Merge branch 'dygraph' of https://github.com/wuyefeilin/PaddleSeg into dygraph

README.md

@@ -89,12 +89,12 @@ pip install -r requirements.txt
 * [数据和配置校验](./docs/check.md)
 * [分割模型介绍](./docs/models.md)
 * [预训练模型下载](./docs/model_zoo.md)
-* [DeepLabv3+模型使用教程](./turtorial/finetune_deeplabv3plus.md)
-* [U-Net模型使用教程](./turtorial/finetune_unet.md)
-* [ICNet模型使用教程](./turtorial/finetune_icnet.md)
-* [PSPNet模型使用教程](./turtorial/finetune_pspnet.md)
-* [HRNet模型使用教程](./turtorial/finetune_hrnet.md)
-* [Fast-SCNN模型使用教程](./turtorial/finetune_fast_scnn.md)
+* [DeepLabv3+模型使用教程](./tutorial/finetune_deeplabv3plus.md)
+* [U-Net模型使用教程](./tutorial/finetune_unet.md)
+* [ICNet模型使用教程](./tutorial/finetune_icnet.md)
+* [PSPNet模型使用教程](./tutorial/finetune_pspnet.md)
+* [HRNet模型使用教程](./tutorial/finetune_hrnet.md)
+* [Fast-SCNN模型使用教程](./tutorial/finetune_fast_scnn.md)
 
 ### 预测部署
 

docs/model_zoo.md

@@ -14,6 +14,7 @@ PaddleSeg对所有内置的分割模型都提供了公开数据集下的预训
 | MobileNetV2_0.5x  | ImageNet | 0.5x | [MobileNetV2_0.5x](https://paddle-imagenet-models-name.bj.bcebos.com/MobileNetV2_x0_5_pretrained.tar) | 65.03%/85.72% |
 | MobileNetV2_1.5x  | ImageNet | 1.5x | [MobileNetV2_1.5x](https://paddle-imagenet-models-name.bj.bcebos.com/MobileNetV2_x1_5_pretrained.tar) | 74.12%/91.67% |
 | MobileNetV2_2.0x  | ImageNet | 2.0x | [MobileNetV2_2.0x](https://paddle-imagenet-models-name.bj.bcebos.com/MobileNetV2_x2_0_pretrained.tar) | 75.23%/92.58% |
+| MobileNetV3_Large_ssld_1.0x  | ImageNet | 1.0x | [MobileNetV3_Large_ssld_1.0x](https://paddleseg.bj.bcebos.com/models/MobileNetV3_large_x1_0_ssld_pretrained.tar) | 79.00%/94.50% |
 
 用户可以结合实际场景的精度和预测性能要求，选取不同`Depth multiplier`参数的MobileNet模型。
 
@@ -58,6 +59,7 @@ train数据集合为Cityscapes训练集合，测试为Cityscapes的验证集合
 | 模型 | 数据集合 | 下载地址 |Output Stride| mutli-scale test| mIoU on val|
 |---|---|---|---|---|---|
 | DeepLabv3+/MobileNetv2/bn | Cityscapes |[mobilenet_cityscapes.tgz](https://paddleseg.bj.bcebos.com/models/mobilenet_cityscapes.tgz) |16|false| 0.698|
+| DeepLabv3+/MobileNetv3_Large/bn | Cityscapes |[deeplabv3p_mobilenetv3_large_cityscapes.tar.gz](https://paddleseg.bj.bcebos.com/models/deeplabv3p_mobilenetv3_large_cityscapes.tar.gz) |32|false| 0.7328|
 | DeepLabv3+/Xception65/gn  | Cityscapes |[deeplabv3p_xception65_gn_cityscapes.tgz](https://paddleseg.bj.bcebos.com/models/deeplabv3p_xception65_cityscapes.tgz) |16|false| 0.7824 |
 | DeepLabv3+/Xception65/bn | Cityscapes |[deeplabv3p_xception65_bn_cityscapes.tgz](https://paddleseg.bj.bcebos.com/models/xception65_bn_cityscapes.tgz) | 16 | false | 0.7930 |
 | DeepLabv3+/ResNet50_vd/bn | Cityscapes |[deeplabv3p_resnet50_vd_cityscapes.tgz](https://paddleseg.bj.bcebos.com/models/deeplabv3p_resnet50_vd_cityscapes.tgz) | 16 | false | 0.8006 |

pretrained_model/download_model.py

@@ -24,6 +24,8 @@ from test_utils import download_file_and_uncompress
 
 model_urls = {
     # ImageNet Pretrained
+    "mobilenetv3_large_ssld_imagenet":
+    "https://paddleseg.bj.bcebos.com/models/MobileNetV3_large_x1_0_ssld_pretrained.tar",
     "mobilenetv2-2-0_bn_imagenet":
     "https://paddle-imagenet-models-name.bj.bcebos.com/MobileNetV2_x2_0_pretrained.tar",
     "mobilenetv2-1-5_bn_imagenet":
@@ -72,6 +74,8 @@ model_urls = {
     "https://paddleseg.bj.bcebos.com/models/pspnet101_coco.tgz",
 
     # Cityscapes pretrained
+    "deeplabv3p_mobilenetv3_large_cityscapes":
+    "https://paddleseg.bj.bcebos.com/models/deeplabv3p_mobilenetv3_large_cityscapes.tar.gz",
     "deeplabv3p_mobilenetv2-1-0_bn_cityscapes":
     "https://paddleseg.bj.bcebos.com/models/mobilenet_cityscapes.tgz",
     "deeplabv3p_xception65_gn_cityscapes":

tutorial/finetune_deeplabv3plus.md

@@ -149,6 +149,7 @@ PaddleSeg在AI Studio平台上提供了在线体验的DeepLabv3+图像分割教
 |deeplabv3p_mobilenetv2-1-0_bn_coco|MobileNetV2|bn|COCO|MODEL.MODEL_NAME: deeplabv3p <br> MODEL.DEEPLAB.BACKBONE: mobilenetv2 <br> MODEL.DEEPLAB.DEPTH_MULTIPLIER: 1.0 <br> MODEL.DEEPLAB.ENCODER_WITH_ASPP: False <br> MODEL.DEEPLAB.ENABLE_DECODER: False <br> MODEL.DEFAULT_NORM_TYPE: bn|
 |**deeplabv3p_xception65_bn_coco**|Xception65|bn|COCO|MODEL.MODEL_NAME: deeplabv3p <br> MODEL.DEEPLAB.BACKBONE: xception_65 <br> MODEL.DEFAULT_NORM_TYPE: bn |
 |deeplabv3p_mobilenetv2-1-0_bn_cityscapes|MobileNetV2|bn|Cityscapes|MODEL.MODEL_NAME: deeplabv3p <br> MODEL.DEEPLAB.BACKBONE: mobilenetv2 <br> MODEL.DEEPLAB.DEPTH_MULTIPLIER: 1.0 <br> MODEL.DEEPLAB.ENCODER_WITH_ASPP: False <br> MODEL.DEEPLAB.ENABLE_DECODER: False <br> MODEL.DEFAULT_NORM_TYPE: bn|
+|deeplabv3p_mobilenetv3_large_cityscapes|MobileNetV3|bn|Cityscapes|MODEL.MODEL_NAME: deeplabv3p <br> MODEL.DEEPLAB.BACKBONE: mobilenetv3_large <br> MODEL.DEFAULT_NORM_TYPE: bn|
 |deeplabv3p_xception65_gn_cityscapes|Xception65|gn|Cityscapes|MODEL.MODEL_NAME: deeplabv3p <br>  MODEL.DEEPLAB.BACKBONE: xception_65 <br> MODEL.DEFAULT_NORM_TYPE: gn|
 |deeplabv3p_xception65_bn_cityscapes|Xception65|bn|Cityscapes|MODEL.MODEL_NAME: deeplabv3p <br> MODEL.DEEPLAB.BACKBONE: xception_65 <br> MODEL.DEFAULT_NORM_TYPE: bn|
 |deeplabv3p_resnet50_vd_cityscapes|resnet50_vd|bn|Cityscapes|MODEL.MODEL_NAME: deeplabv3p <br> MODEL.DEEPLAB.BACKBONE: resnet50_vd <br> MODEL.DEFAULT_NORM_TYPE: bn|