PointNet代码解析

1. 如何对点云使用MLP？

#以分类网络为例，整体代码：
def get_model(point_cloud, is_training, bn_decay=None):
    """ Classification PointNet, input is BxNx3, output Bx40 """
    batch_size = point_cloud.get_shape()[0].value
    num_point = point_cloud.get_shape()[1].value
    end_points = {}
    with tf.variable_scope('transform_net1') as sc:
        transform = input_transform_net(point_cloud, is_training, bn_decay, K=3)
    point_cloud_transformed = tf.matmul(point_cloud, transform)
    input_image = tf.expand_dims(point_cloud_transformed, -1)
    net = tf_util.conv2d(input_image, 64, [1,3],
                         padding='VALID', stride=[1,1],
                         bn=True, is_training=is_training,
                         scope='conv1', bn_decay=bn_decay)
    net = tf_util.conv2d(net, 64, [1,1],
                         padding='VALID', stride=[1,1],
                         bn=True, is_training=is_training,
                         scope='conv2', bn_decay=bn_decay)
    with tf.variable_scope('transform_net2') as sc:
        transform = feature_transform_net(net, is_training, bn_decay, K=64)
    end_points['transform'] = transform
    net_transformed = tf.matmul(tf.squeeze(net, axis=[2]), transform)
    net_transformed = tf.expand_dims(net_transformed, [2])
    net = tf_util.conv2d(net_transformed, 64, [1,1],
                         padding='VALID', stride=[1,1],
                         bn=True, is_training=is_training,
                         scope='conv3', bn_decay=bn_decay)
    net = tf_util.conv2d(net, 128, [1,1],
                         padding='VALID', stride=[1,1],
                         bn=True, is_training=is_training,
                         scope='conv4', bn_decay=bn_decay)
    net = tf_util.conv2d(net, 1024, [1,1],
                         padding='VALID', stride=[1,1],
                         bn=True, is_training=is_training,
                         scope='conv5', bn_decay=bn_decay)
    # Symmetric function: max pooling
    net = tf_util.max_pool2d(net, [num_point,1],
                             padding='VALID', scope='maxpool')
    net = tf.reshape(net, [batch_size, -1])
    net = tf_util.fully_connected(net, 512, bn=True, is_training=is_training,
                                  scope='fc1', bn_decay=bn_decay)
    net = tf_util.dropout(net, keep_prob=0.7, is_training=is_training,
                          scope='dp1')
    net = tf_util.fully_connected(net, 256, bn=True, is_training=is_training,
                                  scope='fc2', bn_decay=bn_decay)
    net = tf_util.dropout(net, keep_prob=0.7, is_training=is_training,
                          scope='dp2')
    net = tf_util.fully_connected(net, 40, activation_fn=None, scope='fc3')
    return net, end_points

'''MLP的核心做法：'''

input_image = tf.expand_dims(point_cloud_transformed, -1)
net = tf_util.conv2d(input_image, 64, [1,3],
                         padding='VALID', stride=[1,1],
                         bn=True, is_training=is_training,
                         scope='conv1', bn_decay=bn_decay)
net = tf_util.conv2d(net, 64, [1,1],
                     padding='VALID', stride=[1,1],
                     bn=True, is_training=is_training,
                     scope='conv2', bn_decay=bn_decay)

这里input_image维度是 B×N×3×1 ，因此将点云看成是W和H分为N和3的2D图像，维度是 1。

然后直接基于这个“2D图像”做卷积，第一个卷积核size是 [1,3] ，正好对应的就是“2D图像”的一行，也就是一个点（三维坐标），输出通道数是64，因此输出张量维度应该是 B×N×1×64

第二个卷积核size是 [1,1] ， 1∗1 卷积只改变通道数，输出张量维度是 B×N×1×64

conv2d就是将卷积封装了一下，核心部分也就是调用tf.nn.conv2d，实现如下：

def conv2d(inputs,
           num_output_channels,
           kernel_size,
           scope,
           stride=[1, 1],
           padding='SAME',
           use_xavier=True,
           stddev=1e-3,
           weight_decay=0.0,
           activation_fn=tf.nn.relu,
           bn=False,
           bn_decay=None,
           is_training=None):
  """ 2D convolution with non-linear operation.
  Args:
    inputs: 4-D tensor variable BxHxWxC
    num_output_channels: int
    kernel_size: a list of 2 ints
    scope: string
    stride: a list of 2 ints
    padding: 'SAME' or 'VALID'
    use_xavier: bool, use xavier_initializer if true
    stddev: float, stddev for truncated_normal init
    weight_decay: float
    activation_fn: function
    bn: bool, whether to use batch norm
    bn_decay: float or float tensor variable in [0,1]
    is_training: bool Tensor variable
  Returns:
    Variable tensor
  """
  with tf.variable_scope(scope) as sc:
      kernel_h, kernel_w = kernel_size
      num_in_channels = inputs.get_shape()[-1].value
      kernel_shape = [kernel_h, kernel_w,
                      num_in_channels, num_output_channels]
      kernel = _variable_with_weight_decay('weights',
                                           shape=kernel_shape,
                                           use_xavier=use_xavier,
                                           stddev=stddev,
                                           wd=weight_decay)
      stride_h, stride_w = stride
      outputs = tf.nn.conv2d(inputs, kernel,
                             [1, stride_h, stride_w, 1],
                             padding=padding)
      biases = _variable_on_cpu('biases', [num_output_channels],
                                tf.constant_initializer(0.0))
      outputs = tf.nn.bias_add(outputs, biases)
      if bn:
        outputs = batch_norm_for_conv2d(outputs, is_training,
                                        bn_decay=bn_decay, scope='bn')
      if activation_fn is not None:
        outputs = activation_fn(outputs)
      return outputs

