Merge from develop (#9)

* Fix compile problem in different python version (#26) * fix some usage problem in linux * Fix compile problem Co-authored-by: root <root@bjyz-sys-gpu-kongming3.bjyz.baidu.com> * Add PaddleDetetion/PPYOLOE model support (#22) * add ppdet/ppyoloe * Add demo code and documents * add convert processor to vision (#27) * update .gitignore * Added checking for cmake include dir * fixed missing trt_backend option bug when init from trt * remove un-need data layout and add pre-check for dtype * changed RGB2BRG to BGR2RGB in ppcls model * add model_zoo yolov6 c++/python demo * fixed CMakeLists.txt typos * update yolov6 cpp/README.md * add yolox c++/pybind and model_zoo demo * move some helpers to private * fixed CMakeLists.txt typos * add normalize with alpha and beta * add version notes for yolov5/yolov6/yolox * add copyright to yolov5.cc * revert normalize * fixed some bugs in yolox * fixed examples/CMakeLists.txt to avoid conflicts * add convert processor to vision * format examples/CMakeLists summary * Fix bug while the inference result is empty with YOLOv5 (#29) * Add multi-label function for yolov5 * Update README.md Update doc * Update fastdeploy_runtime.cc fix variable option.trt_max_shape wrong name * Update runtime_option.md Update resnet model dynamic shape setting name from images to x * Fix bug when inference result boxes are empty * Delete detection.py Co-authored-by: Jason <jiangjiajun@baidu.com> Co-authored-by: root <root@bjyz-sys-gpu-kongming3.bjyz.baidu.com> Co-authored-by: DefTruth <31974251+DefTruth@users.noreply.github.com> Co-authored-by: huangjianhui <852142024@qq.com>
ziqi-jin · Jul 20, 2022 · e8a8439 · e8a8439
1 parent e25e4f2
commit e8a8439
Show file tree

Hide file tree

Showing 29 changed files with 818 additions and 50 deletions.
diff --git a/examples/CMakeLists.txt b/examples/CMakeLists.txt
@@ -1,24 +1,26 @@
-function(add_fastdeploy_executable field url model)   
+function(add_fastdeploy_executable FIELD CC_FILE)
   # temp target name/file var in function scope
-  set(TEMP_TARGET_FILE ${PROJECT_SOURCE_DIR}/examples/${field}/${url}_${model}.cc)
-  set(TEMP_TARGET_NAME ${field}_${url}_${model})
+  set(TEMP_TARGET_FILE ${CC_FILE})
+  string(REGEX MATCHALL "[0-9A-Za-z_]*.cc" FILE_NAME ${CC_FILE})
+  string(REGEX REPLACE ".cc" "" FILE_PREFIX ${FILE_NAME})
+  set(TEMP_TARGET_NAME ${FIELD}_${FILE_PREFIX})
   if (EXISTS ${TEMP_TARGET_FILE} AND TARGET fastdeploy)
     add_executable(${TEMP_TARGET_NAME} ${TEMP_TARGET_FILE})
     target_link_libraries(${TEMP_TARGET_NAME} PUBLIC fastdeploy)
-    message(STATUS "Found source file: [${field}/${url}_${model}.cc], ADD!!! fastdeploy executable: [${TEMP_TARGET_NAME}] !")
-  else ()
-    message(WARNING "Can not found source file: [${field}/${url}_${model}.cc], SKIP!!! fastdeploy executable: [${TEMP_TARGET_NAME}] !")          
+    message(STATUS "  Added FastDeploy Executable       : ${TEMP_TARGET_NAME}")
   endif()
   unset(TEMP_TARGET_FILE)
   unset(TEMP_TARGET_NAME)
 endfunction()
 
 # vision examples
-if (WITH_VISION_EXAMPLES)
-  add_fastdeploy_executable(vision ultralytics yolov5)
-  add_fastdeploy_executable(vision meituan yolov6)
-  add_fastdeploy_executable(vision wongkinyiu yolov7)
-  add_fastdeploy_executable(vision megvii yolox)
+if(WITH_VISION_EXAMPLES AND EXISTS ${PROJECT_SOURCE_DIR}/examples/vision)
+  message(STATUS "")
+  message(STATUS "*************FastDeploy Examples Summary**********")
+  file(GLOB ALL_VISION_EXAMPLE_SRCS ${PROJECT_SOURCE_DIR}/examples/vision/*.cc)
+  foreach(_CC_FILE ${ALL_VISION_EXAMPLE_SRCS})
+    add_fastdeploy_executable(vision ${_CC_FILE})
+  endforeach()
 endif()
 
