meter seg

configs/unet_mechanical_meter.yaml

+EVAL_CROP_SIZE: (2049, 1537) # (width, height), for unpadding rangescaling and stepscaling
+TRAIN_CROP_SIZE: (769, 769) # (width, height), for unpadding rangescaling and stepscaling
+AUG: 
+    AUG_METHOD: u"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
+    RICH_CROP: 
+        ENABLE: False 
+
+BATCH_SIZE: 2
+MEAN: [0.5, 0.5, 0.5]
+STD: [0.5, 0.5, 0.5]
+
+DATALOADER: 
+    BUF_SIZE: 256
+    NUM_WORKERS: 4
+DATASET: 
+    DATA_DIR: "./dataset/mini_mechanical_industry_meter_data/"
+    IMAGE_TYPE: "rgb"  # choice rgb or rgba
+    NUM_CLASSES: 5
+    TEST_FILE_LIST: "./dataset/mini_mechanical_industry_meter_data/val_mini.txt"
+    TEST_TOTAL_IMAGES: 8
+    TRAIN_FILE_LIST: "./dataset/mini_mechanical_industry_meter_data/train_mini.txt"
+    TRAIN_TOTAL_IMAGES: 64
+    VAL_FILE_LIST: "./dataset/mini_mechanical_industry_meter_data/val_mini.txt"
+    VAL_TOTAL_IMAGES: 8
+    SEPARATOR: "|"
+    IGNORE_INDEX: 255
+
+FREEZE:
+    MODEL_FILENAME: "__model__"
+    PARAMS_FILENAME: "__params__"
+MODEL:
+    MODEL_NAME: "unet"
+    DEFAULT_NORM_TYPE: "bn"
+TEST:
+    TEST_MODEL: "./saved_model/unet_mechanical_meter/final/"
+TRAIN:
+    MODEL_SAVE_DIR: "./saved_model/unet_mechanical_meter/"
+    PRETRAINED_MODEL_DIR: "./pretrained_model/unet_bn_coco/"
+    SNAPSHOT_EPOCH: 10
+SOLVER:
+    NUM_EPOCHS: 100
+    LR: 0.001
+    LR_POLICY: "poly"
+    OPTIMIZER: "sgd"

contrib/README.md

@@ -99,6 +99,37 @@ python -u infer.py --example RoadLine
   
   预测结果：![](imgs/RoadLine.png)
 
+## 工业用表分割
+
+### 1. 模型结构
+
+unet
+
+### 2. 数据准备
+ 
+cd到PaddleSeg/dataset文件夹下，执行download_mini_mechanical_industry_meter.py
+
+
+### 3. 训练与评估
+
+```
+CUDA_VISIBLE_DEVICES=0 python ./pdseg/train.py --log_steps 10 --cfg configs/unet_mechanical_meter.yaml --use_gpu --do_eval --use_mpio 
+```
+
+### 4. 可视化
+我们提供了一个训练好的模型，点击[链接](https://paddleseg.bj.bcebos.com/models/unet_mechanical_industry_meter.tar)，下载后放在PaddleSeg/pretrained_model下
+```
+CUDA_VISIBLE_DEVICES=0 python ./pdseg/vis.py --cfg configs/unet_mechanical_meter.yaml --use_gpu --vis_dir vis_meter \
+TEST.TEST_MODEL "./pretrained_model/unet_gongyeyongbiao/" 
+```
+可视化结果会保存在vis_meter文件夹下
+
+### 5. 可视化结果示例：
+
+  原图：![](imgs/1560143028.5_IMG_3091.JPG)
+  
+  预测结果：![](imgs/1560143028.5_IMG_3091.png)
+  
 # 备注
 
 1. 数据及模型路径等详细配置见ACE2P/HumanSeg/RoadLine下的config.py文件

dataset/download_mini_mechanical_industry_meter.py

+# Copyright (c) 2019  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 sys
+import os
+
+LOCAL_PATH = os.path.dirname(os.path.abspath(__file__))
+TEST_PATH = os.path.join(LOCAL_PATH, "..", "test")
+sys.path.append(TEST_PATH)
+
+from test_utils import download_file_and_uncompress
+
+
+def download_deepglobe_road_dataset(savepath, extrapath):
+    url = "https://paddleseg.bj.bcebos.com/dataset/mini_mechanical_industry_meter_data.zip"
+    download_file_and_uncompress(
+        url=url, savepath=savepath, extrapath=extrapath)
+
+
+if __name__ == "__main__":
+    download_deepglobe_road_dataset(LOCAL_PATH, LOCAL_PATH)
+    print("Dataset download finish!")

docs/multiple_gpus_train_and_mixed_precision_train.md

@@ -2,7 +2,7 @@
 
 ### 环境要求
 * PaddlePaddle >= 1.6.0
-* NVIDIA NCCL >= 2.4.7，并在Linux环境下运行
+* NVIDIA NCCL >= 2.4.7
 
 环境配置，数据，预训练模型准备等工作请参考[安装说明](./installation.md)，[PaddleSeg使用说明](./usage.md)
 
@@ -49,12 +49,13 @@ python -m paddle.distributed.launch pdseg/train.py --use_gpu \
 
 ### benchmark
 
-| 模型 | 数据集合 | batch size | number gpu cards | 多进程训练 | 混合精度训练 | 显存占用 | 速度(image/s) | mIoU on val |
-|---|---|---|---|---|---|---|---|---|
-| DeepLabv3+/Xception65/bn | Cityscapes | 16 | 4 | False | False | 15988 MiB | 17.27 | 79.20 |
-| DeepLabv3+/Xception65/bn | Cityscapes | 16 | 4 | True | False | 15814 MiB | 19.80 | 78.90 |
-| DeepLabv3+/Xception65/bn | Cityscapes | 16 | 4 | True | True | 14922 MiB | 25.84 |79.06|
+| 模型 | 数据集合 | batch size | number gpu cards | 多进程训练 | 混合精度训练 | 速度(image/s) | mIoU on val |
+|---|---|---|---|---|---|---|---|
+| DeepLabv3+/Xception65/bn | Cityscapes | 16 | 4 | False | False | 17.27 | 79.20 |
+| DeepLabv3+/Xception65/bn | Cityscapes | 16 | 4 | True | False  | 19.80 | 78.90 |
+| DeepLabv3+/Xception65/bn | Cityscapes | 16 | 4 | True | True  | 25.84 |79.06|
 
+测试环境：python3.7.3，paddle1.6.0，cuda10，cudnn7.6.2，v100。
 
 ### 参考