2. alignment network怎么做的？

这里以input_transform_net为例：

def input_transform_net(point_cloud, is_training, bn_decay=None, K=3):
    """ Input (XYZ) Transform Net, input is BxNx3 gray image
        Return:
            Transformation matrix of size 3xK """
    batch_size = point_cloud.get_shape()[0].value
    num_point = point_cloud.get_shape()[1].value
    input_image = tf.expand_dims(point_cloud, -1)
    net = tf_util.conv2d(input_image, 64, [1,3],
                         padding='VALID', stride=[1,1],
                         bn=True, is_training=is_training,
                         scope='tconv1', bn_decay=bn_decay)
    net = tf_util.conv2d(net, 128, [1,1],
                         padding='VALID', stride=[1,1],
                         bn=True, is_training=is_training,
                         scope='tconv2', bn_decay=bn_decay)
    net = tf_util.conv2d(net, 1024, [1,1],
                         padding='VALID', stride=[1,1],
                         bn=True, is_training=is_training,
                         scope='tconv3', bn_decay=bn_decay)
    net = tf_util.max_pool2d(net, [num_point,1],
                             padding='VALID', scope='tmaxpool')
    net = tf.reshape(net, [batch_size, -1])
    net = tf_util.fully_connected(net, 512, bn=True, is_training=is_training,
                                  scope='tfc1', bn_decay=bn_decay)
    net = tf_util.fully_connected(net, 256, bn=True, is_training=is_training,
                                  scope='tfc2', bn_decay=bn_decay)
    with tf.variable_scope('transform_XYZ') as sc:
        assert(K==3)
        weights = tf.get_variable('weights', [256, 3*K],
                                  initializer=tf.constant_initializer(0.0),
                                  dtype=tf.float32)
        biases = tf.get_variable('biases', [3*K],
                                 initializer=tf.constant_initializer(0.0),
                                 dtype=tf.float32)
        biases += tf.constant([1,0,0,0,1,0,0,0,1], dtype=tf.float32)
        transform = tf.matmul(net, weights)
        transform = tf.nn.bias_add(transform, biases)
    transform = tf.reshape(transform, [batch_size, 3, K])
    return transform

实际上，前半部分就是通过卷积和max_pooling对batch内各个点云提取global feature，再将global feature降到 3×K 维度，并reshape成 3×3 ，得到transform matrix

通过数据增强丰富训练数据集，网络确实应该学习到有效的transform matrix，用来实现transformation invariance

3. 对称函数如何实现来提取global feature的？

max_pooling，这个在论文的图中还是代码中都有体现，代码甚至直接用注释注明了

4. loss?

def get_loss(pred, label, end_points, reg_weight=0.001):
    """ pred: B*NUM_CLASSES,
        label: B, """
    loss = tf.nn.sparse_softmax_cross_entropy_with_logits(logits=pred, labels=label)
    classify_loss = tf.reduce_mean(loss)
    tf.summary.scalar('classify loss', classify_loss)
    # Enforce the transformation as orthogonal matrix
    transform = end_points['transform'] # BxKxK
    K = transform.get_shape()[1].value
    mat_diff = tf.matmul(transform, tf.transpose(transform, perm=[0,2,1]))
    mat_diff -= tf.constant(np.eye(K), dtype=tf.float32)
    mat_diff_loss = tf.nn.l2_loss(mat_diff) 
    tf.summary.scalar('mat loss', mat_diff_loss)
    return classify_loss + mat_diff_loss * reg_weight

无论是分类还是分割，本质上都还是分类任务，只是粒度不同罢了。

因此loss一定有有监督分类任务中常用的交叉熵loss

另外loss还有之前alignment network中提到的约束loss，也就是上面的mat_diff_loss

---

PointNet的代码实际上仅由两部分组成，就是T-Net和一个Encoder-Decoder结构：

class T_Net(nn.Module):
    def __init__(self):
        super(T_Net, self).__init__()
        # 这里需要注意的是上文提到的MLP均由卷积结构完成
        # 比如说将3维映射到64维，其利用64个1x3的卷积核
        self.conv1 = torch.nn.Conv1d(3, 64, 1)
        self.conv2 = torch.nn.Conv1d(64, 128, 1)
        self.conv3 = torch.nn.Conv1d(128, 1024, 1)
        self.fc1 = nn.Linear(1024, 512)
        self.fc2 = nn.Linear(512, 256)
        self.fc3 = nn.Linear(256, 9)
        self.relu = nn.ReLU()