-# other examples ...
+# other examples ...
diff --git a/examples/vision/ppdet_ppyoloe.cc b/examples/vision/ppdet_ppyoloe.cc
@@ -0,0 +1,51 @@
+// Copyright (c) 2022 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.
+
+#include "fastdeploy/vision.h"
+
+int main() {
+  namespace vis = fastdeploy::vision;
+
+  std::string model_file = "ppyoloe_crn_l_300e_coco/model.pdmodel";
+  std::string params_file = "ppyoloe_crn_l_300e_coco/model.pdiparams";
+  std::string config_file = "ppyoloe_crn_l_300e_coco/infer_cfg.yml";
+  std::string img_path = "test.jpeg";
+  std::string vis_path = "vis.jpeg";
+
+  auto model = vis::ppdet::PPYOLOE(model_file, params_file, config_file);
+  if (!model.Initialized()) {
+    std::cerr << "Init Failed." << std::endl;
+    return -1;
+  }
+
+  cv::Mat im = cv::imread(img_path);
+  cv::Mat vis_im = im.clone();
+
+  vis::DetectionResult res;
+  if (!model.Predict(&im, &res)) {
+    std::cerr << "Prediction Failed." << std::endl;
+    return -1;
+  } else {
+    std::cout << "Prediction Done!" << std::endl;
+  }
+
+  // 输出预测框结果
+  std::cout << res.Str() << std::endl;
+
+  // 可视化预测结果
+  vis::Visualize::VisDetection(&vis_im, res);
+  cv::imwrite(vis_path, vis_im);
+  std::cout << "Detect Done! Saved: " << vis_path << std::endl;
+  return 0;
+}
diff --git a/fastdeploy/__init__.py b/fastdeploy/__init__.py
@@ -17,7 +17,7 @@
 from .fastdeploy_runtime import *
 from . import fastdeploy_main as C
 from . import vision
-from .download import download
+from .download import download, download_and_decompress
 
 
 def TensorInfoStr(tensor_info):

diff --git a/fastdeploy/download.py b/fastdeploy/download.py
@@ -156,7 +156,7 @@ def decompress(fname):
 
 def url2dir(url, path, rename=None):
     full_name = download(url, path, rename, show_progress=True)
-    print("SDK is donwloaded, now extracting...")
+    print("File is donwloaded, now extracting...")
     if url.count(".tgz") > 0 or url.count(".tar") > 0 or url.count("zip") > 0:
         return decompress(full_name)
 

diff --git a/fastdeploy/utils/utils.h b/fastdeploy/utils/utils.h
@@ -64,6 +64,10 @@ class FASTDEPLOY_DECL FDLogger {
   bool verbose_ = true;
 };
 
+#ifndef __REL_FILE__
+#define __REL_FILE__ __FILE__
+#endif
+
 #define FDERROR                                                                \
   FDLogger(true, "[ERROR]")                                                    \
       << __REL_FILE__ << "(" << __LINE__ << ")::" << __FUNCTION__ << "\t"

diff --git a/fastdeploy/vision.h b/fastdeploy/vision.h
@@ -16,6 +16,7 @@
 #include "fastdeploy/core/config.h"
 #ifdef ENABLE_VISION
 #include "fastdeploy/vision/ppcls/model.h"
+#include "fastdeploy/vision/ppdet/ppyoloe.h"
 #include "fastdeploy/vision/ultralytics/yolov5.h"
 #include "fastdeploy/vision/wongkinyiu/yolov7.h"
 #include "fastdeploy/vision/meituan/yolov6.h"

diff --git a/fastdeploy/vision/__init__.py b/fastdeploy/vision/__init__.py
@@ -15,6 +15,7 @@
 
 from . import evaluation
 from . import ppcls
+from . import ppdet
 from . import ultralytics
 from . import meituan
 from . import megvii

