Auto aspect-ratio selection

doc/demo_overview.md

@@ -142,7 +142,7 @@ Each flag is divided into flag name, default value, and description.
 
 4. OpenPose Body Pose
 - DEFINE_string(model_pose,               "COCO",         "Model to be used. E.g. `COCO` (18 keypoints), `MPI` (15 keypoints, ~10% faster), `MPI_4_layers` (15 keypoints, even faster but less accurate).");
-- DEFINE_string(net_resolution,           "656x368",      "Multiples of 16. If it is increased, the accuracy usually increases. If it is decreased, the speed increases. For maximum speed-accuracy balance, it should keep the closest aspect ratio possible to the images or videos to be processed. E.g. the default `656x368` is optimal for 16:9 videos, e.g. full HD (1980x1080) and HD (1280x720) videos.");
+- DEFINE_string(net_resolution,           "656x368",      "Multiples of 16. If it is increased, the accuracy potentially increases. If it is decreased, the speed increases. For maximum speed-accuracy balance, it should keep the closest aspect ratio possible to the images or videos to be processed. Using `-1` in any of the dimensions, OP will choose the optimal resolution depending on the other value introduced by the user. E.g. the default `-1x368` is equivalent to `656x368` in 16:9 videos, e.g. full HD (1980x1080) and HD (1280x720) resolutions.");
 - DEFINE_int32(scale_number,              1,              "Number of scales to average.");
 - DEFINE_double(scale_gap,                0.3,            "Scale gap between scales. No effect unless scale_number > 1. Initial scale is always 1. If you want to change the initial scale, you actually want to multiply the `net_resolution` by your desired initial scale.");
 - DEFINE_bool(heatmaps_add_parts,         false,          "If true, it will add the body part heatmaps to the final op::Datum::poseHeatMaps array (program speed will decrease). Not required for our library, enable it only if you intend to process this information later. If more than one `add_heatmaps_X` flag is enabled, it will place then in sequential memory order: body parts + bkg + PAFs. It will follow the order on POSE_BODY_PART_MAPPING in `include/openpose/pose/poseParameters.hpp`.");

doc/release_notes.md

@@ -117,3 +117,4 @@ OpenPose Library - Release Notes
 1. Main improvements:
     1. COCO JSON file outputs 0 as score for non-detected keypoints.
     2. Added example for OpenPose for user asynchronous output and cleaned all `tutorial_wrapper/` examples.
+    3. Added `-1` option for `net_resolution` in order to auto-select the best possible aspect ratio given the user input.

examples/openpose/openpose.cpp

@@ -65,10 +65,12 @@ DEFINE_int32(keypoint_scale,            0,              "Scaling of the (x,y) co
 // OpenPose Body Pose
 DEFINE_string(model_pose,               "COCO",         "Model to be used. E.g. `COCO` (18 keypoints), `MPI` (15 keypoints, ~10% faster), "
                                                         "`MPI_4_layers` (15 keypoints, even faster but less accurate).");
-DEFINE_string(net_resolution,           "656x368",      "Multiples of 16. If it is increased, the accuracy potentially increases. If it is decreased,"
-                                                        " the speed increases. For maximum speed-accuracy balance, it should keep the closest aspect"
-                                                        " ratio possible to the images or videos to be processed. E.g. the default `656x368` is"
-                                                        " optimal for 16:9 videos, e.g. full HD (1980x1080) and HD (1280x720) videos.");
+DEFINE_string(net_resolution,           "656x368",      "Multiples of 16. If it is increased, the accuracy potentially increases. If it is"
+                                                        " decreased, the speed increases. For maximum speed-accuracy balance, it should keep the"
+                                                        " closest aspect ratio possible to the images or videos to be processed. Using `-1` in"
+                                                        " any of the dimensions, OP will choose the optimal aspect ratio depending on the user's"
+                                                        " input value. E.g. the default `-1x368` is equivalent to `656x368` in 16:9 resolutions,"
+                                                        " e.g. full HD (1980x1080) and HD (1280x720) resolutions.");
 DEFINE_int32(scale_number,              1,              "Number of scales to average.");
 DEFINE_double(scale_gap,                0.3,            "Scale gap between scales. No effect unless scale_number > 1. Initial scale is always 1."
                                                         " If you want to change the initial scale, you actually want to multiply the"
@@ -172,7 +174,7 @@ int openPoseDemo()
     // outputSize
     const auto outputSize = op::flagsToPoint(FLAGS_resolution, "1280x720");
     // netInputSize
-    const auto netInputSize = op::flagsToPoint(FLAGS_net_resolution, "656x368");
+    const auto netInputSize = op::flagsToPoint(FLAGS_net_resolution, "-1x368");
     // faceNetInputSize
     const auto faceNetInputSize = op::flagsToPoint(FLAGS_face_net_resolution, "368x368 (multiples of 16)");
     // handNetInputSize

