add bipointnet. (#1727)

example/BiPointNet/README.md

+# BiPointNet
+
+## 1. 简介
+本示例介绍了一种用于点云模型 (PointNet) 的二值化方法（BiPointNet）。BiPointNet 通过引入熵最大化聚合（Entropy-Maximizing Aggregation）来调整聚合前的分布，以获得最大信息熵，并引入分层尺度因子（Layer-wise Scale Recovery）有效恢复特征表达能力，是一种简单而高效的二值化点云模型的方法。
+
+## 2. Benchmark
+
+| 模型         | Accuracy | 权重下载 |
+| ------------- | --------- | --------- |
+| PointNet      | 89.83     | [PointNet.pdparams](https://bj.bcebos.com/v1/paddle-slim-models/PointNet.pdparams) |
+| BiPointNet    | 85.86     | [BiPointNet.pdparams](https://bj.bcebos.com/v1/paddle-slim-models/BiPointNet.pdparams) |
+
+
+## 3. BiPointNet 的训练及测试
+BiPointNet 整体结构如图所示，详情见论文 [BIPOINTNET: BINARY NEURAL NETWORK FOR POINT CLOUDS](https://arxiv.org/abs/2010.05501)
+
+![arch](arch.png)
+
+### 3.1 准备环境
+- PaddlePaddle >= 2.4 （可从[Paddle官网](https://www.paddlepaddle.org.cn/install/quick?docurl=/documentation/docs/zh/install/pip/linux-pip.html)下载安装）
+
+安装 paddlepaddle：
+```shell
+# CPU
+pip install paddlepaddle==2.4.2
+# GPU 以Ubuntu、CUDA 11.2为例
+python -m pip install paddlepaddle-gpu==2.4.2.post112 -f https://www.paddlepaddle.org.cn/whl/linux/mkl/avx/stable.html
+```
+
+### 3.2 准备数据集
+
+本示例在 [ModelNet40](https://modelnet.cs.princeton.edu) 数据集上进行了分类实验。
+
+```shell
+wget https://shapenet.cs.stanford.edu/media/modelnet40_normal_resampled.zip --no-check-certificate
+unzip modelnet40_normal_resampled.zip
+```
+
+### 3.3 启动训练
+
+- 训练基准模型 PointNet
+```
+export CUDA_VISIBLE_DEVICES=0
+python train.py --save_dir 'pointnet'
+```
+
+- 训练 BiPointNet
+```
+export CUDA_VISIBLE_DEVICES=0
+python train.py --save_dir 'bipointnet' --binary
+```
+
+### 3.4 验证精度
+
+- 测试基准模型 PointNet
+```shell
+export CUDA_VISIBLE_DEVICES=0
+python test.py --model_path 'PointNet.pdparams'
+```
+
+- 测试 BiPointNet
+```shell
+export CUDA_VISIBLE_DEVICES=0
+python train.py --model_path 'BiPointNet.pdparams' --binary
+```
+
+## 致谢
+感谢 [Macaronlin](https://github.com/Macaronlin) 贡献 BiPointNet。

example/BiPointNet/basic.py

+# Copyright (c) 2023 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import paddle
+from paddle import nn
+from paddle.autograd import PyLayer
+from paddle.nn import functional as F
+from paddle.nn.layer import Conv1D
+from paddle.nn.layer.common import Linear
+
+
+class BinaryQuantizer(PyLayer):
+    @staticmethod
+    def forward(ctx, input):
+        ctx.save_for_backward(input)
+        out = paddle.sign(input)
+        return out
+
+    @staticmethod
+    def backward(ctx, grad_output):
+        input = ctx.saved_tensor()[0]
+        grad_input = grad_output
+        grad_input[input >= 1] = 0
+        grad_input[input <= -1] = 0
+        return grad_input.clone()
+
+
+class BiLinear(Linear):
+    def __init__(self,
+                 in_features,
+                 out_features,
+                 weight_attr=None,
+                 bias_attr=None,
+                 name=None):
+        super(BiLinear, self).__init__(
+            in_features,
+            out_features,
+            weight_attr=weight_attr,
+            bias_attr=bias_attr,
+            name=name)
+
+        self.scale_weight_init = False
+        self.scale_weight = paddle.create_parameter(shape=[1], dtype='float32')
+
+    def forward(self, input):
+        ba = input
+
+        bw = self.weight
+        bw = bw - bw.mean()
+
+        if self.scale_weight_init == False:
+            scale_weight = F.linear(ba, bw).std() / F.linear(
+                paddle.sign(ba), paddle.sign(bw)).std()
+            if paddle.isnan(scale_weight):
+                scale_weight = bw.std() / paddle.sign(bw).std()
+            self.scale_weight.set_value(scale_weight)
+            self.scale_weight_init = True
+
+        ba = BinaryQuantizer.apply(ba)
+        bw = BinaryQuantizer.apply(bw)
+        bw = bw * self.scale_weight
+
+        out = F.linear(x=ba, weight=bw, bias=self.bias, name=self.name)
+        return out
+
+
+class BiConv1D(Conv1D):
+    def __init__(self,
+                 in_channels,
