delete bak files for SemSegPaddle (#4244)

delete bak files for SemSegPaddle

PaddleCV/Research/SemSegPaddle/src/models/modeling/glore.py.bak

-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
-import sys
-import paddle.fluid as fluid
-from paddle.fluid.param_attr import ParamAttr
-from src.models.libs.model_libs import scope, name_scope
-from src.models.libs.model_libs import avg_pool, conv, bn, conv1d, FCNHead
-from src.models.backbone.resnet import ResNet as resnet_backbone
-from src.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 gcn_module(name_scope, x, num_node, num_state):
-    '''
-    input: any tensor of 3D, B,C,N
-    '''
-    h = fluid.layers.transpose(x, perm=[0, 2, 1]) #B,C,N-->B,N,C
-    h = conv1d(h, num_node, name_scope+'_conv1d1')
-    h = fluid.layers.transpose(h, perm=[0, 2, 1]) #B,C,N
-    h = fluid.layers.elementwise_add(h, x, act='relu')
-    h = conv1d(h, num_state, name_scope+'_conv1d2')
-    return h
-
-def gru_module(x, num_state, num_node):
-    '''
-    Global Reasoning Unit: projection --> graph reasoning --> reverse projection
-    params:
-         x:  B x C x H x W
-         num_state: the dimension of each vertex feature
-         num_node: the number of vertet
-    output: B x C x H x W
-    feature trans:
-            B, C, H, W --> B, N, H, W -->             B, N, H*W -->B, N, C1 -->B, C1, N-->B, C1, N-->B, C1, H*W-->B, C, H, W
-                       --> B, C1,H, W -->B, C1,H*W -->B, H*W, C1
-    '''
-    num_batch, C, H, W = x.shape
-    with scope('projection'):
-        B = conv(x, num_node,
-                filter_size=1,
-                bias_attr=True,
-                name='projection'+'_conv') #num_batch, node, H, W
-        B = fluid.layers.reshape(B, shape=[num_batch, num_node, H*W]) #num_batch, node, L=H*W
-    with scope('reduce_channel'):
-        x_reduce = conv(x, num_state,
-                filter_size=1,
-                bias_attr=True,
-                name='reduce_channel'+'_conv') #num_batch, num_state, H, W
-        x_reduce = fluid.layers.reshape(x_reduce, shape=[num_batch, num_state, H*W]) #num_batch, num_state, L
-        x_reduce = fluid.layers.transpose(x_reduce, perm=[0, 2, 1]) #num_batch, L, num_state
-
-    V = fluid.layers.transpose(fluid.layers.matmul(B, x_reduce), perm=[0,2,1]) #num_batch, num_state, num_node
-    new_V = gcn_module('gru'+'_gcn', V, num_node, num_state)
-
-    D = fluid.layers.transpose(B, perm=[0, 2, 1])
-    Y = fluid.layers.matmul(D, fluid.layers.transpose(new_V, perm=[0, 2, 1]))
-    Y = fluid.layers.transpose(Y, perm=[0, 2, 1])
-    Y = fluid.layers.reshape(Y, shape=[num_batch, num_state, H, W])
-    with scope('extand_dim'):
-        Y = conv(Y, C, filter_size=1, bias_attr=True, name='extend_dim'+'_conv')
-        Y = bn(Y)
-    out = fluid.layers.elementwise_add(Y, x)
-    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.GLORE.DEPTH_MULTIPLIER
-    layers = cfg.MODEL.BACKBONE_LAYERS
-    end_points = layers - 1
-    dilation_dict = {2:2, 3:4}
-    decode_points= [91, 100]
-    model = resnet_backbone(layers, scale)
-    res5, feat_dict = model.net(input,
-                                end_points=end_points,
-                                dilation_dict=dilation_dict,
-                                decode_points= decode_points)
-
-    return res5, feat_dict
-
-def glore(input, num_classes):
-    """
-    Reference:
-       Chen, Yunpeng, et al. "Graph-Based Global Reasoning Networks", In CVPR 2019
-    """
-
-    # Backbone: ResNet
-    res5, feat_dict = resnet(input)
-    res4= feat_dict[91]
-    # Conv_1x1 for reduce dimension
-    with scope('feature'):
-        feature = conv(res5, 512, filter_size=1, bias_attr=True, name='feature_conv')
-        feature = bn(feature, act='relu')
-    # GRU Module
-    gru_output = gru_module(feature, 128, 64)
-    dropout = fluid.layers.dropout(gru_output, 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:])
-
-    if cfg.MODEL.GLORE.AuxHead:
-        aux_logit = FCNHead(res4, 256, num_classes, input.shape[2:])
-        return logit, aux_logit
-
-    return logit

