add hrnet semantic segmentation model w18, w32 and w48 (wang-xinyu#503)

* create psenet

create psenet with weight from tensorflow

* delete some useless code

* repalce tab with 4 blanks

* fix network bug, rewrite post-processing pse algorithm

* update readme

* update readme

* add RepVGG

* fix typo

* add hrnetseg w18 w32 w48

* add hrnetseg with ocr w18 w32 w48

* merge hrnet and small, add hrnet_ocr

* fix warning

* change project name

Author: upczww

hrnet/hrnet-semantic-segmentation/CMakeLists.txt

@@ -24,10 +24,17 @@ set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++11 -Wall -Ofast -Wfatal-errors -
 find_package(OpenCV)
 include_directories(${OpenCV_INCLUDE_DIRS})
 
-add_executable(hrnetseg ${PROJECT_SOURCE_DIR}/hrnetseg.cpp)
-target_link_libraries(hrnetseg nvinfer)
-target_link_libraries(hrnetseg cudart)
-target_link_libraries(hrnetseg ${OpenCV_LIBS})
+add_executable(hrnet ${PROJECT_SOURCE_DIR}/hrnet.cpp)
+target_link_libraries(hrnet nvinfer)
+target_link_libraries(hrnet cudart)
+target_link_libraries(hrnet ${OpenCV_LIBS})
+
+
+add_executable(hrnet_ocr ${PROJECT_SOURCE_DIR}/hrnet_ocr.cpp)
+target_link_libraries(hrnet_ocr nvinfer)
+target_link_libraries(hrnet_ocr cudart)
+target_link_libraries(hrnet_ocr ${OpenCV_LIBS})
+
 
 add_definitions(-O2 -pthread)
 

hrnet/hrnet-semantic-segmentation/README.md

@@ -1,27 +1,76 @@
 # HRNet-Semantic-Segmentation
 
-The Pytorch implementation is [HRNet-Semantic-Segmentation-v1.1](https://github.com/HRNet/HRNet-Semantic-Segmentation/tree/pytorch-v1.1).  The implemented model is **HRNetV2-W18-Small-v2**
+This repo implemtents [HRNet-Semantic-Segmentation-v1.1](https://github.com/HRNet/HRNet-Semantic-Segmentation/tree/pytorch-v1.1) and [HRNet-Semantic-Segmentation-OCR](https://github.com/HRNet/HRNet-Semantic-Segmentation/tree/HRNet-OCR).
 
 
 ## How to Run
-
-* 1. generate .wts
-
-  Download code and model from [HRNet-Semantic-Segmentation-v1.1](https://github.com/HRNet/HRNet-Semantic-Segmentation/tree/pytorch-v1.1) and config your environments.
-
-  Put `demo.py`  in the `YOUR_ROOT_DIR\HRNet-Semantic-Segmentation\tools `  folder, set `savewts in  main()` as `True`, and run, the .wts will be generated.
-
-* 2. cmake and make
+### For HRNet-Semantic-Segmentation-v1.1
+1. generate .wts, use config `experiments/cityscapes/seg_hrnet_w48_train_512x1024_sgd_lr1e-2_wd5e-4_bs_12_epoch484.yaml` and pretrained weight `hrnet_w48_cityscapes_cls19_1024x2048_trainset.pth` as example. change `PRETRAINED` in `experiments/cityscapes/seg_hrnet_w48_train_512x1024_sgd_lr1e-2_wd5e-4_bs_12_epoch484.yaml` to `""`.
+```
+cp gen_wts.py $HRNET--Semantic-Segmentation-PROJECT-ROOT/tools
+cd $HRNET--Semantic-Segmentation-PROJECT-ROOT
+python tools/gen_wts.py --cfg experiments/cityscapes/seg_hrnet_w48_train_512x1024_sgd_lr1e-2_wd5e-4_bs_12_epoch484.yaml --ckpt_path hrnet_w48_cityscapes_cls19_1024x2048_trainset.pth --save_path hrnet_w48.wts
+cp hrnet_w48.wts $HRNET-TENSORRT-ROOT
+cd $HRNET-TENSORRT-ROOT
+```
+2. cmake and make
 
   ```
   mkdir build
   cd build
   cmake ..
   make
-  sudo ./hrnetseg -s             // serialize model to plan file i.e. 'hrnetseg.engine'
-  sudo ./hrnetseg -d  ../samples // deserialize plan file and run inference, the images in samples will be processed.
   ```
+  first serialize model to plan file
+  ```
+  ./hrnet -s [.wts] [.engine] [small or 18 or 32 or 48] # small for W18-Small-v2, 18 for W18, etc.
+  ```
+  such as
+  ```
+  ./hrnet -s ../hrnet_w48.wts ./hrnet_w48.engine 48
+  ```
+  then deserialize plan file and run inference
+  ```
+  ./hrnet -d  [.engine] [image dir]
+  ```
+  such as 
+  ```
+  ./hrnet -d  ./hrnet_w48.engine ../samples
+  ```
+### For HRNet-Semantic-Segmentation-OCR
+
+1. generate .wts, use config `experiments/cityscapes/seg_hrnet_ocr_w48_train_512x1024_sgd_lr1e-2_wd5e-4_bs_12_epoch484.yaml` and pretrained weight `hrnet_ocr_cs_8162_torch11.pth` as example. change `PRETRAINED` in `experiments/cityscapes/seg_hrnet_ocr_w48_train_512x1024_sgd_lr1e-2_wd5e-4_bs_12_epoch484.yaml` to `""`.
+```
+cp gen_wts.py $HRNET-OCR-TRAIN-PROJECT-ROOT/tools
+cd $HRNET-OCR-PROJECT-ROOT
+python tools/gen_wts.py --cfg experiments/cityscapes/seg_hrnet_ocr_w48_train_512x1024_sgd_lr1e-2_wd5e-4_bs_12_epoch484.yaml --ckpt_path hrnet_ocr_cs_8162_torch11.pth --save_path hrnet_ocr_w48.wts
+cp hrnet_ocr_w48.wts $HRNET-OCR-TENSORRT-ROOT
+cd $HRNET-OCR-TENSORRT-ROOT
+```
+2. cmake and make
 
+  ```
+  mkdir build
+  cd build
+  cmake ..
+  make
+  ```
+  first serialize model to plan file
+  ```
+  ./hrnet_ocr -s [.wts] [.engine] [18 or 32 or 48]
+  ```
+  such as
+  ```
+  ./hrnet_ocr -s ../hrnet_ocr_w48.wts ./hrnet_ocr_w48.engine 48
+  ```
+  then deserialize plan file and run inference
+  ```
+  ./hrnet_ocr -d  [.engine] [image dir]
+  ```
+  such as 
+  ```
+  ./hrnet_ocr -d  ./hrnet_ocr_w48.engine ../samples
+  ```
 ## Result
 
 TRT Result: