Removed boost and caffe deps from headers

Author: gineshidalgo99

.github/issue_template.md

@@ -48,6 +48,7 @@ You might select multiple topics, delete the rest:
 3. **General configuration**:
     - **Installation mode**: CMake, sh script, manual Makefile installation, ... (Ubuntu); CMake, ... (Windows); ...?
     - **Operating system** (`lsb_release -a` in Ubuntu):
+    - **Operating system version** (e.g., Ubuntu 16, Windows 10, ...):
     - **Release or Debug mode**? (by default: release):
     - Compiler (`gcc --version` in Ubuntu or VS version in Windows): 5.4.0, ... (Ubuntu); VS2015 Enterprise Update 3, VS2017 community, ... (Windows); ...?
 

README.md

@@ -4,8 +4,8 @@
 
 -----------------
 
-|                 | `Python (CUDA GPU)` | `Python (CPU)` | `CUDA GPU` | `CPU`  | `Debug mode` |
-| :---:           | :---:               | :---:          | :---:      |:---:   | :---:        |
+|                 | `Python (CUDA GPU)` | `Python (CPU)` | `CUDA GPU` | `CPU` | `Debug mode` |
+| :---:           | :---:               | :---:          | :---:      |:---:  | :---:        |
 | **`Linux`**     | [![Status](https://travis-matrix-badges.herokuapp.com/repos/CMU-Perceptual-Computing-Lab/openpose/branches/master/1)](https://travis-ci.org/CMU-Perceptual-Computing-Lab/openpose) | [![Status](https://travis-matrix-badges.herokuapp.com/repos/CMU-Perceptual-Computing-Lab/openpose/branches/master/2)](https://travis-ci.org/CMU-Perceptual-Computing-Lab/openpose) | [![Status](https://travis-matrix-badges.herokuapp.com/repos/CMU-Perceptual-Computing-Lab/openpose/branches/master/3)](https://travis-ci.org/CMU-Perceptual-Computing-Lab/openpose) | [![Status](https://travis-matrix-badges.herokuapp.com/repos/CMU-Perceptual-Computing-Lab/openpose/branches/master/4)](https://travis-ci.org/CMU-Perceptual-Computing-Lab/openpose) | [![Status](https://travis-matrix-badges.herokuapp.com/repos/CMU-Perceptual-Computing-Lab/openpose/branches/master/5)](https://travis-ci.org/CMU-Perceptual-Computing-Lab/openpose) |
 | **`MacOS`**     | | [![Status](https://travis-matrix-badges.herokuapp.com/repos/CMU-Perceptual-Computing-Lab/openpose/branches/master/6)](https://travis-ci.org/CMU-Perceptual-Computing-Lab/openpose) | | [![Status](https://travis-matrix-badges.herokuapp.com/repos/CMU-Perceptual-Computing-Lab/openpose/branches/master/7)](https://travis-ci.org/CMU-Perceptual-Computing-Lab/openpose) | [![Status](https://travis-matrix-badges.herokuapp.com/repos/CMU-Perceptual-Computing-Lab/openpose/branches/master/8)](https://travis-ci.org/CMU-Perceptual-Computing-Lab/openpose) |
 <!-- | **`Windows`**   | | | | | | -->
@@ -46,7 +46,7 @@ It is **authored by [Gines Hidalgo](https://www.gineshidalgo.com), [Zhe Cao](htt
     - Available: command-line demo, C++ wrapper, and C++ API.
     - [**Python API**](doc/modules/python_module.md).
     - [**Unity Plugin**](https://github.com/CMU-Perceptual-Computing-Lab/openpose_unity_plugin).
-    - CUDA (Nvidia GPU), OpenCL (AMD GPU), and CPU versions.
+    - CUDA (Nvidia GPU), OpenCL (AMD GPU), and CPU-only (no GPU) versions.
     - Training code included in the [**original CVPR 2017 GitHub repository**](https://github.com/ZheC/Multi-Person-Pose-Estimation).
 
