Merge pull request #15 from pennypm/master

add pspnet

configs/pspnet.yaml

+EVAL_CROP_SIZE: (2049, 1025) # (width, height), for unpadding rangescaling and stepscaling
+TRAIN_CROP_SIZE: (713, 713) # (width, height), for unpadding rangescaling and stepscaling
+AUG:
+    AUG_METHOD: "stepscaling" # choice unpadding rangescaling and stepscaling
+    FIX_RESIZE_SIZE: (640, 640) # (width, height), for unpadding
+    INF_RESIZE_VALUE: 500  # for rangescaling
+    MAX_RESIZE_VALUE: 600  # for rangescaling
+    MIN_RESIZE_VALUE: 400  # for rangescaling
+    MAX_SCALE_FACTOR: 2.0  # for stepscaling
+    MIN_SCALE_FACTOR: 0.5  # for stepscaling
+    SCALE_STEP_SIZE: 0.25  # for stepscaling
+    MIRROR: True
+BATCH_SIZE: 4
+DATASET:
+    DATA_DIR: "./dataset/cityscapes/"
+    IMAGE_TYPE: "rgb"  # choice rgb or rgba
+    NUM_CLASSES: 19
+    TEST_FILE_LIST: "dataset/cityscapes/val.list"
+    TRAIN_FILE_LIST: "dataset/cityscapes/train.list"
+    VAL_FILE_LIST: "dataset/cityscapes/val.list"
+    IGNORE_INDEX: 255
+FREEZE:
+    MODEL_FILENAME: "model"
+    PARAMS_FILENAME: "params"
+MODEL:
+    MODEL_NAME: "pspnet"
+    DEFAULT_NORM_TYPE: "bn"
+TEST:
+    TEST_MODEL: "pretrained_model/pspnet50_ADE20K/"
+TRAIN:
+    MODEL_SAVE_DIR: "snapshots/cityscape_pspnet50/"
+    PRETRAINED_MODEL_DIR: u"pretrained_model/pspnet50_ADE20K/"
+    SNAPSHOT_EPOCH: 10
+SOLVER:
+    LR: 0.001
+    LR_POLICY: "poly"
+    OPTIMIZER: "sgd"
+    NUM_EPOCHS: 700

pdseg/data_aug.py

@@ -374,7 +374,7 @@ def rand_crop(crop_img, crop_seg, mode=ModelPhase.TRAIN):
     Args:
         crop_img(numpy.ndarray): 输入图像
         crop_seg(numpy.ndarray): 标签图
-        mode(string): 模式, 默认训练模式，验证或预测模式时crop尺寸需大于原始图片尺寸, 其他模式无限制
+        mode(string): 模式, 默认训练模式，验证或预测、可视化模式时crop尺寸需大于原始图片尺寸
 
     Returns：
         裁剪后的图片和标签图
@@ -391,7 +391,7 @@ def rand_crop(crop_img, crop_seg, mode=ModelPhase.TRAIN):
         crop_width = cfg.EVAL_CROP_SIZE[0]
         crop_height = cfg.EVAL_CROP_SIZE[1]
 
-    if ModelPhase.is_eval(mode) or ModelPhase.is_predict(mode):
+    if not ModelPhase.is_train(mode): 
         if (crop_height < img_height or crop_width < img_width):
             raise Exception(
                 "Crop size({},{}) must large than img size({},{}) when in EvalPhase."

pdseg/models/backbone/resnet.py

@@ -85,7 +85,7 @@ class ResNet():
             depth = [3, 8, 36, 3]
         num_filters = [64, 128, 256, 512]
 
