Open Source Computer Vision Library https://opencv.org/
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

286 lines
9.0 KiB

Merge pull request #17020 from dbudniko:dbudniko/serialization_backend G-API Serialization routines * Serialization backend in tests, initial version * S11N/00: A Great Rename - "Serialization" is too long and too error-prone to type, so now it is renamed to "s11n" everywhere; - Same applies to "SRLZ"; - Tests also renamed to start with 'S11N.*' (easier to run); - Also updated copyright years in new files to 2020. * S11N/01: Some basic interface segregation - Moved some details (low-level functions) out of serialization.hpp; - Introduced I::IStream and I::OStream interfaces; - Implemented those via the existing [De]SerializationStream classes; - Moved all operators to use interfaces instead of classes; - Moved the htonl/ntohl handling out of operators (to the classes). The implementation didn't change much, it is a subject to the further refactoring * S11N/02: Basic operator reorg, basic tests, vector support - Reorganized operators on atomic types to follow >>/<< model (put them closer in the code for the respective types); - Introduce more operators for basic (scalar) types; - Drop all vector s11n overloads -- replace with a generic (template-based) one; - Introduced a new test suite where low-level s11n functionality is tested (for the basic types). * S11N/03: Operators reorganization - Sorted the Opaque types enum by complexity; - Reorganized the existing operators for basic types, also ordered by complexity; - Organized operators in three groups (Basics, OpenCV, G-API); - Added a generic serialization for variant<>; - Reimplemented some of the existing operators (for OpenCV and G-API data structures); - Introduced new operators for cv::gimpl data types. These operators (and so, the data structures) are not yet used in the graph dump/reconstruction routine, it will be done as a next step. * S11N/04: The Great Clean-up - Drop the duplicates of GModel data structures from the serialization, serialize the GModel data structures themselve instead (hand-written code replaced with operators). - Also removed usuned code for printing, etc. * S11N/05: Internal API Clean-up - Minimize the serialization API to just Streams and Operators; - Refactor and fix the graph serialization (deconstruction and reconstruction) routines, fix data addressing problems there; - Move the serialization.[ch]pp files to the core G-API library * S11N/06: Top-level API introduction - !!!This is likely the most invasive commit in the series!!! - Introduced a top-level API to serialize and deserialize a GComputation - Extended the compiler to support both forms of a GComputation: an expession based and a deserialized one. This has led to changes in the cv::GComputation::Priv and in its dependent components (even the transformation tests); - Had to extend the kernel API (GKernel) with extra information on operations (mainly `outMeta`) which was only available for expression based graphs. Now the `outMeta` can be taken from kernels too (and for the deserialized graphs it is the only way); - Revisited the internal serialization API, had to expose previously hidden entities (like `GSerialized`); - Extended the serialized graph info with new details (object counter, protocol). Added unordered_map generic serialization for that; - Reworked the very first pipeline test to be "proper"; GREEN now, the rest is to be reworked in the next iteration. * S11N/07: Tests reworked - Moved the sample pipeline tests w/serialization to test the public API (`cv::gapi::serialize`, then followed by `cv::gapi::deserialize<>`). All GREEN. - As a consequence, dropped the "Serialization" test backend as no longer necessary. * S11N/08: Final touches - Exposed the C++ native data types at Streams level; - Switched the ByteMemoryIn/OutStreams to store data in `char` internally (2x less memory for sample pipelines); - Fixed and refactored Mat dumping to the stream; - Renamed S11N pipeline tests to their new meaning. * linux build fix * fix RcDesc and int uint warnings * more Linux build fix * white space and virtual android error fix (attempt) * more warnings to be fixed * android warnings fix attempt * one more attempt for android build fix * android warnings one more fix * return back override * avoid size_t * static deserialize * and how do you like this, elon? anonymous namespace to fix android warning. * static inline * trying to fix standalone build * mat dims fix * fix mat r/w for standalone Co-authored-by: Dmitry Matveev <dmitry.matveev@intel.com>
5 years ago
// This file is part of OpenCV project.