 

doc/installation.md

@@ -48,14 +48,14 @@ This installation section is only intended if you plan to modify the OpenPose co
 
 ## Requirements and Dependencies
 - **Requirements** for the default configuration (you might need more resources with a greater `--net_resolution` and/or `scale_number` or less resources by reducing the net resolution and/or using the MPI and MPI_4 models):
-    - Nvidia GPU version:
+    - CUDA (Nvidia GPU) version:
         - NVIDIA graphics card with at least 1.6 GB available (the `nvidia-smi` command checks the available GPU memory in Ubuntu).
         - At least 2.5 GB of free RAM memory for BODY_25 model or 2 GB for COCO model (assuming cuDNN installed).
         - Highly recommended: cuDNN.
-    - AMD GPU version:
+    - OpenCL (AMD GPU) version:
         - Vega series graphics card
         - At least 2 GB of free RAM memory.
-    - CPU version:
+    - CPU-only (no GPU) version:
         - Around 8GB of free RAM memory.
     - Highly recommended: a CPU with at least 8 cores.
 - **Dependencies**:

doc/modules/python_module.md

@@ -18,7 +18,7 @@ This module exposes a Python API for OpenPose. It is effectively a wrapper that
 ## Compatibility
 The OpenPose Python module is compatible with both Python 2 and Python 3 (default and recommended). In addition, it will also run in all OpenPose compatible operating systems. It uses [Pybind11](https://github.com/pybind/pybind11) for mapping between C++ and Python datatypes.
 
-To compile, enable `BUILD_PYTHON` in CMake-gui. In Windows, make sure you compile the whole solution (clicking the green play button does not compile the whole solution!). You can do that by right-click on the OpenPose project solution, and clicking in `Build solution`.
+To compile, enable `BUILD_PYTHON` in CMake-gui. In Windows, make sure you compile the whole solution (clicking the green play button does not compile the whole solution!). You can do that by right-click on the OpenPose project solution, and clicking in `Build Solution` (or individually building the PyOpenPose module).
 
 Pybind selects the latest version of Python by default (Python 3). To use Python 2, change `PYTHON_EXECUTABLE` and `PYTHON_LIBRARY` flags in CMake-gui to your desired Python version.
 

doc/release_notes.md

@@ -316,6 +316,7 @@ OpenPose Library - Release Notes
     33. Added the flags `--face_detector` and `--hand_detector`, that enable the user to select the face/hand rectangle detector that is used for the later face/hand keypoint detection. It includes OpenCV (for face), and also allows the user to provide its own input. Flag `--hand_tracking` is removed and integrated into this flag too.
     34. Maximum queue size per OpenPose thread is configurable through the Wrapper class.
     35. Added pre-processing capabilities to Wrapper (WorkerType::PreProcessing), which will be run right after the image has been read.
+    36. Removed boost::shared_ptr and caffe::Blob dependencies from the headers. No 3rdparty dependencies left on headers (except dim3 for CUDA).
 2. Functions or parameters renamed:
     1. By default, python example `tutorial_developer/python_2_pose_from_heatmaps.py` was using 2 scales starting at -1x736, changed to 1 scale at -1x368.
     2. WrapperStructPose default parameters changed to match those of the OpenPose demo binary.

doc/released_features.md

@@ -1,8 +1,8 @@
 OpenPose Library - All Released Features
 ====================================
 
-- Jan 2018: [**Unity plugin released**](https://github.com/CMU-Perceptual-Computing-Lab/openpose_unity_plugin)!
-- Jan 2018: [**Improved Python API**](doc/modules/python_module.md) released! Including body, face, hands, and all the functionality of the C++ API!
+- Jan 2019: [**Unity plugin released**](https://github.com/CMU-Perceptual-Computing-Lab/openpose_unity_plugin)!
+- Jan 2019: [**Improved Python API**](doc/modules/python_module.md) released! Including body, face, hands, and all the functionality of the C++ API!
 - Dec 2018: [**Foot dataset**](https://cmu-perceptual-computing-lab.github.io/foot_keypoint_dataset) and [**new paper released**](https://arxiv.org/abs/1812.08008)!
 - Sep 2018: [**Experimental tracker**](./quick_start.md#tracking)!
 - Jun 2018: [**Combined body-foot model released! 40% faster and 5% more accurate**](./installation.md)!

examples/tests/clTest.cpp

@@ -5,6 +5,10 @@
 #include <openpose/flags.hpp>
 // OpenPose dependencies
 #include <openpose/headers.hpp>
+// Caffe dependencies
+#ifdef USE_CAFFE
+    #include <caffe/blob.hpp>
+#endif
 // OpenCL dependencies
 #ifdef USE_OPENCL
 #include <openpose/gpu/opencl.hcl>
@@ -72,7 +76,8 @@ DEFINE_string(image_path,               "examples/media/COCO_val2014_00000000019
 