+                 out_channels,
+                 kernel_size,
+                 stride=1,
+                 padding=0,
+                 dilation=1,
+                 groups=1,
+                 padding_mode='zeros',
+                 weight_attr=None,
+                 bias_attr=None,
+                 data_format="NCL"):
+        super(BiConv1D, self).__init__(
+            in_channels, out_channels, kernel_size, stride, padding, dilation,
+            groups, padding_mode, weight_attr, bias_attr, data_format)
+        self.scale_weight_init = False
+        self.scale_weight = paddle.create_parameter(shape=[1], dtype='float32')
+
+    def forward(self, input):
+        ba = input
+
+        bw = self.weight
+        bw = bw - bw.mean()
+
+        padding = 0
+        if self._padding_mode != "zeros":
+            ba = F.pad(
+                ba,
+                self._reversed_padding_repeated_twice,
+                mode=self._padding_mode,
+                data_format=self._data_format)
+        else:
+            padding = self._padding
+
+        if self.scale_weight_init == False:
+            scale_weight = F.conv1d(ba, bw, bias=self.bias, padding=padding, stride=self._stride, dilation=self._dilation, groups=self._groups, data_format=self._data_format).std() / \
+                F.conv1d(paddle.sign(ba), paddle.sign(bw), bias=self.bias, padding=padding, stride=self._stride, dilation=self._dilation, groups=self._groups, data_format=self._data_format).std()
+            if paddle.isnan(scale_weight):
+                scale_weight = bw.std() / paddle.sign(bw).std()
+
+            self.scale_weight.set_value(scale_weight)
+            self.scale_weight_init = True
+
+        ba = BinaryQuantizer.apply(ba)
+        bw = BinaryQuantizer.apply(bw)
+        bw = bw * self.scale_weight
+
+        return F.conv1d(
+            ba,
+            bw,
+            bias=self.bias,
+            padding=padding,
+            stride=self._stride,
+            dilation=self._dilation,
+            groups=self._groups,
+            data_format=self._data_format)
+
+
+def _to_bi_function(model, fp_layers=[]):
+    for name, layer in model.named_children():
+        if id(layer) in fp_layers:
+            continue
+        if isinstance(layer, Linear):
+            new_layer = BiLinear(layer.weight.shape[0], layer.weight.shape[1],
+                                 layer._weight_attr, layer._bias_attr,
+                                 layer.name)
+            new_layer.weight = layer.weight
+            new_layer.bias = layer.bias
+            model._sub_layers[name] = new_layer
+        elif isinstance(layer, Conv1D):
+            new_layer = BiConv1D(layer._in_channels, layer._out_channels,
+                                 layer._kernel_size, layer._stride,
+                                 layer._padding, layer._dilation, layer._groups,
+                                 layer._padding_mode, layer._param_attr,
+                                 layer._bias_attr, layer._data_format)
+            new_layer.weight = layer.weight
+            new_layer.bias = layer.bias
+            model._sub_layers[name] = new_layer
+        elif isinstance(layer, nn.ReLU):
+            model._sub_layers[name] = nn.Hardtanh()
+        else:
+            model._sub_layers[name] = _to_bi_function(layer, fp_layers)
+    return model

example/BiPointNet/data.py

+# Copyright (c) 2023 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import os
+import pickle
+import warnings
+import numpy as np
+import paddle
+from paddle.io import Dataset
+from tqdm import tqdm
+
+warnings.filterwarnings("ignore")
+
+
+def normalize_point_cloud(pc):
+    centroid = np.mean(pc, axis=0)
+    pc = pc - centroid
+    m = np.max(np.sqrt(np.sum(pc**2, axis=1)))
+    pc = pc / m
+    return pc
+
+
+def random_point_dropout(pc, max_dropout_ratio=0.875):
+    dropout_ratio = np.random.random() * max_dropout_ratio  # 0~0.875