        self.bn1 = nn.BatchNorm1d(64)
        self.bn2 = nn.BatchNorm1d(128)
        self.bn3 = nn.BatchNorm1d(1024)
        self.bn4 = nn.BatchNorm1d(512)
        self.bn5 = nn.BatchNorm1d(256)


    def forward(self, x):
        batchsize = x.size()[0]
        x = F.relu(self.bn1(self.conv1(x)))
        x = F.relu(self.bn2(self.conv2(x)))
        x = F.relu(self.bn3(self.conv3(x)))
        x = torch.max(x, 2, keepdim=True)[0]
        x = x.view(-1, 1024)

        x = F.relu(self.bn4(self.fc1(x)))
        x = F.relu(self.bn5(self.fc2(x)))
        x = self.fc3(x)

        iden = Variable(torch.from_numpy(np.array([1,0,0,0,1,0,0,0,1]).astype(np.float32))).view(1,9).repeat(batchsize,1)
        if x.is_cuda:
            iden = iden.cuda()
        x = x + iden
        x = x.view(-1, 3, 3) # 输出为Batch*3*3的张量
        return x

再看PointNet的主体：

class PointNetEncoder(nn.Module):
    def __init__(self, global_feat = True):
        super(PointNetEncoder, self).__init__()
        self.stn = T_Net()
        self.conv1 = torch.nn.Conv1d(3, 64, 1)
        self.conv2 = torch.nn.Conv1d(64, 128, 1)
        self.conv3 = torch.nn.Conv1d(128, 1024, 1)
        self.bn1 = nn.BatchNorm1d(64)
        self.bn2 = nn.BatchNorm1d(128)
        self.bn3 = nn.BatchNorm1d(1024)
        self.global_feat = global_feat
    def forward(self, x):
   	    '''生成全局特征'''
        n_pts = x.size()[2]
        trans = self.stn(x)
        x = x.transpose(2,1)
        x = torch.bmm(x, trans) # batch matrix multiply 即乘以T-Net的结果
        x = x.transpose(2,1)
        x = self.conv1(x)

        x = F.relu(self.bn1(x))
        pointfeat = x
        x_skip = self.conv2(x)

        x = F.relu(self.bn2(x_skip))
        x = self.bn3(self.conv3(x))
        x = torch.max(x, 2, keepdim=True)[0]
        x = x.view(-1, 1024)
        if self.global_feat:
            return x, trans
        else:
            x = x.view(-1, 1024, 1).repeat(1, 1, n_pts)
            return torch.cat([x, pointfeat], 1), trans

class PointNetCls(nn.Module):
    def __init__(self, k = 2):
        super(PointNetCls, self).__init__()
        self.k = k
        self.feat = PointNetEncoder(global_feat=False)
        self.conv1 = torch.nn.Conv1d(1088, 512, 1)
        self.conv2 = torch.nn.Conv1d(512, 256, 1)
        self.conv3 = torch.nn.Conv1d(256, 128, 1)
        self.conv4 = torch.nn.Conv1d(128, self.k, 1)
        self.bn1 = nn.BatchNorm1d(512)
        self.bn2 = nn.BatchNorm1d(256)
        self.bn3 = nn.BatchNorm1d(128)

    def forward(self, x):
    	'''分类网络'''
        batchsize = x.size()[0]
        n_pts = x.size()[2]
        x, trans = self.feat(x)
        x = F.relu(self.bn1(self.conv1(x)))
        x = F.relu(self.bn2(self.conv2(x)))
        x = F.relu(self.bn3(self.conv3(x)))
        x = self.conv4(x)
        x = x.transpose(2,1).contiguous()
        x = F.log_softmax(x.view(-1,self.k), dim=-1)
        x = x.view(batchsize, n_pts, self.k)
        return x


class PointNetPartSeg(nn.Module):
    def __init__(self,num_class):
        super(PointNetPartSeg, self).__init__()
        self.k = num_class
        self.feat = PointNetEncoder(global_feat=False)
        self.conv1 = torch.nn.Conv1d(1088, 512, 1)
        self.conv2 = torch.nn.Conv1d(512, 256, 1)
        self.conv3 = torch.nn.Conv1d(256, 128, 1)
        self.conv4 = torch.nn.Conv1d(128, self.k, 1)
        self.bn1 = nn.BatchNorm1d(512)
        self.bn1_1 = nn.BatchNorm1d(1024)
        self.bn2 = nn.BatchNorm1d(256)
        self.bn3 = nn.BatchNorm1d(128)

    def forward(self, x):
        '''分割网络'''
        batchsize = x.size()[0]
        n_pts = x.size()[2]
        x, trans = self.feat(x)
        x = F.relu(self.bn1(self.conv1(x)))
        x = F.relu(self.bn2(self.conv2(x)))
        x = F.relu(self.bn3(self.conv3(x)))
        x = self.conv4(x)
        x = x.transpose(2,1).contiguous()
        x = F.log_softmax(x.view(-1,self.k), dim=-1)
        x = x.view(batchsize, n_pts, self.k)
        return x, trans