examples/openpose/openposeBasicDemo.cpp

+// ------------------------- OpenPose Library Tutorial - Real Time Pose Estimation -------------------------
+// C++ std library dependencies
+#include <chrono> // `std::chrono::` functions and classes, e.g. std::chrono::milliseconds
+#include <thread> // std::this_thread
+// Other 3rdparty dependencies
+#include <gflags/gflags.h> // DEFINE_bool, DEFINE_int32, DEFINE_int64, DEFINE_uint64, DEFINE_double, DEFINE_string
+#include <glog/logging.h> // google::InitGoogleLogging
+
+// OpenPose dependencies
+#include <openpose/headers.hpp>
+
+// See all the available parameter options withe the `--help` flag. E.g. `./build/examples/openpose/openpose.bin --help`.
+// Note: This command will show you flags for other unnecessary 3rdparty files. Check only the flags for the OpenPose
+// executable. E.g. for `openpose.bin`, look for `Flags from examples/openpose/openpose.cpp:`.
+// Producer
+DEFINE_int32(camera,                    -1,             "The camera index for cv::VideoCapture. Integer in the range [0, 9]. Select a negative"
+                                                        " number (by default), to auto-detect and open the first available camera.");
+DEFINE_string(camera_resolution,        "1280x720",     "Size of the camera frames to ask for.");
+DEFINE_double(camera_fps,               30.0,           "Frame rate for the webcam (only used when saving video from webcam). Set this value to the"
+                                                        " minimum value between the OpenPose displayed speed and the webcam real frame rate.");
+DEFINE_string(video,                    "",             "Use a video file instead of the camera. Use `examples/media/video.avi` for our default"
+                                                        " example video.");
+DEFINE_string(image_dir,                "",             "Process a directory of images. Use `examples/media/` for our default example folder with 20"
+                                                        " images. Read all standard formats (jpg, png, bmp, etc.).");
+// Display
+DEFINE_bool(no_gui_verbose,             false,          "Do not write text on output images on GUI (e.g. number of current frame and people). It"
+                                                        " does not affect the pose rendering.");
+DEFINE_bool(no_display,                 false,          "Do not open a display window. Useful if there is no X server and/or to slightly speed up"
+                                                        " the processing if visual output is not required.");
+
+int openPoseDemo()
+{
+    op::log("Starting pose estimation demo.", op::Priority::High);
+    const auto timerBegin = std::chrono::high_resolution_clock::now();
+
+    // Applying user defined configuration - Google flags to program variables
+    const auto producerSharedPtr = op::flagsToProducer(FLAGS_image_dir, FLAGS_video, FLAGS_camera, FLAGS_camera_resolution, FLAGS_camera_fps);
+    op::log("", op::Priority::Low, __LINE__, __FUNCTION__, __FILE__);
+
+    // OpenPose wrapper
+    op::log("Configuring OpenPose wrapper.", op::Priority::Low, __LINE__, __FUNCTION__, __FILE__);
+    op::Wrapper<std::vector<op::Datum>> opWrapper;
+    // Pose configuration (use WrapperStructPose{} for default and recommended configuration)
+    op::WrapperStructPose wrapperStructPose{};
+    wrapperStructPose.renderMode = op::RenderMode::Gpu;
+    // Producer (use default to disable any input)
+    const op::WrapperStructInput wrapperStructInput{producerSharedPtr};
+    // Consumer (comment or use default argument to disable any output)
+    const op::WrapperStructOutput wrapperStructOutput{!FLAGS_no_display, !FLAGS_no_gui_verbose};
+    // Configure wrapper
+    opWrapper.configure(wrapperStructPose, wrapperStructInput, wrapperStructOutput);
+    // Set to single-thread running (e.g. for debugging purposes)
+    // opWrapper.disableMultiThreading();
+
+    // Start processing
+    // Two different ways of running the program on multithread environment
+    op::log("Starting thread(s)", op::Priority::High);
+    opWrapper.exec();  // It blocks this thread until all threads have finished
+
+    // Measuring total time
+    const auto now = std::chrono::high_resolution_clock::now();
+    const auto totalTimeSec = (double)std::chrono::duration_cast<std::chrono::nanoseconds>(now-timerBegin).count() * 1e-9;
+    const auto message = "Real-time pose estimation demo successfully finished. Total time: " + std::to_string(totalTimeSec) + " seconds.";
+    op::log(message, op::Priority::High);
+
+    return 0;
+}
+
+int main(int argc, char *argv[])
+{
+    // Initializing google logging (Caffe uses it for logging)
+    google::InitGoogleLogging("openPoseDemo");
+
+    // Parsing command line flags
+    gflags::ParseCommandLineFlags(&argc, &argv, true);
+
+    // Running openPoseDemo
+    return openPoseDemo();
+}