// It is subject to the license terms in the LICENSE file found in the top-level directory
// of this distribution and at http://opencv.org/license.html.
//
// Copyright (C) 2020 Intel Corporation
#include "../test_precomp.hpp"
#include <ade/util/iota_range.hpp>
#include <opencv2/gapi/s11n.hpp>
namespace opencv_test
{
TEST(S11N, Pipeline_Crop_Rect)
{
cv::Rect rect_to{ 4,10,37,50 };
cv::Size sz_in = cv::Size(1920, 1080);
cv::Size sz_out = cv::Size(37, 50);
cv::Mat in_mat = cv::Mat::eye(sz_in, CV_8UC1);
cv::Mat out_mat_gapi(sz_out, CV_8UC1);
cv::Mat out_mat_ocv(sz_out, CV_8UC1);
// G-API code //////////////////////////////////////////////////////////////
cv::GMat in;
auto out = cv::gapi::crop(in, rect_to);
auto p = cv::gapi::serialize(cv::GComputation(in, out));
auto c = cv::gapi::deserialize<cv::GComputation>(p);
c.apply(in_mat, out_mat_gapi);
// OpenCV code /////////////////////////////////////////////////////////////
{
out_mat_ocv = in_mat(rect_to);
}
// Comparison //////////////////////////////////////////////////////////////
{
EXPECT_EQ(0, cvtest::norm(out_mat_ocv, out_mat_gapi, NORM_INF));
}
}
TEST(S11N, Pipeline_Canny_Bool)
{
const cv::Size sz_in(1280, 720);
cv::GMat in;
double thrLow = 120.0;
double thrUp = 240.0;
int apSize = 5;
bool l2gr = true;
cv::Mat in_mat = cv::Mat::eye(1280, 720, CV_8UC1);
cv::Mat out_mat_gapi(sz_in, CV_8UC1);
cv::Mat out_mat_ocv(sz_in, CV_8UC1);
// G-API code //////////////////////////////////////////////////////////////
auto out = cv::gapi::Canny(in, thrLow, thrUp, apSize, l2gr);
auto p = cv::gapi::serialize(cv::GComputation(in, out));
auto c = cv::gapi::deserialize<cv::GComputation>(p);
c.apply(in_mat, out_mat_gapi);
// OpenCV code /////////////////////////////////////////////////////////////
{
cv::Canny(in_mat, out_mat_ocv, thrLow, thrUp, apSize, l2gr);
}
// Comparison //////////////////////////////////////////////////////////////
EXPECT_EQ(0, cvtest::norm(out_mat_gapi, out_mat_ocv, NORM_INF));
}
TEST(S11N, Pipeline_Not)
{
cv::GMat in;
auto p = cv::gapi::serialize(cv::GComputation(in, cv::gapi::bitwise_not(in)));
auto c = cv::gapi::deserialize<cv::GComputation>(p);
cv::Mat in_mat = cv::Mat::eye(32, 32, CV_8UC1);
cv::Mat ref_mat = ~in_mat;
cv::Mat out_mat;
c.apply(in_mat, out_mat);
EXPECT_EQ(0, cvtest::norm(out_mat, ref_mat, NORM_INF));
out_mat = cv::Mat();
auto cc = c.compile(cv::descr_of(in_mat));
cc(in_mat, out_mat);
EXPECT_EQ(0, cvtest::norm(out_mat, ref_mat, NORM_INF));
}
TEST(S11N, Pipeline_Sum_Scalar)
{
cv::GMat in;
auto p = cv::gapi::serialize(cv::GComputation(in, cv::gapi::sum(in)));
auto c = cv::gapi::deserialize<cv::GComputation>(p);
cv::Mat in_mat = cv::Mat::eye(32, 32, CV_8UC1);
