom_feats = ((geom_feats - (self.bx - self.dx/2.)) / self.dx).long()geom_feats = geom_feats.view(Nprime, 3)batch_ix = torch.cat([torch.full([Nprime//B, 1], ix,device=x.device, dtype=torch.long) for ix in range(B)])geom_feats = torch.cat((geom_feats, batch_ix), 1)#如果self.nx[2]=1，那么只会保留网格高度为1的物体，也就是实际高度20米以内的物体# filter out points that are outside boxkept = (geom_feats[:, 0] >= 0) & (geom_feats[:, 0] < self.nx[0])\& (geom_feats[:, 1] >= 0) & (geom_feats[:, 1] < self.nx[1])\& (geom_feats[:, 2] >= 0) & (geom_feats[:, 2] < self.nx[2])x = x[kept]geom_feats = geom_feats[kept]# 在同一个bev网络中的下采样图像像素所对应的rank是一样的# get tensors from the same voxel next to each otherranks = geom_feats[:, 0] * (self.nx[1] * self.nx[2] * B)\+ geom_feats[:, 1] * (self.nx[2] * B)\+ geom_feats[:, 2] * B\+ geom_feats[:, 3]sorts = ranks.argsort()x, geom_feats, ranks = x[sorts], geom_feats[sorts], ranks[sorts]# cumsum trickif not self.use_quickcumsum:x, geom_feats = cumsum_trick(x, geom_feats, ranks)else:x, geom_feats = QuickCumsum.apply(x, geom_feats, ranks)# griddify (B x C x Z x X x Y)final = torch.zeros((B, C, self.nx[2], self.nx[0], self.nx[1]), device=x.device)final[geom_feats[:, 3], :, geom_feats[:, 2], geom_feats[:, 0], geom_feats[:, 1]] = x# collapse Zfinal = torch.cat(final.unbind(dim=2), 1)return final