diff --git a/fastdeploy/vision/common/processors/convert.cc b/fastdeploy/vision/common/processors/convert.cc
@@ -0,0 +1,62 @@
+// Copyright (c) 2022 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.
+
+#include "fastdeploy/vision/common/processors/convert.h"
+
+namespace fastdeploy {
+
+namespace vision {
+
+Convert::Convert(const std::vector<float>& alpha,
+                 const std::vector<float>& beta) {
+  FDASSERT(alpha.size() == beta.size(),
+           "Convert: requires the size of alpha equal to the size of beta.");
+  FDASSERT(alpha.size() != 0,
+           "Convert: requires the size of alpha and beta > 0.");
+  alpha_.assign(alpha.begin(), alpha.end());
+  beta_.assign(beta.begin(), beta.end());
+}
+
+bool Convert::CpuRun(Mat* mat) {
+  cv::Mat* im = mat->GetCpuMat();
+  std::vector<cv::Mat> split_im;
+  cv::split(*im, split_im);
+  for (int c = 0; c < im->channels(); c++) {
+    split_im[c].convertTo(split_im[c], CV_32FC1, alpha_[c], beta_[c]);
+  }
+  cv::merge(split_im, *im);
+  return true;
+}
+
+#ifdef ENABLE_OPENCV_CUDA
+bool Convert::GpuRun(Mat* mat) {
+  cv::cuda::GpuMat* im = mat->GetGpuMat();
+  std::vector<cv::cuda::GpuMat> split_im;
+  cv::cuda::split(*im, split_im);
+  for (int c = 0; c < im->channels(); c++) {
+    split_im[c].convertTo(split_im[c], CV_32FC1, alpha_[c], beta_[c]);
+  }
+  cv::cuda::merge(split_im, *im);
+  return true;
+}
+#endif
+
+bool Convert::Run(Mat* mat, const std::vector<float>& alpha,
+                  const std::vector<float>& beta, ProcLib lib) {
+  auto c = Convert(alpha, beta);
+  return c(mat, lib);
+}
+
+}  // namespace vision
+}  // namespace fastdeploy
diff --git a/fastdeploy/vision/common/processors/convert.h b/fastdeploy/vision/common/processors/convert.h
@@ -0,0 +1,42 @@
+// Copyright (c) 2022 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.
+
+#pragma once
+
+#include "fastdeploy/vision/common/processors/base.h"
+
+namespace fastdeploy {
+namespace vision {
+class Convert : public Processor {
+ public:
+  Convert(const std::vector<float>& alpha, const std::vector<float>& beta);
+
+  bool CpuRun(Mat* mat);
+#ifdef ENABLE_OPENCV_CUDA
+  bool GpuRun(Mat* mat);
+#endif
+  std::string Name() { return "Convert"; }
+
+  // Compute `result = mat * alpha + beta` directly by channel.
+  // The default behavior is the same as OpenCV's convertTo method.
+  static bool Run(Mat* mat, const std::vector<float>& alpha,
+                  const std::vector<float>& beta,
+                  ProcLib lib = ProcLib::OPENCV_CPU);
+
+ private:
+  std::vector<float> alpha_;
+  std::vector<float> beta_;
+};
+}  // namespace vision
+}  // namespace fastdeploy
diff --git a/fastdeploy/vision/common/processors/transform.h b/fastdeploy/vision/common/processors/transform.h
@@ -17,6 +17,7 @@
 #include "fastdeploy/vision/common/processors/cast.h"
 #include "fastdeploy/vision/common/processors/center_crop.h"
 #include "fastdeploy/vision/common/processors/color_space_convert.h"
+#include "fastdeploy/vision/common/processors/convert.h"
 #include "fastdeploy/vision/common/processors/hwc2chw.h"
 #include "fastdeploy/vision/common/processors/normalize.h"
 #include "fastdeploy/vision/common/processors/pad.h"

diff --git a/fastdeploy/vision/meituan/yolov6.cc b/fastdeploy/vision/meituan/yolov6.cc
@@ -25,14 +25,14 @@ namespace meituan {
 void LetterBox(Mat* mat, std::vector<int> size, std::vector<float> color,
                bool _auto, bool scale_fill = false, bool scale_up = true,
                int stride = 32) {
-  float scale = std::min(size[1] * 1.0f / static_cast<float>(mat->Height()), 
-                         size[0] * 1.0f / static_cast<float>(mat->Width()));       
+  float scale = std::min(size[1] * 1.0f / static_cast<float>(mat->Height()),
+                         size[0] * 1.0f / static_cast<float>(mat->Width()));
   if (!scale_up) {
     scale = std::min(scale, 1.0f);
   }
 
   int resize_h = int(round(static_cast<float>(mat->Height()) * scale));
-  int resize_w = int(round(static_cast<float>(mat->Width())  * scale));
+  int resize_w = int(round(static_cast<float>(mat->Width()) * scale));
 