+    drop_idx = np.where(np.random.random((pc.shape[0])) <= dropout_ratio)[0]
+    if len(drop_idx) > 0:
+        pc[drop_idx, :] = pc[0, :]  # set to the first point
+    return pc
+
+
+def random_scale_point_cloud(data, scale_low=0.8, scale_high=1.25):
+    scales = np.random.uniform(scale_low, scale_high)
+    data *= scales
+    return data
+
+
+def shift_point_cloud(data, shift_range=0.1):
+    shifts = np.random.uniform(-shift_range, shift_range, (3))
+    data += shifts
+    return data
+
+
+def jitter_point_cloud(data: np.ndarray, sigma: float=0.02, clip: float=0.05):
+    assert clip > 0
+    jittered_data = np.clip(sigma * np.random.randn(*data.shape), -clip, clip)
+    data = data + jittered_data
+    return data
+
+
+def random_rotate_point_cloud(data: np.ndarray):
+    angle = np.random.uniform() * 2 * np.pi
+    cosval = np.cos(angle)
+    sinval = np.sin(angle)
+    rotation_matrix = np.array(
+        [[cosval, sinval, 0], [-sinval, cosval, 0], [0, 0, 1]],
+        dtype=data.dtype)
+    data = data @ rotation_matrix
+    return data
+
+
+def farthest_point_sample(point, npoint):
+    """
+    Input:
+        xyz: pointcloud data, [N, D]
+        npoint: number of samples
+    Return:
+        centroids: sampled pointcloud index, [npoint, D]
+    """
+    N, D = point.shape
+    xyz = point[:, :3]
+    centroids = np.zeros((npoint, ))
+    distance = np.ones((N, )) * 1e10
+    farthest = np.random.randint(0, N)
+    for i in range(npoint):
+        centroids[i] = farthest
+        centroid = xyz[farthest, :]
+        dist = np.sum((xyz - centroid)**2, -1)
+        mask = dist < distance
+        distance[mask] = dist[mask]
+        farthest = np.argmax(distance, -1)
+    point = point[centroids.astype(np.int32)]
+    return point
+
+
+class ModelNetDataset(Dataset):
+    def __init__(
+            self,
+            root,
+            num_point,
+            use_uniform_sample=False,
+            use_normals=False,
+            num_category=40,
+            split="train",
+            process_data=False, ):
+        self.root = root
+        self.npoints = num_point
+        self.split = split
+        self.process_data = process_data
+        self.uniform = use_uniform_sample
+        self.use_normals = use_normals
+        self.num_category = num_category
+
+        if self.num_category == 10:
+            self.catfile = os.path.join(self.root, "modelnet10_shape_names.txt")
+        else:
+            self.catfile = os.path.join(self.root, "modelnet40_shape_names.txt")
+
+        self.cat = [line.rstrip() for line in open(self.catfile)]
+        self.classes = dict(zip(self.cat, range(len(self.cat))))
+
+        shape_ids = {}
+        if self.num_category == 10:
+            shape_ids["train"] = [
+                line.rstrip() for line in
+                open(os.path.join(self.root, "modelnet10_train.txt"))
+            ]
+            shape_ids["test"] = [
+                line.rstrip() for line in
+                open(os.path.join(self.root, "modelnet10_test.txt"))
+            ]
+        else:
+            shape_ids["train"] = [
+                line.rstrip() for line in
+                open(os.path.join(self.root, "modelnet40_train.txt"))
+            ]
+            shape_ids["test"] = [
+                line.rstrip() for line in
+                open(os.path.join(self.root, "modelnet40_test.txt"))
+            ]
+
+        assert split == "train" or split == "test"
+        shape_names = ["_".join(x.split("_")[0:-1]) for x in shape_ids[split]]
+        self.datapath = [(shape_names[i], os.path.join(
+            self.root, shape_names[i], shape_ids[split][i]) + ".txt", )
+                         for i in range(len(shape_ids[split]))]
+        print("The size of %s data is %d" % (split, len(self.datapath)))
+
+        if self.uniform:
+            self.save_path = os.path.join(
+                root,
+                "modelnet%d_%s_%dpts_fps.dat" % (self.num_category, split,