PaddleCV/Research/SemSegPaddle/src/models/modeling/glore.py.bak1

-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
-import sys
-import paddle.fluid as fluid
-from paddle.fluid.param_attr import ParamAttr
-from src.models.libs.model_libs import scope, name_scope
-from src.models.libs.model_libs import avg_pool, conv, bn, bn_zero, conv1d, FCNHead
-from src.models.backbone.resnet import ResNet as resnet_backbone
-from src.utils.config import cfg
-
-
-def get_logit_interp(input, num_classes, out_shape, name="logit"):
-    # 1x1_Conv
-    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 gcn_module(name_scope, x, num_node, num_state):
-    '''
-    input: any tensor of 3D, B,C,N
-    '''
-    print(x.shape)
-    h = fluid.layers.transpose(x, perm=[0, 2, 1]) #B,C,N-->B,N,C
-    h = conv1d(h, num_node, name_scope+'_conv1d1')
-    h = fluid.layers.transpose(h, perm=[0, 2, 1]) #B,C,N
-    h = fluid.layers.elementwise_add(h, x, act='relu')
-    h = conv1d(h, num_state, name_scope+'_conv1d2')
-    return h
-
-def gru_module(x, num_state, num_node):
-    '''
-    Global Reasoning Unit: projection --> graph reasoning --> reverse projection
-    params:
-         x:  B x C x H x W
-         num_state: the dimension of each vertex feature
-         num_node: the number of vertet
-    output: B x C x H x W
-    feature trans:
-            B, C, H, W --> B, N, H, W -->             B, N, H*W -->B, N, C1 -->B, C1, N-->B, C1, N-->B, C1, H*W-->B, C, H, W
-                       --> B, C1,H, W -->B, C1,H*W -->B, H*W, C1
-    '''
-    num_batch, C, H, W = x.shape
-    with scope('projection'):
-        B = conv(x, num_node,
-                filter_size=1,
-                bias_attr=True,
-                name='projection'+'_conv') #num_batch, node, H, W
-        B = fluid.layers.reshape(B, shape=[num_batch, num_node, H*W]) #num_batch, node, L=H*W
-    with scope('reduce_channel'):
-        x_reduce = conv(x, num_state,
-                filter_size=1,
-                bias_attr=True,
-                name='reduce_channel'+'_conv') #num_batch, num_state, H, W
-        x_reduce = fluid.layers.reshape(x_reduce, shape=[num_batch, num_state, H*W]) #num_batch, num_state, L
-        x_reduce = fluid.layers.transpose(x_reduce, perm=[0, 2, 1]) #num_batch, L, num_state
-
-    V = fluid.layers.transpose(fluid.layers.matmul(B, x_reduce), perm=[0,2,1]) #num_batch, num_state, num_node
-    L = fluid.layers.fill_constant(shape=[1], value=H*W, dtype='float32')
-    V = fluid.layers.elementwise_div(V, L)
-    new_V = gcn_module('gru'+'_gcn', V, num_node, num_state)
-    D = fluid.layers.transpose(B, perm=[0, 2, 1])
-    Y = fluid.layers.matmul(D, fluid.layers.transpose(new_V, perm=[0, 2, 1]))
-    Y = fluid.layers.transpose(Y, perm=[0, 2, 1])
-    Y = fluid.layers.reshape(Y, shape=[num_batch, num_state, H, W])
-    with scope('extend_dim'):
-        Y = conv(Y, C, filter_size=1, bias_attr=True, name='extend_dim'+'_conv')
-        #Y = bn_zero(Y)
-        Y = bn(Y)
-    out = fluid.layers.elementwise_add(Y, x)
-    return out
-
-def resnet(input):
-    # end_points: end_layer of resnet backbone 
-    # dilation_dict: dilation factor for stages_key
-    scale = cfg.MODEL.GLORE.DEPTH_MULTIPLIER
-    layers = cfg.MODEL.BACKBONE_LAYERS
-    end_points = layers - 1
-    dilation_dict = {2:2, 3:4}
-    decode_points= [91, 100]
-    model = resnet_backbone(layers, scale)
-    res5, feat_dict = model.net(input,
-                                end_points=end_points,
-                                dilation_dict=dilation_dict,
-                                decode_points= decode_points)
-
-    return res5, feat_dict
-
-def glore(input, num_classes):
-    """
-    Reference:
-       Chen, Yunpeng, et al. "Graph-Based Global Reasoning Networks", In CVPR 2019
-    """
-
-    # Backbone: ResNet
-    res5, feat_dict = resnet(input)
-    res4= feat_dict[91]
-    # Conv_1x1 for reduce dimension
-    with scope('feature'):
-        feature = conv(res5, 512, filter_size=3, bias_attr=True, name='feature_conv')
-        feature = bn(feature, act='relu')
-    # GRU Module
-    gru_output = gru_module(feature, 128, 64)
-    dropout = fluid.layers.dropout(gru_output, dropout_prob=0.1, name="dropout")
-
-    logit = get_logit_interp(dropout, num_classes, input.shape[2:])
-    if cfg.MODEL.GLORE.AuxHead:
-        aux_logit = FCNHead(res4, 256, num_classes, input.shape[2:])
-        return logit, aux_logit
-
-    return logit
-