examples/tutorial_pose/1_extract_from_image.cpp

@@ -31,10 +31,12 @@ DEFINE_string(image_path,               "examples/media/COCO_val2014_00000000019
 DEFINE_string(model_pose,               "COCO",         "Model to be used. E.g. `COCO` (18 keypoints), `MPI` (15 keypoints, ~10% faster), "
                                                         "`MPI_4_layers` (15 keypoints, even faster but less accurate).");
 DEFINE_string(model_folder,             "models/",      "Folder path (absolute or relative) where the models (pose, face, ...) are located.");
-DEFINE_string(net_resolution,           "656x368",      "Multiples of 16. If it is increased, the accuracy potentially increases. If it is decreased,"
-                                                        " the speed increases. For maximum speed-accuracy balance, it should keep the closest aspect"
-                                                        " ratio possible to the images or videos to be processed. E.g. the default `656x368` is"
-                                                        " optimal for 16:9 videos, e.g. full HD (1980x1080) and HD (1280x720) videos.");
+DEFINE_string(net_resolution,           "656x368",      "Multiples of 16. If it is increased, the accuracy potentially increases. If it is"
+                                                        " decreased, the speed increases. For maximum speed-accuracy balance, it should keep the"
+                                                        " closest aspect ratio possible to the images or videos to be processed. Using `-1` in"
+                                                        " any of the dimensions, OP will choose the optimal aspect ratio depending on the user's"
+                                                        " input value. E.g. the default `-1x368` is equivalent to `656x368` in 16:9 resolutions,"
+                                                        " e.g. full HD (1980x1080) and HD (1280x720) resolutions.");
 DEFINE_string(resolution,               "1280x720",     "The image resolution (display and output). Use \"-1x-1\" to force the program to use the"
                                                         " default images resolution.");
 DEFINE_int32(num_gpu_start,             0,              "GPU device start number.");
@@ -67,7 +69,7 @@ int openPoseTutorialPose1()
     // outputSize
     const auto outputSize = op::flagsToPoint(FLAGS_resolution, "1280x720");
     // netInputSize
-    const auto netInputSize = op::flagsToPoint(FLAGS_net_resolution, "656x368");
+    const auto netInputSize = op::flagsToPoint(FLAGS_net_resolution, "-1x368");
     // netOutputSize
     const auto netOutputSize = netInputSize;
     // poseModel