-        if self.stem == 'icnet':
+        if self.stem == 'icnet' or self.stem == 'pspnet':
             conv = self.conv_bn_layer(
                 input=input,
                 num_filters=int(64 * self.scale),
@@ -139,7 +139,7 @@ class ResNet():
                         else:
                             conv_name = "res" + str(block + 2) + "b" + str(i)
                     else:
-                        conv_name = "conv" + str(block + 2) + '_' + str(1 + i)
+                        conv_name = "res" + str(block + 2) + chr(97 + i)
                     dilation_rate = get_dilated_rate(dilation_dict, block)
                     
                     conv = self.bottleneck_block(
@@ -215,6 +215,12 @@ class ResNet():
                       groups=1,
                       act=None,
                       name=None):
+   
+        if self.stem == 'pspnet':
+            bias_attr=ParamAttr(name=name + "_biases")
+        else:
+            bias_attr=False
+
         conv = fluid.layers.conv2d(
             input=input,
             num_filters=num_filters,
@@ -224,20 +230,21 @@ class ResNet():
             dilation=dilation,
             groups=groups,
             act=None,
-            param_attr=ParamAttr(name=name + "/weights"),
-            bias_attr=False,
+            param_attr=ParamAttr(name=name + "_weights"),
+            bias_attr=bias_attr,
             name=name + '.conv2d.output.1')
 
-        bn_name = name + '/BatchNorm/'
-        return fluid.layers.batch_norm(
-            input=conv,
+        if name == "conv1":
+            bn_name = "bn_" + name
+        else:
+            bn_name = "bn" + name[3:]
+        return fluid.layers.batch_norm(input=conv,
                                        act=act,
                                        name=bn_name + '.output.1',
-            param_attr=ParamAttr(name=bn_name + 'gamma'),
-            bias_attr=ParamAttr(bn_name + 'beta'),
-            moving_mean_name=bn_name + 'moving_mean',
-            moving_variance_name=bn_name + 'moving_variance',
-        )
+                                       param_attr=ParamAttr(name=bn_name + '_scale'),
+                                       bias_attr=ParamAttr(bn_name + '_offset'),
+                                       moving_mean_name=bn_name + '_mean',
+                                       moving_variance_name=bn_name + '_variance', )
 
     def shortcut(self, input, ch_out, stride, is_first, name):
         ch_in = input.shape[1]
@@ -247,12 +254,17 @@ class ResNet():
             return input
 
     def bottleneck_block(self, input, num_filters, stride, name, dilation=1):
+        if self.stem == 'pspnet' and self.layers == 101:
+            strides = [1, stride]
+        else:
+            strides = [stride, 1]
+        
         conv0 = self.conv_bn_layer(
             input=input,
             num_filters=num_filters,
             filter_size=1,
             dilation=1,
-            stride=stride,
+            stride=strides[0],
             act='relu',
             name=name + "_branch2a")
         if dilation > 1:
@@ -262,6 +274,7 @@ class ResNet():
             num_filters=num_filters,
             filter_size=3,
             dilation=dilation,
+            stride=strides[1],
             act='relu',
             name=name + "_branch2b")
         conv2 = self.conv_bn_layer(

pdseg/models/model_builder.py

@@ -73,6 +73,7 @@ def map_model_name(model_name):
         "unet": "unet.unet",
         "deeplabv3p": "deeplab.deeplabv3p",
         "icnet": "icnet.icnet",
+        "pspnet": "pspnet.pspnet",
     }
     if model_name in name_dict.keys():
         return name_dict[model_name]