README.md

@@ -75,9 +75,9 @@ PaddlePaddle 提供了丰富的计算单元，使得用户可以采用模块化
 
 | 模型名称                                                     | 模型简介                                                     | 数据集    | 评估指标        |
 | ------------------------------------------------------------ | ------------------------------------------------------------ | --------- | --------------- |
-| [ICNet](https://github.com/PaddlePaddle/models/tree/develop/PaddleCV/icnet) | 主要用于图像实时语义分割，能够兼顾速度和准确性，易于线上部署 | Cityscape | Mean IoU=67.0%  |
-| [DeepLab   V3+](https://github.com/PaddlePaddle/models/tree/develop/PaddleCV/deeplabv3%2B) | 通过 encoder-decoder 进行多尺度信息的融合，同时保留了原来的空洞卷积和 ASSP 层，   其骨干网络使用了 Xception 模型，提高了语义分割的健壮性和运行速率 | Cityscape | Mean IoU=78.81% |
-| [PSPNet (res101)](https://github.com/PaddlePaddle/models/tree/develop/PaddleCV/Research/SemSegPaddle) | 通过利用不同子区域和全局的上下文信息来增强语义分割质量，同时提出deeply supervised 的辅助loss去改善模型的优化 | Cityscape | Mean IoU = 78.1 |
+| [ICNet](https://github.com/PaddlePaddle/models/tree/develop/PaddleCV/icnet) | 主要用于图像实时语义分割，能够兼顾速度和准确性，易于线上部署 | Cityscapes | Mean IoU=67.0%  |
+| [DeepLab   V3+](https://github.com/PaddlePaddle/models/tree/develop/PaddleCV/deeplabv3%2B) | 通过 encoder-decoder 进行多尺度信息的融合，同时保留了原来的空洞卷积和 ASSP 层，   其骨干网络使用了 Xception 模型，提高了语义分割的健壮性和运行速率 | Cityscapes | Mean IoU=78.81% |
+| [PSPNet (res101)](https://github.com/PaddlePaddle/models/tree/develop/PaddleCV/Research/SemSegPaddle) | 通过利用不同子区域和全局的上下文信息来增强语义分割质量，同时提出deeply supervised 的辅助loss去改善模型的优化 | Cityscapes | Mean IoU = 78.1 |
 | [GloRe (res101)](https://github.com/PaddlePaddle/models/tree/develop/PaddleCV/Research/SemSegPaddle)| 提出一个轻量级的、可端到端训练的全局推理单元GloRe来高效推理image regions之间的关系，增强了模型上下文建模能力| Cityscapes | Mean IoU = 78.4 |
 | [PSPNet (res101)](https://github.com/PaddlePaddle/models/tree/develop/PaddleCV/Research/SemSegPaddle) | -| PASCAL Context | Mean IoU = 48.9  |
 | [GloRe (res101)](https://github.com/PaddlePaddle/models/tree/develop/PaddleCV/Research/SemSegPaddle)| -| PASCAL Context | Mean IoU = 48.4 |