+                                                 self.npoints), )
+        else:
+            self.save_path = os.path.join(
+                root,
+                "modelnet%d_%s_%dpts.dat" % (self.num_category, split,
+                                             self.npoints), )
+
+        if self.process_data:
+            if not os.path.exists(self.save_path):
+                print("Processing data %s (only running in the first time)..." %
+                      self.save_path)
+                self.list_of_points = [None] * len(self.datapath)
+                self.list_of_labels = [None] * len(self.datapath)
+
+                for index in tqdm(
+                        range(len(self.datapath)), total=len(self.datapath)):
+                    fn = self.datapath[index]
+                    cls = self.classes[self.datapath[index][0]]
+                    cls = np.array([cls]).astype(np.int32)
+                    point_set = np.loadtxt(
+                        fn[1], delimiter=",").astype(np.float32)
+
+                    if self.uniform:
+                        point_set = farthest_point_sample(
+                            point_set, self.npoints)
+                    else:
+                        point_set = point_set[0:self.npoints, :]
+
+                    self.list_of_points[index] = point_set
+                    self.list_of_labels[index] = cls
+
+                with open(self.save_path, "wb") as f:
+                    pickle.dump([self.list_of_points, self.list_of_labels], f)
+            else:
+                print("Load processed data from %s..." % self.save_path)
+                with open(self.save_path, "rb") as f:
+                    self.list_of_points, self.list_of_labels = pickle.load(f)
+
+    def __len__(self):
+        return len(self.datapath)
+
+    def _get_item(self, index):
+        if self.process_data:
+            point_set, label = self.list_of_points[index], self.list_of_labels[
+                index]
+        else:
+            fn = self.datapath[index]
+            cls = self.classes[self.datapath[index][0]]
+            label = np.array([cls]).astype(np.int32)
+            point_set = np.loadtxt(fn[1], delimiter=",").astype(np.float32)
+
+            if self.uniform:
+                point_set = farthest_point_sample(point_set, self.npoints)
+            else:
+                point_set = point_set[0:self.npoints, :]
+
+        point_set[:, 0:3] = normalize_point_cloud(point_set[:, 0:3])
+        if not self.use_normals:
+            point_set = point_set[:, 0:3]
+
+        if self.split == "train":
+            # point_set[:, 0:3] = random_rotate_point_cloud(point_set[:, 0:3])
+            point_set[:, 0:3] = jitter_point_cloud(point_set[:, 0:3])
+            # point_set[:, 0:3] = random_point_dropout(point_set[:, 0:3])
+            point_set[:, 0:3] = random_scale_point_cloud(point_set[:, 0:3])
+            point_set[:, 0:3] = shift_point_cloud(point_set[:, 0:3])
+
+        return paddle.to_tensor(point_set, dtype=paddle.float32), int(label[0])
+
+    def __getitem__(self, index):
+        return self._get_item(index)

example/BiPointNet/model.py

+# Copyright (c) 2023 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import paddle
+import paddle.nn as nn
+import paddle.nn.functional as F
+
+offset_map = {1024: -3.2041, 2048: -3.4025, 4096: -3.5836}
+
+
+class TNet(nn.Layer):
+    def __init__(self, k=64, binary=False):
+        super(TNet, self).__init__()
+        self.conv1 = nn.Conv1D(k, 64, 1)
+        self.conv2 = nn.Conv1D(64, 128, 1)
+        self.conv3 = nn.Conv1D(128, 1024, 1)
+        self.fc1 = nn.Linear(1024, 512)
+        self.fc2 = nn.Linear(512, 256)
+        self.fc3 = nn.Linear(256, k * k)
+        self.act_function = 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)
+
+        self.k = k
+        self.binary = binary
+        self.iden = paddle.eye(self.k, self.k, dtype=paddle.float32)
+
+    def forward(self, x):