cv::Scalar ref_scl = cv::sum(in_mat);
cv::Scalar out_scl;
c.apply(in_mat, out_scl);
EXPECT_EQ(out_scl, ref_scl);
out_scl = cv::Scalar();
auto cc = c.compile(cv::descr_of(in_mat));
cc(in_mat, out_scl);
EXPECT_EQ(out_scl, ref_scl);
}
TEST(S11N, Pipeline_BinaryOp)
{
cv::GMat a, b;
auto p = cv::gapi::serialize(cv::GComputation(a, b, cv::gapi::add(a, b)));
auto c = cv::gapi::deserialize<cv::GComputation>(p);
cv::Mat in_mat = cv::Mat::eye(32, 32, CV_8UC1);
cv::Mat ref_mat = (in_mat + in_mat);
cv::Mat out_mat;
c.apply(in_mat, in_mat, out_mat);
EXPECT_EQ(0, cvtest::norm(out_mat, ref_mat, NORM_INF));
out_mat = cv::Mat();
auto cc = c.compile(cv::descr_of(in_mat), cv::descr_of(in_mat));
cc(in_mat, in_mat, out_mat);
EXPECT_EQ(0, cvtest::norm(out_mat, ref_mat, NORM_INF));
}
TEST(S11N, Pipeline_Binary_Sum_Scalar)
{
cv::GMat a, b;
auto p = cv::gapi::serialize(cv::GComputation(a, b, cv::gapi::sum(a + b)));
auto c = cv::gapi::deserialize<cv::GComputation>(p);
cv::Mat in_mat = cv::Mat::eye(32, 32, CV_8UC1);
cv::Scalar ref_scl = cv::sum(in_mat + in_mat);
cv::Scalar out_scl;
c.apply(in_mat, in_mat, out_scl);
EXPECT_EQ(out_scl, ref_scl);
out_scl = cv::Scalar();
auto cc = c.compile(cv::descr_of(in_mat), cv::descr_of(in_mat));
cc(in_mat, in_mat, out_scl);
EXPECT_EQ(out_scl, ref_scl);
}
TEST(S11N, Pipeline_Sharpen)
{
const cv::Size sz_in (1280, 720);
const cv::Size sz_out( 640, 480);
cv::Mat in_mat (sz_in, CV_8UC3);
in_mat = cv::Scalar(128, 33, 53);
cv::Mat out_mat(sz_out, CV_8UC3);
cv::Mat out_mat_y;
cv::Mat out_mat_ocv(sz_out, CV_8UC3);
float sharpen_coeffs[] = {
0.0f, -1.f, 0.0f,
-1.0f, 5.f, -1.0f,
0.0f, -1.f, 0.0f
};
cv::Mat sharpen_kernel(3, 3, CV_32F, sharpen_coeffs);
// G-API code //////////////////////////////////////////////////////////////
cv::GMat in;
auto vga = cv::gapi::resize(in, sz_out);
auto yuv = cv::gapi::RGB2YUV(vga);
auto yuv_p = cv::gapi::split3(yuv);
auto y_sharp = cv::gapi::filter2D(std::get<0>(yuv_p), -1, sharpen_kernel);
auto yuv_new = cv::gapi::merge3(y_sharp, std::get<1>(yuv_p), std::get<2>(yuv_p));
auto out = cv::gapi::YUV2RGB(yuv_new);
auto p = cv::gapi::serialize(cv::GComputation(cv::GIn(in), cv::GOut(y_sharp, out)));
auto c = cv::gapi::deserialize<cv::GComputation>(p);
c.apply(cv::gin(in_mat), cv::gout(out_mat_y, out_mat));
// OpenCV code /////////////////////////////////////////////////////////////
{
cv::Mat smaller;
cv::resize(in_mat, smaller, sz_out);
cv::Mat yuv_mat;
cv::cvtColor(smaller, yuv_mat, cv::COLOR_RGB2YUV);
std::vector<cv::Mat> yuv_planar(3);
cv::split(yuv_mat, yuv_planar);
cv::filter2D(yuv_planar[0], yuv_planar[0], -1, sharpen_kernel);
cv::merge(yuv_planar, yuv_mat);