pdseg/models/modeling/pspnet.py

+# coding: utf8
+# copyright (c) 2019 PaddlePaddle Authors. All Rights Reserve.
+#
+# 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.
+
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+
+import paddle.fluid as fluid
+from paddle.fluid.param_attr import ParamAttr
+from models.libs.model_libs import scope, name_scope
+from models.libs.model_libs import avg_pool, conv, bn
+from models.backbone.resnet import ResNet as resnet_backbone
+from utils.config import cfg
+
+def get_logit_interp(input, num_classes, out_shape, name="logit"):
+    # 根据类别数决定最后一层卷积输出, 并插值回原始尺寸
+    param_attr = fluid.ParamAttr(
+        name=name + 'weights',
+        regularizer=fluid.regularizer.L2DecayRegularizer(
+            regularization_coeff=0.0),
+        initializer=fluid.initializer.TruncatedNormal(loc=0.0, scale=0.01))
+
+    with scope(name):
+        logit = conv(input,
+                    num_classes,
+                    filter_size=1,
+                    param_attr=param_attr,
+                    bias_attr=True,
+                    name=name+'_conv')
+        logit_interp = fluid.layers.resize_bilinear(
+                    logit, 
+                    out_shape=out_shape,
+                    name=name+'_interp') 
+    return logit_interp
+
+
+def psp_module(input, out_features):
+    # Pyramid Scene Parsing 金字塔池化模块
+    # 输入：backbone输出的特征
+    # 输出：对输入进行不同尺度pooling, 卷积操作后插值回原始尺寸，并concat
+    #       最后进行一个卷积及BN操作
+    
+    cat_layers = []
+    sizes = (1,2,3,6)
+    for size in sizes:
+        psp_name = "psp" + str(size)
+        with scope(psp_name):
+            pool = fluid.layers.adaptive_pool2d(input, 
+                    pool_size=[size, size], 
+                    pool_type='avg', 
+                    name=psp_name+'_adapool')
+            data = conv(pool, out_features, 
+                    filter_size=1,
+                    bias_attr=True, 
+                    name= psp_name + '_conv')
+            data_bn = bn(data, act='relu')
+            interp = fluid.layers.resize_bilinear(data_bn, 
+                    out_shape=input.shape[2:], 
+                    name=psp_name+'_interp') 
+        cat_layers.append(interp)
+    cat_layers = [input] + cat_layers[::-1]
+    cat = fluid.layers.concat(cat_layers, axis=1, name='psp_cat')
+
+    psp_end_name = "psp_end"
+    with scope(psp_end_name):
+        data = conv(cat, 
+                    out_features, 
+                    filter_size=3,
+                    padding=1, 
+                    bias_attr=True,
+                    name=psp_end_name)
+        out = bn(data, act='relu')
+
+    return out
+
+def resnet(input):
+    # PSPNET backbone: resnet, 默认resnet50
+    # end_points: resnet终止层数
+    # dilation_dict: resnet block数及对应的膨胀卷积尺度
+    scale = cfg.MODEL.PSPNET.DEPTH_MULTIPLIER
+    layers = cfg.MODEL.PSPNET.LAYERS
+    end_points = layers - 1
+    dilation_dict = {2:2, 3:4}
+    model = resnet_backbone(layers, scale, stem='pspnet')
+    data, _ = model.net(input, 
+                        end_points=end_points, 
+                        dilation_dict=dilation_dict)
+
+    return data
+
+def pspnet(input, num_classes):
+    # Backbone: ResNet
+    res = resnet(input)
+    # PSP模块
+    psp = psp_module(res, 512)
+    dropout = fluid.layers.dropout(psp, dropout_prob=0.1, name="dropout")
+    # 根据类别数决定最后一层卷积输出, 并插值回原始尺寸
+    logit = get_logit_interp(dropout, num_classes, input.shape[2:])
+    return logit
+

pdseg/utils/config.py

@@ -196,6 +196,12 @@ cfg.MODEL.ICNET.DEPTH_MULTIPLIER = 0.5
 # RESNET 层数 设置
 cfg.MODEL.ICNET.LAYERS = 50
 
+########################## PSPNET模型配置 ######################################
+# RESNET backbone scale 设置
+cfg.MODEL.PSPNET.DEPTH_MULTIPLIER = 1
+# RESNET 层数 设置 50或101
+cfg.MODEL.PSPNET.LAYERS = 50
+
 ########################## 预测部署模型配置 ###################################
 # 预测保存的模型名称
 cfg.FREEZE.MODEL_FILENAME = '__model__'