+        B, D, N = x.shape
+
+        x = self.act_function(self.bn1(self.conv1(x)))
+        x = self.act_function(self.bn2(self.conv2(x)))
+
+        if self.binary:
+            x = self.bn3(self.conv3(x))
+            x = paddle.max(x, 2, keepdim=True) + offset_map[N]
+        else:
+            x = self.act_function(self.bn3(self.conv3(x)))
+            x = paddle.max(x, 2, keepdim=True)
+        x = x.reshape((-1, 1024))
+
+        x = self.act_function(self.bn4(self.fc1(x)))
+        x = self.act_function(self.bn5(self.fc2(x)))
+        x = self.fc3(x)
+
+        x = x.reshape((-1, self.k, self.k)) + self.iden
+        return x
+
+
+class PointNetEncoder(nn.Layer):
+    def __init__(self,
+                 global_feat=True,
+                 input_transform=True,
+                 feature_transform=False,
+                 channel=3,
+                 binary=False):
+        super(PointNetEncoder, self).__init__()
+
+        self.global_feat = global_feat
+        if input_transform:
+            self.input_transfrom = TNet(k=channel)
+        else:
+            self.input_transfrom = lambda x: paddle.eye(
+                channel, channel, dtype=paddle.float32
+            )
+
+        self.conv1 = nn.Conv1D(channel, 64, 1)
+        self.conv2 = nn.Conv1D(64, 128, 1)
+        self.conv3 = nn.Conv1D(128, 1024, 1)
+
+        self.bn1 = nn.BatchNorm1D(64)
+        self.bn2 = nn.BatchNorm1D(128)
+        self.bn3 = nn.BatchNorm1D(1024)
+
+        if feature_transform:
+            self.feature_transform = TNet(k=64)
+        else:
+            self.feature_transform = lambda x: paddle.eye(64, 64, dtype=paddle.float32)
+
+        self.act_function = nn.ReLU()
+        self.binary = binary
+
+    def forward(self, x):
+        x = paddle.transpose(x, (0, 2, 1))
+        B, D, N = x.shape
+        trans_input = self.input_transfrom(x)
+        x = paddle.transpose(x, (0, 2, 1))
+        if D > 3:
+            feature = x[:, :, 3:]
+            x = x[:, :, :3]
+        x = paddle.bmm(x, trans_input)
+        if D > 3:
+            x = paddle.concat([x, feature], axis=2)
+        x = paddle.transpose(x, (0, 2, 1))
+        x = self.act_function(self.bn1(self.conv1(x)))
+
+        trans_feat = self.feature_transform(x)
+        x = paddle.transpose(x, (0, 2, 1))
+        x = paddle.bmm(x, trans_feat)
+        x = paddle.transpose(x, (0, 2, 1))
+
+        pointfeat = x
+        x = self.act_function(self.bn2(self.conv2(x)))
+        x = self.bn3(self.conv3(x))
+
+        if self.binary:
+            x = paddle.max(x, 2, keepdim=True) + offset_map[N]
+        else:
+            x = paddle.max(x, 2, keepdim=True)
+
+        x = x.reshape((-1, 1024))
+
+        if self.global_feat:
+            return x, trans_input, trans_feat
+        else:
+            x = x.reshape((-1, 1024, 1)).repeat(1, 1, N)
+            return paddle.cat([x, pointfeat], 1), trans_input, trans_feat
+
+
+class PointNetClassifier(nn.Layer):
+    def __init__(self, k=40, normal_channel=False, binary=False):
+        super(PointNetClassifier, self).__init__()
+        if normal_channel:
+            channel = 6
+        else:
+            channel = 3
+        self.feat = PointNetEncoder(
+            global_feat=True,
+            input_transform=True,
+            feature_transform=True,
+            channel=channel,
+            binary=binary, )
+        self.fc1 = nn.Linear(1024, 512)
+        self.fc2 = nn.Linear(512, 256)
+        self.fc3 = nn.Linear(256, k)
+        self.dropout = nn.Dropout(p=0.4)
+        self.bn1 = nn.BatchNorm1D(512)
+        self.bn2 = nn.BatchNorm1D(256)
+        self.act_function = nn.ReLU()
+        self.binary = binary
+
+    def forward(self, x):
+        x, trans_input, trans_feat = self.feat(x)
+        x = self.act_function(self.bn1(self.fc1(x)))
+        x = self.act_function(self.bn2(self.fc2(x)))
+        if not self.binary:
+            x = self.dropout(x)
+        x = self.fc3(x)
