G-API Smart Framing Demo #3421

dbudniko · 2022-03-29T16:11:40Z

@TolyaTalamanov @mpashchenkov @OrestChura please review this draft implementation

mpashchenkov · 2022-03-30T06:39:39Z

demos/smart_framing_demo/cpp_gapi/include/custom_kernels.hpp

+
+namespace custom {
+
+const std::vector<std::string> coco_classes = {


I think you can load this from omz/data/dataset_calsses/coco_80cl.txt

@mpashchenkov are aware about any examples of doing so for classes?

FYI omz/demos/instance_segmentation_demo (cases.py example for labels flag)

I think you can just add --labels flag for path to file with labels.

mpashchenkov · 2022-03-30T06:56:40Z

@dbudniko, need to add tests for demo to omz/demos/tests/cases.py

mpashchenkov

Please delete or move to NOTE / TODO all commented code.

mpashchenkov · 2022-03-31T12:16:36Z

demos/README.md

@@ -38,6 +38,7 @@
   omz_demos_image_retrieval_demo_python
   omz_demos_segmentation_demo_cpp
   omz_demos_segmentation_demo_python
+   omz_demos_smart_framing_demo_cpp


omz_demos_smart_framing_demo_cpp -->
omz_demos_smart_framing_demo_cpp_gapi

mpashchenkov · 2022-03-31T12:19:16Z

demos/README.md

@@ -139,7 +140,8 @@ The Open Model Zoo includes the following demos:
 - [Pedestrian Tracker C++ Demo](./pedestrian_tracker_demo/cpp/README.md) - Demo application for pedestrian tracking scenario.
 - [Place Recognition Python\* Demo](./place_recognition_demo/python/README.md) - This demo demonstrates how to run Place Recognition models using OpenVINO™.
 - [Security Barrier Camera C++ Demo](./security_barrier_camera_demo/cpp/README.md) - Vehicle Detection followed by the Vehicle Attributes and License-Plate Recognition, supports images/video and camera inputs.
- [Speech Recognition DeepSpeech Python\* Demo](./speech_recognition_deepspeech_demo/python/README.md) - Speech recognition demo: accepts an audio file with an English phrase on input and converts it into text. This demo does streaming audio data processing and can optionally provide current transcription of the processed part.
+- [Smart Framing C++ Demo](./smart_framing_demo/cpp_gapi/README.md) - Person Detection followed by the Smart Framing/Croping and optionally Super Resolution, supports images/video and camera inputs.


Smart Framing C++ G-API Demo

mpashchenkov · 2022-03-31T12:58:41Z

demos/smart_framing_demo/cpp_gapi/main.cpp

+        cv::GArray<custom::DetectedObject> yolo_detections;
+
+        // Now build the graph
+        cv::GMat in;


Please put graph 140 - 166 + GComputation to the beginning of main().

Please put graph after 221 line if it is possible

mpashchenkov · 2022-03-31T13:04:03Z

demos/smart_framing_demo/cpp_gapi/main.cpp

+            if (!FLAGS_no_show) {
+                //Draw detections on original image
+                for (const auto& el : objects) {
+                    std::cout << el << std::endl;


slog::debug << ...

mpashchenkov · 2022-03-31T13:05:14Z

demos/smart_framing_demo/cpp_gapi/main.cpp

+                cv::Rect init_rect;
+                for (const auto& el : objects) {
+                    if (el.labelID == 0) {//person ID
+                        init_rect = init_rect | (cv::Rect)el;


static_cast < cv::Rect > (el)

mpashchenkov · 2022-03-31T13:23:10Z

demos/smart_framing_demo/cpp_gapi/main.cpp

+            }
+            else if (FLAGS_at_sr == "3ch") {


Suggested change

}

else if (FLAGS_at_sr == "3ch") {

} else if (FLAGS_at_sr == "3ch") {

mpashchenkov · 2022-03-31T13:23:37Z

demos/smart_framing_demo/cpp_gapi/main.cpp

+            }
+            else {


Suggested change

}

else {

} else {

mpashchenkov · 2022-03-31T13:23:55Z

demos/smart_framing_demo/cpp_gapi/main.cpp

+            }
+            else {


Suggested change

}

else {

} else {

mpashchenkov · 2022-03-31T13:24:11Z

demos/smart_framing_demo/cpp_gapi/main.cpp

+        }
+        else {


Suggested change

}

else {

} else {

mpashchenkov · 2022-03-31T13:24:33Z

demos/smart_framing_demo/cpp_gapi/main.cpp

+            }
+            else {


Suggested change

}

else {

} else {

mpashchenkov · 2022-03-31T17:29:55Z

demos/smart_framing_demo/cpp_gapi/main.cpp

+                    cv::rectangle(image, el, cv::Scalar{ 0,255,0 }, 2, cv::LINE_8, 0);
+                    cv::putText(image, el.label, el.tl(), cv::FONT_HERSHEY_SIMPLEX, 1.0, { 0,255,0 }, 2);


Would it be better to draw these with render primitives in graph?
Up to you.

mpashchenkov · 2022-03-31T17:35:41Z

demos/smart_framing_demo/cpp_gapi/src/custom_kernels.cpp

+        auto entriesNum = sideW * sideH;
+        const float* output_blob = (float*)blob.data;
+
+        auto postprocessRawData = /*(yoloVersion == YOLO_V4 || yoloVersion == YOLO_V4_TINY) ?*/ sigmoid;// : linear;


Does demo have unsupported scenarios?

mpashchenkov · 2022-03-31T17:36:15Z

demos/smart_framing_demo/cpp_gapi/src/custom_kernels.cpp

+        for (int i = 0; i < entriesNum; ++i) {
+            int row = i / sideW;
+            int col = i % sideW;
+            //for (int n = 0; n < region.num; ++n) {


Looks like legacy

mpashchenkov · 2022-03-31T17:37:13Z

demos/smart_framing_demo/cpp_gapi/src/custom_kernels.cpp

+                int obj_index = calculateEntryIndex(entriesNum, /*region.coords*/ region_coords, /*region.classes*/ region_classes, n * entriesNum + i, /*region.coords*/ region_coords);
+                int box_index = calculateEntryIndex(entriesNum, /*region.coords*/ region_coords, /*region.classes*/ region_classes, n * entriesNum + i, 0);


mpashchenkov · 2022-03-31T17:37:30Z

demos/smart_framing_demo/cpp_gapi/src/custom_kernels.cpp

+                    //double height = std::exp(output_blob[box_index + 3 * entriesNum]) * region.anchors[2 * n + 1] * original_im_h / scaleH;
+                    //double width = std::exp(output_blob[box_index + 2 * entriesNum]) * region.anchors[2 * n] * original_im_w / scaleW;


mpashchenkov · 2022-03-31T17:45:58Z

demos/smart_framing_demo/cpp_gapi/src/custom_kernels.cpp

+                cv::BORDER_CONSTANT, cv::Scalar::all(127));
+        }
+        else {
+            image.copyTo(out);


Is this copy necessary? You can get image for next operations from previous graph stage.

yes. we need this copy.

Why don't you use frame from cv::gapi::copy() instead?

mpashchenkov · 2022-03-31T17:49:04Z

demos/smart_framing_demo/cpp_gapi/src/custom_kernels.cpp

+    static void run(const cv::Mat & in,
+        cv::Mat & out) {
+        int h = in.size[2];
+        int w = in.size[3];


Suggested change

static void run(const cv::Mat & in,

cv::Mat & out) {

int h = in.size[2];

int w = in.size[3];

static void run(const cv::Mat& in,

cv::Mat& out) {

int h = in.size[2];

int w = in.size[3];

mpashchenkov · 2022-03-31T17:52:05Z

demos/smart_framing_demo/cpp_gapi/src/custom_kernels.cpp

+        std::transform(in_p, in_p + h * w, out_p,
+            [](float v) { return static_cast<uint8_t>(v * 255); });
+    }


Can it be:
out = in * 255;
Are in and out 1-channel cv::Mat?

Need to check with Tolik. This is his kernel.

mpashchenkov · 2022-03-31T17:54:10Z

demos/tests/cases.py

+            test_cases=combine_cases(
+        TestCase(options={'--no_show': None, '-at_sr': '3ch',
+            **MONITORS,
+            '-i': DataPatternArg('smart-classroom-demo'),


Ok. I will add comment about it bit later.

mpashchenkov · 2022-03-31T17:54:32Z

demos/smart_framing_demo/cpp_gapi/src/custom_kernels.cpp

+
+GAPI_OCV_KERNEL(OCVSuperResolutionPostProcessingKernel, custom::GSuperResolutionPostProcessingKernel) {
+    static void run(const cv::Mat & image, cv::Mat & out) {
+        float* outputData = (float*)image.data;


c-style cast

There's still C-style cast

…ing_demo

demos/smart_framing_demo/cpp_gapi/main.cpp

dbudniko · 2022-04-04T12:52:05Z

@Wovchena please review.

Wovchena · 2022-04-04T13:08:34Z

I'm not sure when I get to it

ivikhrev · 2022-04-04T15:54:11Z

demos/smart_framing_demo/cpp_gapi/README.md

+    -t_conf_yolo               Optional. YOLO v4 Tiny confidence threshold.
+    -t_box_iou_yolo            Optional. YOLO v4 Tiny box IOU threshold.
+    -advanced_pp               Optional. Use advanced post-processing for the YOLO v4 Tiny.
+    -apply_sr                  Optional. Use Super Resolution post processing model.


Can't we decide whether to apply super resolution postprocessing based on was model provided or not?

ivikhrev · 2022-04-04T15:58:25Z

demos/smart_framing_demo/cpp_gapi/README.md

+    -limit "<num>"             Optional. Number of frames to store in output. If 0 is set, all frames are stored.
+    -res "<WxH>"               Optional. Set camera resolution in format WxH.
+    -m_yolo "<path>"           Required. Path to an .xml file with a trained YOLO v4 Tiny model.
+    -at_sr "<type>"            Required if Super Resolution is not disabled by apply_sr=false flag. Architecture type: Super Resolution - 3 channels input (3ch) or 1 channel input (1ch).


Both single-image-super-resolution (1032 and 1033) models accept 3 channel images. What for 1 channel input support?

Some use cases and SR models requires 1ch input. 3ch OMZ models are currently default. But we want to have optional 1ch scenario support in place.

ivikhrev · 2022-04-04T16:07:17Z

demos/smart_framing_demo/cpp_gapi/smart_framing_demo_gapi.hpp

+* \brief This function shows a help message
+*/
+
+static void showUsage() {


We've redesigned a usage message for demos, see example here:

open_model_zoo/demos/interactive_face_detection_demo/cpp_gapi/main.cpp

Line 154 in 3bd940d

void parse(int argc, char* argv[]) {

Please, update in accordance with it:

move to main.cpp

wrap into anonymous namespace

define flag and it's message nearby like this:

constexpr char h_msg[] = "show the help message and exit"; DEFINE_bool(h, false, h_msg);

Mark optional parameters with [ ]

ivikhrev · 2022-04-04T16:26:53Z

demos/smart_framing_demo/cpp_gapi/smart_framing_demo_gapi.hpp

+
+#include <gflags/gflags.h>
+#include <utils/default_flags.hpp>
+#include <models/detection_model.h>


Transitive includes is a bad thing, please include models/detection_model.h in file where you actually use it.

mpashchenkov · 2022-04-05T13:24:20Z

demos/smart_framing_demo/cpp_gapi/include/custom_kernels.hpp

+
+#include <utils/slog.hpp>
+
+namespace IE = InferenceEngine;


Does it make sense?

mpashchenkov · 2022-04-05T13:26:21Z

demos/smart_framing_demo/cpp_gapi/main.cpp

+
+
+
+//#include "smart_framing_demo_gapi.hpp"


??? And demo works without this include. It is weird

Understood. Need to remove.

mpashchenkov · 2022-04-05T13:28:22Z

demos/smart_framing_demo/cpp_gapi/main.cpp

+
+
+namespace util {
+//bool ParseAndCheckCommandLine(int argc, char *argv[]) {


Ok. Understood. New behavior is the same as in IFD.

MB need remove this code + include

mpashchenkov · 2022-04-05T13:33:19Z

demos/smart_framing_demo/cpp_gapi/main.cpp

+        }
+
+
+        /** ---------------- Main graph of demo ---------------- **/


Comment without graph :)

mpashchenkov · 2022-04-05T13:34:31Z

demos/smart_framing_demo/cpp_gapi/main.cpp

+        cv::GArray<custom::DetectedObject> yolo_detections;
+
+        // Now build the graph
+        cv::GMat in;


Please put graph after 221 line if it is possible

mpashchenkov · 2022-04-05T13:35:39Z

demos/smart_framing_demo/cpp_gapi/smart_framing_demo_gapi.hpp

+DEFINE_INPUT_FLAGS
+DEFINE_OUTPUT_FLAGS
+
+static const char help_message[] = "Print a usage message.";


It is unused file now, isn't it?

Removed after cleanup

Graph code rearranged. Don't completely understand "put graph after 221 line "

mpashchenkov · 2022-04-05T13:37:39Z

demos/smart_framing_demo/cpp_gapi/src/custom_kernels.cpp

+    }
+};
+
+using GPostProc = cv::GOpaque<YOLOv4TinyPostProcessing>;


mpashchenkov · 2022-04-05T13:42:07Z

demos/smart_framing_demo/cpp_gapi/src/custom_kernels.cpp

+                cv::BORDER_CONSTANT, cv::Scalar::all(127));
+        }
+        else {
+            image.copyTo(out);


Why don't you use frame from cv::gapi::copy() instead?

dbudniko · 2022-04-06T10:22:58Z

Will we continue and merge?

dbudniko · 2022-04-07T10:36:37Z

@dmatveev FYI

dmatveev · 2022-04-08T06:13:53Z

demos/smart_framing_demo/cpp_gapi/include/custom_kernels.hpp

+using GDetections = cv::GArray<DetectedObject>;
+using GLabels = cv::GArray<std::string>;
+
+G_API_OP(GYOLOv4TinyPostProcessingKernel, < GDetections(cv::GMat, cv::GMat, cv::GMat, GLabels, float, float, bool) >, "custom.yolov4_tiny_post_processing") {


Operation shouldn't be called Kernel, as a Kernel is an implementation to the Operation.

dmatveev · 2022-04-08T06:17:01Z

demos/smart_framing_demo/cpp_gapi/include/custom_kernels.hpp

+struct DetectedObject : public cv::Rect2f
+{
+    unsigned int labelID;
+    std::string label;
+    float confidence;
+};


I doubt is subclassing from Rect2f is a good idea.. Not sure if it is bad, too.

The reason we usually don't use structures like this in pipeline and instead produce multiple arrays of different types is that it is easier to process those arrays with other functions like infer-list without any extra custom code.

dmatveev · 2022-04-08T06:17:38Z

demos/smart_framing_demo/cpp_gapi/include/custom_kernels.hpp

+        }
+};
+
+G_API_OP(GSmartFramingKernel, <cv::GMat(cv::GMat, GDetections)>, "custom.smart_framing") {


I assume it should be SmartCrop isn't it?

dmatveev · 2022-04-08T06:18:01Z

demos/smart_framing_demo/cpp_gapi/include/custom_kernels.hpp

+            cv::GMatDesc out_desc(CV_8U /* depth */, in.dims[1] /* channels */, cv::Size(in.dims[3], in.dims[2]), false /* planar */);
+            return out_desc;
+        }
+};


Please drop Kernel suffix from all the operations

dmatveev · 2022-04-08T06:18:38Z

demos/smart_framing_demo/cpp_gapi/include/custom_kernels.hpp

+        // NB: Input is ND mat.
+        return cv::GMatDesc{ CV_8U, in.dims[1], cv::Size(in.dims[3], in.dims[2]) };
+    }
+};


Don't we have this in our base framework? Should we plan extending the existing function instead?

dmatveev · 2022-04-08T06:23:27Z

demos/smart_framing_demo/cpp_gapi/main.cpp

+static cv::gapi::GKernelPackage getKernelPackage(const std::string& type) {
+    if (type == "opencv") {
+        return cv::gapi::combine(cv::gapi::core::cpu::kernels(),
+                                 cv::gapi::imgproc::cpu::kernels());
+    } else if (type == "fluid") {
+        return cv::gapi::combine(cv::gapi::core::fluid::kernels(),
+                                 cv::gapi::imgproc::fluid::kernels());
+    } else {
+        throw std::logic_error("Unsupported kernel package type: " + type);
+    }
+    GAPI_Assert(false && "Unreachable code!");
+}


This probably doesn't make sense as so far there is no image processing kernels in the pipeline?

dmatveev · 2022-04-08T06:26:23Z

demos/smart_framing_demo/cpp_gapi/main.cpp

+                auto out_b = custom::GCvt32Fto8U::on(cv::gapi::infer<SRNet>(b));
+                auto out_g = custom::GCvt32Fto8U::on(cv::gapi::infer<SRNet>(g));
+                auto out_r = custom::GCvt32Fto8U::on(cv::gapi::infer<SRNet>(r));
+                out_sr_pp = cv::gapi::merge3(out_b, out_g, out_r);


Ok, so this in theory could leverage Fluid but in this face "Cvf32Fto8U" should be implemented as a Fluid kernel.

..yes they can be custom

dmatveev · 2022-04-08T06:27:25Z

demos/smart_framing_demo/cpp_gapi/src/custom_kernels.cpp

+
+        cv::merge(imgPlanes, out);
+    }
+};


Can we unroll this kernel into a graph is possible? I see its treshold/convertTo/merge path can run on Fluid as well.

dmatveev · 2022-04-08T06:29:10Z

demos/smart_framing_demo/cpp_gapi/src/custom_kernels.cpp

+    }
+};
+
+GAPI_OCV_KERNEL(OCVSmartFramingMakeBorderKernel, custom::GSmartFramingMakeBorderKernel) {


This is harder than I thought...

As it has a lot of temporary Mats created inside, those can be moved to State (as we turn this kernel into Stateful) to avoid run-time memory allocations.

Wovchena

There are debug prints. Please remove them

Wovchena · 2022-09-16T09:54:25Z

demos/smart_framing_demo/cpp_gapi/README.md

@@ -0,0 +1,110 @@
+# G-API Smart Framing Demo
+
+This demo shows how to perform smart framing using G-API.


Please add a gif to show how demo looks.

Wovchena · 2022-09-16T09:55:23Z

demos/smart_framing_demo/cpp_gapi/README.md

+    --t_box_iou_yolo            Optional. YOLO v4 Tiny box IOU threshold.
+    --advanced_pp               Optional. Use advanced post-processing for the YOLO v4 Tiny.
+    --show                      Optional. (Don't) show output.
+    -u                          Optional. List of monitors to show initially.


This is misaligned with what the demo actually prints.

Wovchena · 2022-09-16T09:57:53Z

demos/smart_framing_demo/cpp_gapi/main.cpp

+                  << "\n\t[--loop]                                      " << loop_msg
+                  << "\n\t[ -o <OUTPUT>]                                " << o_msg
+                  << "\n\t[--labels]                                    " << labels_msg
+                  << "\n\t[--crop_with_borders]                         " << crop_with_borders_msg


Please sort flags:

-h

--help

required flags

-i

optional flags sorted alphabetically

Wovchena · 2022-09-16T10:38:41Z

demos/smart_framing_demo/cpp_gapi/main.cpp

+        cv::GArray<std::string> labels;
+
+        std::tie(blob26x26, blob13x13) = cv::gapi::infer<YOLOv4TinyNet>(in);
+        yolo_detections = custom::GYOLOv4TinyPostProcessingKernel::on(in, blob26x26, blob13x13, labels, FLAGS_t_conf_yolo, FLAGS_t_box_iou_yolo, FLAGS_advanced_pp);


There's warning on windows warning C4244: 'argument': conversion from 'double' to 'float', possible loss of data. Please put explicit cast

Wovchena · 2022-09-16T10:39:48Z

demos/smart_framing_demo/cpp_gapi/src/custom_kernels.cpp

+            cv::Mat SF_resized_ROI;
+            cv::Size target_size;
+            target_size.height = image.size().height;
+            target_size.width = even_rect.width * (static_cast<float>(image.size().height) / static_cast<float>(even_rect.height));


There's warning on Windows: '=': conversion from 'float' to '_Tp', possible loss of data. Please resolve it

Wovchena · 2022-09-16T12:23:17Z

demos/smart_framing_demo/cpp_gapi/src/custom_kernels.cpp

+                    auto* in_p = in.ptr<const float>();
+
+                    std::transform(in_p, in_p + h * w, out_p,
+                                   [](float v) { return static_cast<uint8_t>(v * 255); });


Why not in.convertTo()?

There's even comment about it on top of this kernel

Wovchena · 2022-09-16T12:32:25Z

demos/smart_framing_demo/cpp_gapi/README.md

+To run the demo, please provide paths to the model in the IR format, and to an input video, image, or folder with images:
+
+```bash
+./smart_framing_demo_gapi/ -m <path_to_model> -i <path_to_file>


Suggested change

./smart_framing_demo_gapi/ -m <path_to_model> -i <path_to_file>

./smart_framing_demo_gapi -m <path_to_model> -i <path_to_file>

Wovchena · 2022-09-16T12:34:02Z

demos/smart_framing_demo/cpp_gapi/main.cpp

+constexpr char yolo_m_msg[] = "path to an .xml file with a trained YOLO v4 Tiny model.";
+DEFINE_string(m_yolo, "", yolo_m_msg);


Suggested change

constexpr char yolo_m_msg[] = "path to an .xml file with a trained YOLO v4 Tiny model.";

DEFINE_string(m_yolo, "", yolo_m_msg);

constexpr char m_msg[] = "path to an .xml file with a trained YOLO v4 Tiny model.";

DEFINE_string(m, "", m_msg);

Could you make it just m: --m_yolo->-m because it's the primal model of the demo

Wovchena · 2022-09-16T12:41:09Z

demos/smart_framing_demo/cpp_gapi/src/custom_kernels.cpp

+
+GAPI_OCV_KERNEL(OCVSuperResolutionPostProcessingKernel, custom::GSuperResolutionPostProcessingKernel) {
+    static void run(const cv::Mat & image, cv::Mat & out) {
+        float* outputData = (float*)image.data;


There's still C-style cast

Wovchena · 2022-09-16T12:45:07Z

demos/smart_framing_demo/cpp_gapi/src/custom_kernels.cpp

+
+} // anonymous namespace
+
+


Suggested change

Two blank lines are usually prohibited for C++. It's not Python

Dmitry Budnikov added 3 commits March 29, 2022 19:04

smart framing non-tested draft

67eb1bc

Merge branch 'master' into dbudniko/smart_framing_demo

00fe67d

try to fix models name

e3d2ea4

mpashchenkov reviewed Mar 30, 2022

View reviewed changes

Dmitry Budnikov added 9 commits March 30, 2022 11:25

try to fix build

cf9d9a5

use kernel package

f631294

README update

d172846

add test

601e7eb

reference README in main README

6b447fd

fix path

957ef22

trailing space

9564048

trailing space in README

cf727d7

fix output plus more windows imshow

cad71ef

mpashchenkov reviewed Mar 31, 2022

View reviewed changes

Dmitry Budnikov added 11 commits March 31, 2022 16:40

draft labels files logic

655d663

Address comments from Maxim

65b92d6

add header for slog

3946a42

correct include

179f281

fix cases.py syntax error

677472c

move graph few lines upper

fb93f26

add dependency

c5bb31a

rename labels

5f1e821

getting closer

33edc4b

coco labels remove

dcc40d3

remove from header

abffc8a

mpashchenkov requested a review from OrestChura March 31, 2022 17:14

mpashchenkov reviewed Mar 31, 2022

View reviewed changes

mpashchenkov self-requested a review March 31, 2022 17:56

More comments from Maxim are addressed

4e49964

Dmitry Budnikov added 5 commits April 1, 2022 11:28

fix misprint

da2289b

remove space

0b14d58

correct labels graph logic and kernels

cd87578

float*

b093956

Merge remote-tracking branch 'origin/master' into dbudniko/smart_fram…

e27d3e2

…ing_demo

mpashchenkov reviewed Apr 1, 2022

View reviewed changes

demos/smart_framing_demo/cpp_gapi/main.cpp Show resolved Hide resolved

Dmitry Budnikov added 8 commits April 1, 2022 14:11

add header

ba9cb11

images capture header

e9cd275

fix header name

fcec580

add const

4656702

more consts

cff4130

clean up

d84d675

new smart cropping algorithm

d905a6c

try to fix build

5eb8a56

Wovchena requested a review from ivikhrev April 4, 2022 13:09

ivikhrev reviewed Apr 4, 2022

View reviewed changes

Dmitry Budnikov added 4 commits April 5, 2022 15:08

address comments from Ivan

f6b70df

try to comment out unused code

c7e6f2b

more misprints are fixed

cbdb3dd

comment out temporary unused flags

5288e5d

mpashchenkov reviewed Apr 5, 2022

View reviewed changes

Dmitry Budnikov added 3 commits April 5, 2022 17:49

clean up

91c3267

more comments from Maxim are addressed

893ec91

fix flag for test cases.py

3b7d36e

dmatveev reviewed Apr 8, 2022

View reviewed changes

Wovchena requested changes Sep 16, 2022

View reviewed changes


		namespace custom {

		const std::vector<std::string> coco_classes = {

	}
	else if (FLAGS_at_sr == "3ch") {
	} else if (FLAGS_at_sr == "3ch") {

		cv::rectangle(image, el, cv::Scalar{ 0,255,0 }, 2, cv::LINE_8, 0);
		cv::putText(image, el.label, el.tl(), cv::FONT_HERSHEY_SIMPLEX, 1.0, { 0,255,0 }, 2);

		int obj_index = calculateEntryIndex(entriesNum, /region.coords/ region_coords, /region.classes/ region_classes, n * entriesNum + i, /region.coords/ region_coords);
		int box_index = calculateEntryIndex(entriesNum, /region.coords/ region_coords, /region.classes/ region_classes, n * entriesNum + i, 0);

		//double height = std::exp(output_blob[box_index + 3 * entriesNum]) * region.anchors[2 * n + 1] * original_im_h / scaleH;
		//double width = std::exp(output_blob[box_index + 2 * entriesNum]) * region.anchors[2 * n] * original_im_w / scaleW;



		namespace util {
		//bool ParseAndCheckCommandLine(int argc, char *argv[]) {

		}


		/ ---------------- Main graph of demo ---------------- /

		@@ -0,0 +1,110 @@
		# G-API Smart Framing Demo

		This demo shows how to perform smart framing using G-API.

	./smart_framing_demo_gapi/ -m <path_to_model> -i <path_to_file>
	./smart_framing_demo_gapi -m <path_to_model> -i <path_to_file>

		constexpr char yolo_m_msg[] = "path to an .xml file with a trained YOLO v4 Tiny model.";
		DEFINE_string(m_yolo, "", yolo_m_msg);




		//#include "smart_framing_demo_gapi.hpp"

G-API Smart Framing Demo #3421

Are you sure you want to change the base?

G-API Smart Framing Demo #3421

Conversation

dbudniko commented Mar 29, 2022

Choose a reason for hiding this comment

Choose a reason for hiding this comment

mpashchenkov Mar 30, 2022 • edited

Choose a reason for hiding this comment

mpashchenkov commented Mar 30, 2022

mpashchenkov left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

mpashchenkov Mar 31, 2022 • edited

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

dbudniko commented Apr 4, 2022

Wovchena commented Apr 4, 2022

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ivikhrev Apr 4, 2022 • edited

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ivikhrev Apr 4, 2022 • edited

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

mpashchenkov Mar 30, 2022 •

edited

mpashchenkov Mar 31, 2022 •

edited

ivikhrev Apr 4, 2022 •

edited

ivikhrev Apr 4, 2022 •

edited