 typedef cl::KernelFunctor<cl::Buffer, int, int, float> ScaleFunctor;
 const std::string scaleKernelString = MULTI_LINE_STRING(
-            __kernel void scaleKernel(__global float* targetPtr, const int targetWidth, const int targetHeight, const float scale)
+            __kernel void scaleKernel(__global float* targetPtr, const int targetWidth, const int targetHeight,
+                                      const float scale)
 {
                 int x = get_global_id(0);
                 int y = get_global_id(1);
@@ -111,7 +116,8 @@ int clTest()
         std::unique_ptr<caffe::Net<float>> upCaffeNet;
         caffe::Caffe::set_mode(caffe::Caffe::GPU);
         caffe::Caffe::SelectDevice(0, true);
-        upCaffeNet.reset(new caffe::Net<float>{"models/pose/coco/pose_deploy_linevec.prototxt", caffe::TEST, caffe::Caffe::GetDefaultDevice()});
+        upCaffeNet.reset(new caffe::Net<float>{
+            "models/pose/coco/pose_deploy_linevec.prototxt", caffe::TEST, caffe::Caffe::GetDefaultDevice()});
         upCaffeNet->CopyTrainedLayersFrom("models/pose/coco/pose_iter_440000.caffemodel");
         op::OpenCL::getInstance(0, CL_DEVICE_TYPE_GPU, true);
 
@@ -125,19 +131,16 @@ int clTest()
         blob_proto.set_height(imgResize.size().width);
         blob_proto.set_width(imgResize.size().height);
         blob_proto.clear_data();
-        for (int c = 0; c < 3; ++c) {
-            for (int h = 0; h < imgResize.size().height; ++h) {
-                for (int w = 0; w < imgResize.size().width; ++w) {
+        for (int c = 0; c < 3; ++c)
+            for (int h = 0; h < imgResize.size().height; ++h)
+                for (int w = 0; w < imgResize.size().width; ++w)
                     blob_proto.add_data(imgResize.at<cv::Vec3f>(h, w)[c]);
-                }
-            }
-        }
         blob_proto.set_num(1);
         caffe::Blob<float>* input_layer = upCaffeNet->input_blobs()[0];
         input_layer->FromProto(blob_proto);
         upCaffeNet->Forward(0);
 
-        boost::shared_ptr<caffe::Blob<float> > output_blob = upCaffeNet->blob_by_name("net_output");
+        boost::shared_ptr<caffe::Blob<float>> output_blob = upCaffeNet->blob_by_name("net_output");
 
         // Test
         cl::Device& device = op::OpenCL::getInstance(0)->getDevice();
@@ -156,8 +159,9 @@ int clTest()
             // Read it
             // Read back image to GPU
             float* heatmaps = new float[output_blob->shape()[1] * output_blob->shape()[2] * output_blob->shape()[3]];
-            op::OpenCL::getInstance(0)->getQueue().enqueueReadBuffer(outputBuffer, CL_TRUE, 0,
-                                                                     output_blob->shape()[1] * output_blob->shape()[2] * output_blob->shape()[3] * sizeof(float), heatmaps);
+            op::OpenCL::getInstance(0)->getQueue().enqueueReadBuffer(
+                outputBuffer, CL_TRUE, 0,
+                output_blob->shape()[1] * output_blob->shape()[2] * output_blob->shape()[3] * sizeof(float), heatmaps);
 
             int heatmapChannels = output_blob->shape()[1];
             int shape = output_blob->shape()[2] * output_blob->shape()[3];
@@ -172,11 +176,11 @@ int clTest()
             }
         }
         #if defined(USE_OPENCL) && defined(CL_HPP_ENABLE_EXCEPTIONS)
-        catch (const cl::Error& e)
-        {
-            op::error(std::string(e.what()) + " : " + op::OpenCL::clErrorToString(e.err()) + " ID: " +
-                      std::to_string(0), __LINE__, __FUNCTION__, __FILE__);
-        }
+            catch (const cl::Error& e)
+            {
+                op::error(std::string(e.what()) + " : " + op::OpenCL::clErrorToString(e.err()) + " ID: " +
+                          std::to_string(0), __LINE__, __FUNCTION__, __FILE__);
+            }
         #endif
         catch (const std::exception& e)
         {
@@ -223,6 +227,6 @@ int main()
 #else
     op::error("OpenPose must be compiled with the `USE_CAFFE` & `USE_OPENCL` macro definitions in order to run"
               " this functionality.", __LINE__, __FUNCTION__, __FILE__);
-    return 0;
+    return -1;
 #endif
 }

examples/tests/resizeTest.cpp

@@ -15,7 +15,7 @@
 
     cv::Mat gpuResize(cv::Mat& img, const cv::Size& newSize)
     {
-        #ifdef USE_CUDA
+        #if defined USE_CAFFE && defined USE_CUDA
             // Upload to Source to GPU
             float* cpuPtr = &img.at<float>(0);
             float* gpuPtr;
@@ -49,6 +49,7 @@
             UNUSED(newSize);
             op::error("OpenPose must be compiled with the `USE_CAFFE` & `USE_CUDA` macro definitions in order to run"
                   " this functionality.", __LINE__, __FUNCTION__, __FILE__);
+            return cv::Mat();
         #endif
     }
 