+        return x, trans_input, trans_feat
+
+
+class CrossEntropyMatrixRegularization(nn.Layer):
+    def __init__(self, mat_diff_loss_scale=1e-3):
+        super(CrossEntropyMatrixRegularization, self).__init__()
+        self.mat_diff_loss_scale = mat_diff_loss_scale
+
+    def forward(self, pred, target, trans_feat=None):
+        loss = F.cross_entropy(pred, target)
+
+        if trans_feat is None:
+            mat_diff_loss = 0
+        else:
+            mat_diff_loss = self.feature_transform_reguliarzer(trans_feat)
+
+        total_loss = loss + mat_diff_loss * self.mat_diff_loss_scale
+        return total_loss
+
+    def feature_transform_reguliarzer(self, trans):
+        d = trans.shape[1]
+        I = paddle.eye(d)
+        loss = paddle.mean(
+            paddle.norm(
+                paddle.bmm(trans, paddle.transpose(trans, (0, 2, 1))) - I,
+                axis=(1, 2)))
+        return loss

example/BiPointNet/test.py

+# Copyright (c) 2023 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import argparse
+import paddle
+from paddle.io import DataLoader
+from paddle.metric import Accuracy
+from data import ModelNetDataset
+from model import PointNetClassifier
+
+
+def parse_args():
+    parser = argparse.ArgumentParser("Test")
+    parser.add_argument("--batch_size", type=int, default=32, help="batch size")
+    parser.add_argument(
+        "--num_point", type=int, default=1024, help="point number")
+    parser.add_argument(
+        "--num_workers", type=int, default=32, help="num wrokers")
+    parser.add_argument("--log_freq", type=int, default=10)
+    parser.add_argument(
+        "--model_path", type=str, default="./BiPointNet.pdparams")
+    parser.add_argument(
+        "--data_dir",
+        type=str,
+        default="./modelnet40_normal_resampled", )
+    parser.add_argument(
+        "--binary",
+        action='store_true',
+        help="whehter to build binary pointnet")
+    return parser.parse_args()
+
+
+def test(args):
+
+    test_data = ModelNetDataset(
+        args.data_dir, split="test", num_point=args.num_point)
+    test_loader = DataLoader(
+        test_data,
+        shuffle=False,
+        num_workers=args.num_workers,
+        batch_size=args.batch_size, )
+
+    model = PointNetClassifier(binary=args.binary)
+    if args.binary:
+        import basic
+        fp_layers = [
+            id(model.feat.input_transfrom.conv1),
+            id(model.feat.conv1),
+            id(model.fc3)
+        ]
+        model = basic._to_bi_function(model, fp_layers=fp_layers)
+
+        def func(model):
+            if hasattr(model, "scale_weight_init"):
+                model.scale_weight_init = True
+
+        model.apply(func)
+
+    model_state_dict = paddle.load(args.model_path)
+    model.set_state_dict(model_state_dict)
+
+    metrics = Accuracy()
+    metrics.reset()
+    model.eval()
+    for iter, data in enumerate(test_loader):
+        x, y = data
+        pred, _, _ = model(x)
+
+        correct = metrics.compute(pred, y)
+        metrics.update(correct)
+        if iter % args.log_freq == 0:
+            print("Eval iter:", iter)
+    test_acc = metrics.accumulate()
+    print("Test Accuracy: {}".format(test_acc))
+
+
+if __name__ == "__main__":
+    args = parse_args()
+    print(args)
+    test(args)

example/BiPointNet/train.py

+# Copyright (c) 2023 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import os
+import argparse
+import paddle
+from paddle.io import DataLoader
+from paddle.metric import Accuracy
+from paddle.optimizer import Adam
+from paddle.optimizer.lr import CosineAnnealingDecay
+from data import ModelNetDataset
+from model import CrossEntropyMatrixRegularization, PointNetClassifier
+
+
+def parse_args():
+    parser = argparse.ArgumentParser("Train")
+    parser.add_argument(
+        "--batch_size", type=int, default=32, help="batch size in training")
+    parser.add_argument(
+        "--learning_rate",
+        type=float,
+        default=1e-3,
+        help="learning rate in training")
+    parser.add_argument(
+        "--num_point", type=int, default=1024, help="point number")
+    parser.add_argument(
+        "--max_epochs", type=int, default=200, help="max epochs")
+    parser.add_argument(
+        "--num_workers", type=int, default=32, help="num wrokers")
+    parser.add_argument(
+        "--weight_decay", type=float, default=1e-4, help="weight decay")
+    parser.add_argument("--log_freq", type=int, default=50)
+    parser.add_argument(
+        "--pretrained",
+        type=str,
+        default='pointnet.pdparams',
+        help='pretrained model path')
+    parser.add_argument(
+        "--save_dir", type=str, default='./save_model', help='save model path')
+    parser.add_argument(
+        "--data_dir",
+        type=str,
+        default="./modelnet40_normal_resampled",
+        help='dataset dir')
+    parser.add_argument(
+        "--binary",
+        action='store_true',
+        help="whehter to build binary pointnet")
+    return parser.parse_args()
+
+
+def train(args):
+    train_data = ModelNetDataset(
+        args.data_dir, split="train", num_point=args.num_point)
+    test_data = ModelNetDataset(
+        args.data_dir, split="test", num_point=args.num_point)
+    train_loader = DataLoader(
+        train_data,
+        shuffle=True,
+        num_workers=args.num_workers,
+        batch_size=args.batch_size, )
+    test_loader = DataLoader(
+        test_data,
+        shuffle=False,
+        num_workers=args.num_workers,
+        batch_size=args.batch_size, )
+
+    model = PointNetClassifier(binary=args.binary)
+    if args.binary:
+        import basic
+        fp_layers = [
+            id(model.feat.input_transfrom.conv1),
+            id(model.feat.conv1),
+            id(model.fc3)
+        ]
+        model = basic._to_bi_function(model, fp_layers=fp_layers)
+        print(model)
+
+    model_state_dict = paddle.load(path=args.pretrained)
+    model.set_state_dict(model_state_dict)
+
+    scheduler = CosineAnnealingDecay(
+        learning_rate=args.learning_rate,
+        T_max=args.max_epochs, )
+
+    optimizer = Adam(
+        learning_rate=scheduler,
+        parameters=model.parameters(),
+        weight_decay=args.weight_decay, )
+    loss_fn = CrossEntropyMatrixRegularization()
+    metrics = Accuracy()
+
+    best_test_acc = 0
+    for epoch in range(args.max_epochs):
+        metrics.reset()
+        model.train()
+        for batch_id, data in enumerate(train_loader):
+
+            x, y = data
+            pred, trans_input, trans_feat = model(x)
+
+            loss = loss_fn(pred, y, trans_feat)
+
+            correct = metrics.compute(pred, y)
+            metrics.update(correct)
+            loss.backward()
+            optimizer.step()
+            optimizer.clear_grad()
+
+            if (batch_id + 1) % args.log_freq == 0:
+                print("Epoch: {}, Batch ID: {}, Loss: {}, ACC: {}".format(
+                    epoch, batch_id + 1, loss.item(), metrics.accumulate()))
+
+        scheduler.step()
+
+        metrics.reset()
+        model.eval()
+        for batch_id, data in enumerate(test_loader):
+            x, y = data
+            pred, trans_input, trans_feat = model(x)
+
+            correct = metrics.compute(pred, y)
+            metrics.update(correct)
+        test_acc = metrics.accumulate()
+        print("Test epoch: {}, acc is: {}".format(epoch, test_acc))
+
+        if test_acc > best_test_acc:
+            best_test_acc = test_acc
+            save_path = os.path.join(args.save_dir, 'best_model.pdparams')
+            paddle.save(model.state_dict(), save_path)
+            print("Best Test ACC: {}, Model saved in {}".format(
+                test_acc, save_path))
+        else:
+            print("Current Best Test ACC: {}".format(best_test_acc))
+
+
+if __name__ == "__main__":
+    args = parse_args()
+    print(args)
+    train(args)