cv::cvtColor(yuv_mat, out_mat_ocv, cv::COLOR_YUV2RGB);
}
// Comparison //////////////////////////////////////////////////////////////
{
cv::Mat diff = out_mat_ocv != out_mat;
std::vector<cv::Mat> diffBGR(3);
cv::split(diff, diffBGR);
EXPECT_EQ(0, cvtest::norm(diffBGR[0], NORM_INF));
EXPECT_EQ(0, cvtest::norm(diffBGR[1], NORM_INF));
EXPECT_EQ(0, cvtest::norm(diffBGR[2], NORM_INF));
}
// Metadata check /////////////////////////////////////////////////////////
{
auto cc = c.compile(cv::descr_of(in_mat));
auto metas = cc.outMetas();
ASSERT_EQ(2u, metas.size());
auto out_y_meta = cv::util::get<cv::GMatDesc>(metas[0]);
auto out_meta = cv::util::get<cv::GMatDesc>(metas[1]);
// Y-output
EXPECT_EQ(CV_8U, out_y_meta.depth);
EXPECT_EQ(1, out_y_meta.chan);
EXPECT_EQ(640, out_y_meta.size.width);
EXPECT_EQ(480, out_y_meta.size.height);
// Final output
EXPECT_EQ(CV_8U, out_meta.depth);
EXPECT_EQ(3, out_meta.chan);
EXPECT_EQ(640, out_meta.size.width);
EXPECT_EQ(480, out_meta.size.height);
}
}
TEST(S11N, Pipeline_CustomRGB2YUV)
{
const cv::Size sz(1280, 720);
const int INS = 3;
std::vector<cv::Mat> in_mats(INS);
for (auto i : ade::util::iota(INS))
{
in_mats[i].create(sz, CV_8U);
cv::randu(in_mats[i], cv::Scalar::all(0), cv::Scalar::all(255));
}
const int OUTS = 3;
std::vector<cv::Mat> out_mats_cv(OUTS);
std::vector<cv::Mat> out_mats_gapi(OUTS);
for (auto i : ade::util::iota(OUTS))
{
out_mats_cv[i].create(sz, CV_8U);
out_mats_gapi[i].create(sz, CV_8U);
}
// G-API code //////////////////////////////////////////////////////////////
{
cv::GMat r, g, b;
cv::GMat y = 0.299f*r + 0.587f*g + 0.114f*b;
cv::GMat u = 0.492f*(b - y);
cv::GMat v = 0.877f*(r - y);
auto p = cv::gapi::serialize(cv::GComputation({r, g, b}, {y, u, v}));
auto c = cv::gapi::deserialize<cv::GComputation>(p);
c.apply(in_mats, out_mats_gapi);
}
// OpenCV code /////////////////////////////////////////////////////////////
{
cv::Mat r = in_mats[0], g = in_mats[1], b = in_mats[2];
cv::Mat y = 0.299f*r + 0.587f*g + 0.114f*b;
cv::Mat u = 0.492f*(b - y);
cv::Mat v = 0.877f*(r - y);
out_mats_cv[0] = y;
out_mats_cv[1] = u;
out_mats_cv[2] = v;
}
// Comparison //////////////////////////////////////////////////////////////
{
const auto diff = [](cv::Mat m1, cv::Mat m2, int t) {
return cv::abs(m1 - m2) > t;
};
// FIXME: Not bit-accurate even now!
cv::Mat
diff_y = diff(out_mats_cv[0], out_mats_gapi[0], 2),
diff_u = diff(out_mats_cv[1], out_mats_gapi[1], 2),
diff_v = diff(out_mats_cv[2], out_mats_gapi[2], 2);
EXPECT_EQ(0, cvtest::norm(diff_y, NORM_INF));
EXPECT_EQ(0, cvtest::norm(diff_u, NORM_INF));
EXPECT_EQ(0, cvtest::norm(diff_v, NORM_INF));
}
}
} // namespace opencv_test