   int pad_w = size[0] - resize_w;
   int pad_h = size[1] - resize_h;
@@ -85,13 +85,13 @@ bool YOLOv6::Initialize() {
   is_scale_up = false;
   stride = 32;
   max_wh = 4096.0f;
-  
+
   if (!InitRuntime()) {
     FDERROR << "Failed to initialize fastdeploy backend." << std::endl;
     return false;
   }
-  // Check if the input shape is dynamic after Runtime already initialized, 
-  // Note that, We need to force is_mini_pad 'false' to keep static 
+  // Check if the input shape is dynamic after Runtime already initialized,
+  // Note that, We need to force is_mini_pad 'false' to keep static
   // shape after padding (LetterBox) when the is_dynamic_shape is 'false'.
   is_dynamic_input_ = false;
   auto shape = InputInfoOfRuntime(0).shape;
@@ -102,7 +102,7 @@ bool YOLOv6::Initialize() {
       break;
     }
   }
-  if (!is_dynamic_input_) {  
+  if (!is_dynamic_input_) {
     is_mini_pad = false;
   }
   return true;
@@ -111,15 +111,15 @@ bool YOLOv6::Initialize() {
 bool YOLOv6::Preprocess(Mat* mat, FDTensor* output,
                         std::map<std::string, std::array<float, 2>>* im_info) {
   // process after image load
-  float ratio = std::min(size[1] * 1.0f / static_cast<float>(mat->Height()), 
-                         size[0] * 1.0f / static_cast<float>(mat->Width()));                                          
+  float ratio = std::min(size[1] * 1.0f / static_cast<float>(mat->Height()),
+                         size[0] * 1.0f / static_cast<float>(mat->Width()));
   if (ratio != 1.0) {
     int interp = cv::INTER_AREA;
     if (ratio > 1.0) {
       interp = cv::INTER_LINEAR;
     }
     int resize_h = int(round(static_cast<float>(mat->Height()) * ratio));
-    int resize_w = int(round(static_cast<float>(mat->Width())  * ratio));
+    int resize_w = int(round(static_cast<float>(mat->Width()) * ratio));
     Resize::Run(mat, resize_w, resize_h, -1, -1, interp);
   }
   // yolov6's preprocess steps
@@ -129,8 +129,12 @@ bool YOLOv6::Preprocess(Mat* mat, FDTensor* output,
   LetterBox(mat, size, padding_value, is_mini_pad, is_no_pad, is_scale_up,
             stride);
   BGR2RGB::Run(mat);
-  Normalize::Run(mat, std::vector<float>(mat->Channels(), 0.0),
-                 std::vector<float>(mat->Channels(), 1.0));
+  // Normalize::Run(mat, std::vector<float>(mat->Channels(), 0.0),
+  //                std::vector<float>(mat->Channels(), 1.0));
+  // Compute `result = mat * alpha + beta` directly by channel
+  std::vector<float> alpha = {1.0f / 255.0f, 1.0f / 255.0f, 1.0f / 255.0f};
+  std::vector<float> beta = {0.0f, 0.0f, 0.0f};
+  Convert::Run(mat, alpha, beta);
 
   // Record output shape of preprocessed image
   (*im_info)["output_shape"] = {static_cast<float>(mat->Height()),

diff --git a/fastdeploy/vision/ppcls/model.cc b/fastdeploy/vision/ppcls/model.cc
@@ -1,3 +1,16 @@
+// Copyright (c) 2022 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.
 
 #include "fastdeploy/vision/ppcls/model.h"
 #include "fastdeploy/vision/utils/utils.h"
@@ -135,6 +148,6 @@ bool Model::Predict(cv::Mat* im, ClassifyResult* result, int topk) {
   return true;
 }
 
-}  // namespace ppcls
-}  // namespace vision
-}  // namespace fastdeploy
+} // namespace ppcls
+} // namespace vision
+} // namespace fastdeploy
diff --git a/fastdeploy/vision/ppcls/model.h b/fastdeploy/vision/ppcls/model.h
@@ -1,7 +1,21 @@
+// Copyright (c) 2022 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.
+
 #pragma once
 #include "fastdeploy/fastdeploy_model.h"
-#include "fastdeploy/vision/common/result.h"
 #include "fastdeploy/vision/common/processors/transform.h"
+#include "fastdeploy/vision/common/result.h"
 
 namespace fastdeploy {
 namespace vision {

diff --git a/fastdeploy/vision/ppcls/ppcls_pybind.cc b/fastdeploy/vision/ppcls/ppcls_pybind.cc
@@ -14,7 +14,7 @@
 #include "fastdeploy/pybind/main.h"
 
 namespace fastdeploy {
-void BindPpClsModel(pybind11::module& m) {
+void BindPPCls(pybind11::module& m) {
   auto ppcls_module = m.def_submodule("ppcls", "Module to deploy PaddleClas.");
   pybind11::class_<vision::ppcls::Model, FastDeployModel>(ppcls_module, "Model")
       .def(pybind11::init<std::string, std::string, std::string, RuntimeOption,