examples/tutorial_pose/2_extract_pose_or_heatmat_from_image.cpp

@@ -31,10 +31,12 @@ DEFINE_string(image_path,               "examples/media/COCO_val2014_00000000019
 DEFINE_string(model_pose,               "COCO",         "Model to be used. E.g. `COCO` (18 keypoints), `MPI` (15 keypoints, ~10% faster), "
                                                         "`MPI_4_layers` (15 keypoints, even faster but less accurate).");
 DEFINE_string(model_folder,             "models/",      "Folder path (absolute or relative) where the models (pose, face, ...) are located.");
-DEFINE_string(net_resolution,           "656x368",      "Multiples of 16. If it is increased, the accuracy potentially increases. If it is decreased,"
-                                                        " the speed increases. For maximum speed-accuracy balance, it should keep the closest aspect"
-                                                        " ratio possible to the images or videos to be processed. E.g. the default `656x368` is"
-                                                        " optimal for 16:9 videos, e.g. full HD (1980x1080) and HD (1280x720) videos.");
+DEFINE_string(net_resolution,           "656x368",      "Multiples of 16. If it is increased, the accuracy potentially increases. If it is"
+                                                        " decreased, the speed increases. For maximum speed-accuracy balance, it should keep the"
+                                                        " closest aspect ratio possible to the images or videos to be processed. Using `-1` in"
+                                                        " any of the dimensions, OP will choose the optimal aspect ratio depending on the user's"
+                                                        " input value. E.g. the default `-1x368` is equivalent to `656x368` in 16:9 resolutions,"
+                                                        " e.g. full HD (1980x1080) and HD (1280x720) resolutions.");
 DEFINE_string(resolution,               "1280x720",     "The image resolution (display and output). Use \"-1x-1\" to force the program to use the"
                                                         " default images resolution.");
 DEFINE_int32(num_gpu_start,             0,              "GPU device start number.");
@@ -72,7 +74,7 @@ int openPoseTutorialPose2()
     // outputSize
     const auto outputSize = op::flagsToPoint(FLAGS_resolution, "1280x720");
     // netInputSize
-    const auto netInputSize = op::flagsToPoint(FLAGS_net_resolution, "656x368");
+    const auto netInputSize = op::flagsToPoint(FLAGS_net_resolution, "-1x368");
     // netOutputSize
     const auto netOutputSize = netInputSize;
     // poseModel