include/openpose/core/arrayCpuGpu.hpp

@@ -0,0 +1,100 @@
+#ifndef OPENPOSE_CORE_ARRAY_CPU_GPU_HPP
+#define OPENPOSE_CORE_ARRAY_CPU_GPU_HPP
+
+#include <memory> // std::shared_ptr
+#include <vector>
+#include <openpose/core/macros.hpp>
+
+namespace op
+{
+    /**
+     * ArrayCpuGpu<T>: Bind of caffe::Blob<T> to avoid Caffe as dependency in the headers. 
+     */
+    template<typename T>
+    class ArrayCpuGpu
+    {
+    public:
+        ArrayCpuGpu();
+        explicit ArrayCpuGpu(const void* caffeBlobTPtr);
+        explicit ArrayCpuGpu(const int num, const int channels, const int height, const int width);
+        // explicit ArrayCpuGpu(const std::vector<int>& shape);
+
+        void Reshape(const int num, const int channels, const int height, const int width);
+        void Reshape(const std::vector<int>& shape);
+        // // void Reshape(const BlobShape& shape);
+        // // void ReshapeLike(const Blob& other);
+        // void ReshapeLike(const ArrayCpuGpu& other);
+        std::string shape_string() const;
+        const std::vector<int>& shape() const;
+        int shape(const int index) const;
+        int num_axes() const;
+        int count() const;
+        int count(const int start_axis, const int end_axis) const;
+        int count(const int start_axis) const;
+
+        int CanonicalAxisIndex(const int axis_index) const;
+
+        // int num() const;
+        // int channels() const;
+        // int height() const;
+        // int width() const;
+        // int LegacyShape(const int index) const;
+
+        // int offset(const int n, const int c = 0, const int h = 0, const int w = 0) const;
+        // int offset(const std::vector<int>& indices) const;
+
+        // // void CopyFrom(const Blob<T>& source, bool copy_diff = false, bool reshape = false);
+        // void CopyFrom(const ArrayCpuGpu<T>& source, bool copy_diff = false, bool reshape = false);
+
+        T data_at(const int n, const int c, const int h, const int w) const;
+        T diff_at(const int n, const int c, const int h, const int w) const;
+        // T data_at(const std::vector<int>& index) const; // Caffe warning
+        // T diff_at(const std::vector<int>& index) const;
+
+        // // const boost::shared_ptr<SyncedMemory>& data() const;
+        // // const boost::shared_ptr<SyncedMemory>& diff() const;
+
+        const T* cpu_data() const;
+        void set_cpu_data(T* data);
+        const int* gpu_shape() const;
+        const T* gpu_data() const;
+        void set_gpu_data(T* data);
+        const T* cpu_diff() const;
+        const T* gpu_diff() const;
+        T* mutable_cpu_data();
+        T* mutable_gpu_data();
+        T* mutable_cpu_diff();
+        T* mutable_gpu_diff();
+        void Update();
+        // void FromProto(const BlobProto& proto, bool reshape = true);
+        // void ToProto(BlobProto* proto, bool write_diff = false) const;
+
+        // T asum_data() const;
+        // T asum_diff() const;
+        // T sumsq_data() const;
+        // T sumsq_diff() const;
+
+        // void scale_data(T scale_factor);
+        // void scale_diff(T scale_factor);
+
+        // // void ShareData(const Blob& other);
+        // // void ShareDiff(const Blob& other);
+
+        // // bool ShapeEquals(const BlobProto& other);
+
+    private:
+        // PIMPL idiom
+        // http://www.cppsamples.com/common-tasks/pimpl.html
+        struct ImplArrayCpuGpu;
+        std::shared_ptr<ImplArrayCpuGpu> spImpl;
+
+        // PIMP requires DELETE_COPY & destructor, or extra code
+        // http://oliora.github.io/2015/12/29/pimpl-and-rule-of-zero.html
+        DELETE_COPY(ArrayCpuGpu);
+    };
+
+    // // Static methods
+    // OVERLOAD_C_OUT(ArrayCpuGpu)
+}
+
+#endif // OPENPOSE_CORE_ARRAY_CPU_GPU_HPP

include/openpose/core/common.hpp

@@ -8,6 +8,7 @@
 #include <vector>
 // OpenPose most used classes
 #include <openpose/core/array.hpp>
+#include <openpose/core/arrayCpuGpu.hpp>
 #include <openpose/core/macros.hpp>
 #include <openpose/core/point.hpp>
 #include <openpose/core/rectangle.hpp>