examples/tutorial_wrapper/1_user_asynchronous_output.cpp

@@ -65,10 +65,12 @@ DEFINE_int32(keypoint_scale,            0,              "Scaling of the (x,y) co
 // OpenPose Body Pose
 DEFINE_string(model_pose,               "COCO",         "Model to be used. E.g. `COCO` (18 keypoints), `MPI` (15 keypoints, ~10% faster), "
                                                         "`MPI_4_layers` (15 keypoints, even faster but less accurate).");
-DEFINE_string(net_resolution,           "656x368",      "Multiples of 16. If it is increased, the accuracy potentially increases. If it is decreased,"
-                                                        " the speed increases. For maximum speed-accuracy balance, it should keep the closest aspect"
-                                                        " ratio possible to the images or videos to be processed. E.g. the default `656x368` is"
-                                                        " optimal for 16:9 videos, e.g. full HD (1980x1080) and HD (1280x720) videos.");
+DEFINE_string(net_resolution,           "656x368",      "Multiples of 16. If it is increased, the accuracy potentially increases. If it is"
+                                                        " decreased, the speed increases. For maximum speed-accuracy balance, it should keep the"
+                                                        " closest aspect ratio possible to the images or videos to be processed. Using `-1` in"
+                                                        " any of the dimensions, OP will choose the optimal aspect ratio depending on the user's"
+                                                        " input value. E.g. the default `-1x368` is equivalent to `656x368` in 16:9 resolutions,"
+                                                        " e.g. full HD (1980x1080) and HD (1280x720) resolutions.");
 DEFINE_int32(scale_number,              1,              "Number of scales to average.");
 DEFINE_double(scale_gap,                0.3,            "Scale gap between scales. No effect unless scale_number > 1. Initial scale is always 1."
                                                         " If you want to change the initial scale, you actually want to multiply the"
@@ -239,7 +241,7 @@ int openPoseTutorialWrapper3()
     // outputSize
     const auto outputSize = op::flagsToPoint(FLAGS_resolution, "1280x720");
     // netInputSize
-    const auto netInputSize = op::flagsToPoint(FLAGS_net_resolution, "656x368");
+    const auto netInputSize = op::flagsToPoint(FLAGS_net_resolution, "-1x368");
     // faceNetInputSize
     const auto faceNetInputSize = op::flagsToPoint(FLAGS_face_net_resolution, "368x368 (multiples of 16)");
     // handNetInputSize

examples/tutorial_wrapper/2_user_synchronous.cpp

@@ -49,10 +49,12 @@ DEFINE_int32(keypoint_scale,            0,              "Scaling of the (x,y) co
 // OpenPose Body Pose
 DEFINE_string(model_pose,               "COCO",         "Model to be used. E.g. `COCO` (18 keypoints), `MPI` (15 keypoints, ~10% faster), "
                                                         "`MPI_4_layers` (15 keypoints, even faster but less accurate).");
-DEFINE_string(net_resolution,           "656x368",      "Multiples of 16. If it is increased, the accuracy potentially increases. If it is decreased,"
-                                                        " the speed increases. For maximum speed-accuracy balance, it should keep the closest aspect"
-                                                        " ratio possible to the images or videos to be processed. E.g. the default `656x368` is"
-                                                        " optimal for 16:9 videos, e.g. full HD (1980x1080) and HD (1280x720) videos.");
+DEFINE_string(net_resolution,           "656x368",      "Multiples of 16. If it is increased, the accuracy potentially increases. If it is"
+                                                        " decreased, the speed increases. For maximum speed-accuracy balance, it should keep the"
+                                                        " closest aspect ratio possible to the images or videos to be processed. Using `-1` in"
+                                                        " any of the dimensions, OP will choose the optimal aspect ratio depending on the user's"
+                                                        " input value. E.g. the default `-1x368` is equivalent to `656x368` in 16:9 resolutions,"
+                                                        " e.g. full HD (1980x1080) and HD (1280x720) resolutions.");
 DEFINE_int32(scale_number,              1,              "Number of scales to average.");
 DEFINE_double(scale_gap,                0.3,            "Scale gap between scales. No effect unless scale_number > 1. Initial scale is always 1."
                                                         " If you want to change the initial scale, you actually want to multiply the"
@@ -319,7 +321,7 @@ int openPoseTutorialWrapper2()
     // outputSize
     const auto outputSize = op::flagsToPoint(FLAGS_resolution, "1280x720");
     // netInputSize
-    const auto netInputSize = op::flagsToPoint(FLAGS_net_resolution, "656x368");
+    const auto netInputSize = op::flagsToPoint(FLAGS_net_resolution, "-1x368");
     // faceNetInputSize
     const auto faceNetInputSize = op::flagsToPoint(FLAGS_face_net_resolution, "368x368 (multiples of 16)");
     // handNetInputSize

examples/tutorial_wrapper/3_user_asynchronous.cpp

@@ -49,10 +49,12 @@ DEFINE_int32(keypoint_scale,            0,              "Scaling of the (x,y) co
 // OpenPose Body Pose
 DEFINE_string(model_pose,               "COCO",         "Model to be used. E.g. `COCO` (18 keypoints), `MPI` (15 keypoints, ~10% faster), "
                                                         "`MPI_4_layers` (15 keypoints, even faster but less accurate).");
-DEFINE_string(net_resolution,           "656x368",      "Multiples of 16. If it is increased, the accuracy potentially increases. If it is decreased,"
-                                                        " the speed increases. For maximum speed-accuracy balance, it should keep the closest aspect"
-                                                        " ratio possible to the images or videos to be processed. E.g. the default `656x368` is"
-                                                        " optimal for 16:9 videos, e.g. full HD (1980x1080) and HD (1280x720) videos.");
+DEFINE_string(net_resolution,           "656x368",      "Multiples of 16. If it is increased, the accuracy potentially increases. If it is"
+                                                        " decreased, the speed increases. For maximum speed-accuracy balance, it should keep the"
+                                                        " closest aspect ratio possible to the images or videos to be processed. Using `-1` in"
+                                                        " any of the dimensions, OP will choose the optimal aspect ratio depending on the user's"
+                                                        " input value. E.g. the default `-1x368` is equivalent to `656x368` in 16:9 resolutions,"
+                                                        " e.g. full HD (1980x1080) and HD (1280x720) resolutions.");
 DEFINE_int32(scale_number,              1,              "Number of scales to average.");
 DEFINE_double(scale_gap,                0.3,            "Scale gap between scales. No effect unless scale_number > 1. Initial scale is always 1."
                                                         " If you want to change the initial scale, you actually want to multiply the"
@@ -278,7 +280,7 @@ int openPoseTutorialWrapper1()
     // outputSize
     const auto outputSize = op::flagsToPoint(FLAGS_resolution, "1280x720");
     // netInputSize
-    const auto netInputSize = op::flagsToPoint(FLAGS_net_resolution, "656x368");
+    const auto netInputSize = op::flagsToPoint(FLAGS_net_resolution, "-1x368");
     // faceNetInputSize
     const auto faceNetInputSize = op::flagsToPoint(FLAGS_face_net_resolution, "368x368 (multiples of 16)");
     // handNetInputSize

examples_beta/openpose3d/openpose3d.cpp

@@ -49,10 +49,12 @@ DEFINE_int32(keypoint_scale,            0,              "Scaling of the (x,y) co
 // OpenPose Body Pose
 DEFINE_string(model_pose,               "COCO",         "Model to be used. E.g. `COCO` (18 keypoints), `MPI` (15 keypoints, ~10% faster), "
                                                         "`MPI_4_layers` (15 keypoints, even faster but less accurate).");
-DEFINE_string(net_resolution,           "656x368",      "Multiples of 16. If it is increased, the accuracy potentially increases. If it is decreased,"
-                                                        " the speed increases. For maximum speed-accuracy balance, it should keep the closest aspect"
-                                                        " ratio possible to the images or videos to be processed. E.g. the default `656x368` is"
-                                                        " optimal for 16:9 videos, e.g. full HD (1980x1080) and HD (1280x720) videos.");
+DEFINE_string(net_resolution,           "656x368",      "Multiples of 16. If it is increased, the accuracy potentially increases. If it is"
+                                                        " decreased, the speed increases. For maximum speed-accuracy balance, it should keep the"
+                                                        " closest aspect ratio possible to the images or videos to be processed. Using `-1` in"
+                                                        " any of the dimensions, OP will choose the optimal aspect ratio depending on the user's"
+                                                        " input value. E.g. the default `-1x368` is equivalent to `656x368` in 16:9 resolutions,"
+                                                        " e.g. full HD (1980x1080) and HD (1280x720) resolutions.");
 DEFINE_int32(scale_number,              1,              "Number of scales to average.");
 DEFINE_double(scale_gap,                0.3,            "Scale gap between scales. No effect unless scale_number > 1. Initial scale is always 1."
                                                         " If you want to change the initial scale, you actually want to multiply the"
@@ -153,7 +155,7 @@ int openpose3d()
     // outputSize
     const auto outputSize = op::flagsToPoint(FLAGS_resolution, "1280x720");
     // netInputSize
-    const auto netInputSize = op::flagsToPoint(FLAGS_net_resolution, "656x368");
+    const auto netInputSize = op::flagsToPoint(FLAGS_net_resolution, "-1x368");
     // faceNetInputSize
     const auto faceNetInputSize = op::flagsToPoint(FLAGS_face_net_resolution, "368x368 (multiples of 16)");
     // handNetInputSize

include/openpose/wrapper/wrapper.hpp

@@ -548,12 +548,27 @@ namespace op
                         error(message, __LINE__, __FUNCTION__, __FILE__);
                     }
                 }
+
             }
             else if (finalOutputSize.x == -1 || finalOutputSize.y == -1)
             {
                 const auto message = "Output resolution cannot be (-1 x -1) unless wrapperStructInput.producerSharedPtr is also set.";
                 error(message, __LINE__, __FUNCTION__, __FILE__);
             }
+            // Set poseNetInputSize if -1 used
+            Point<int> poseNetInputSize = wrapperStructPose.netInputSize;
+            if (poseNetInputSize.x == -1 || poseNetInputSize.y == -1)
+            {
+                if (producerSize.area() <= 0)
+                {
+                    const auto message = "Net resolution cannot be -1 for image_dir, only for video and webcam.";
+                    error(message, __LINE__, __FUNCTION__, __FILE__);
+                }
+                else if (poseNetInputSize.x == -1)
+                    poseNetInputSize.x = 16 * intRound(poseNetInputSize.y * producerSize.x / (float) producerSize.y / 16.f);
+                else // if (poseNetInputSize.y == -1)
+                    poseNetInputSize.y = 16 * intRound(poseNetInputSize.x * producerSize.y / (float) producerSize.x / 16.f);
+            }
 
             // Producer
             if (wrapperStructInput.producerSharedPtr != nullptr)
@@ -567,11 +582,11 @@ namespace op
                 wDatumProducer = nullptr;
 
             // Pose estimators
-            const Point<int>& poseNetOutputSize = wrapperStructPose.netInputSize;
+            const Point<int>& poseNetOutputSize = poseNetInputSize;
             std::vector<std::shared_ptr<PoseExtractor>> poseExtractors;
             for (auto gpuId = 0; gpuId < gpuNumber; gpuId++)
                 poseExtractors.emplace_back(std::make_shared<PoseExtractorCaffe>(
-                    wrapperStructPose.netInputSize, poseNetOutputSize, finalOutputSize, wrapperStructPose.scalesNumber,
+                    poseNetInputSize, poseNetOutputSize, finalOutputSize, wrapperStructPose.scalesNumber,
                     wrapperStructPose.poseModel, wrapperStructPose.modelFolder, gpuId + gpuNumberStart,
                     wrapperStructPose.heatMapTypes, wrapperStructPose.heatMapScale
                 ));
@@ -613,7 +628,7 @@ namespace op
 
             // Input cvMat to OpenPose format
             const auto cvMatToOpInput = std::make_shared<CvMatToOpInput>(
-                wrapperStructPose.netInputSize, wrapperStructPose.scalesNumber, wrapperStructPose.scaleGap
+                poseNetInputSize, wrapperStructPose.scalesNumber, wrapperStructPose.scaleGap
             );
             spWCvMatToOpInput = std::make_shared<WCvMatToOpInput<TDatumsPtr>>(cvMatToOpInput);
             const auto cvMatToOpOutput = std::make_shared<CvMatToOpOutput>(finalOutputSize, renderOutput);