IntenseWebs/openvino
3
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

[GPU] Added NV12toBGR single plane surface tests (#15417)

Browse Source
This commit is contained in:
Roman Lyamin 2023-02-01 18:26:39 +04:00 committed by GitHub
parent 1550a98bd2
commit 2d9a213ed6
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
2 changed files with 220 additions and 24 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
src/plugins/intel_gpu/src/plugin/ops/convert_color.cpp
View File

@ -36,10 +36,10 @@ static void CreateCommonConvertColorOp(Program& p, const std::shared_ptr<ngraph:
if (outShape.batch[0] > 1 && memory_type == cldnn::convert_color::memory_type::image) {
std::vector<cldnn::input_info> convert_color_names;
for (int b = 0; b < outShape.batch[0]; ++b) {
cldnn::primitive::input_info_arr batched_inputs = {
cldnn::input_info(inputs[0].pid + "_" + std::to_string(b), inputs[0].idx),
cldnn::input_info(inputs[1].pid + "_" + std::to_string(b), inputs[1].idx)
};
cldnn::primitive::input_info_arr batched_inputs;
for (size_t i = 0; i < inputs.size(); ++i) {
batched_inputs.emplace_back(cldnn::input_info(inputs[i].pid + "_" + std::to_string(b), inputs[i].idx));
}
cldnn::primitive_id batched_prim_id = layerName + "_" + std::to_string(b);
convert_color_names.emplace_back(cldnn::input_info(batched_prim_id));
auto new_shape = outShape;

236
src/tests/functional/plugin/gpu/remote_blob_tests/gpu_remote_tensor_tests.cpp
View File

@ -845,9 +845,9 @@ TEST_F(OVRemoteTensor_Test, NV12toBGR_image_ConvertTranspose) {
using namespace ov::preprocess;
auto p = PrePostProcessor(fn_ptr_remote);
p.input().tensor().set_element_type(ov::element::u8)
.set_color_format(ov::preprocess::ColorFormat::NV12_TWO_PLANES, {"y", "uv"})
.set_color_format(ColorFormat::NV12_TWO_PLANES, {"y", "uv"})
.set_memory_type(GPU_CONFIG_KEY(SURFACE));
p.input().preprocess().convert_color(ov::preprocess::ColorFormat::BGR);
p.input().preprocess().convert_color(ColorFormat::BGR);
p.input().model().set_layout("NCHW");
auto function = p.build();
@ -910,9 +910,9 @@ TEST_F(OVRemoteTensor_Test, NV12toBGR_image_ConvertTranspose) {
using namespace ov::preprocess;
auto p_reg = PrePostProcessor(fn_ptr_regular);
p_reg.input().tensor().set_element_type(ov::element::u8)
.set_color_format(ov::preprocess::ColorFormat::NV12_TWO_PLANES, {"y", "uv"})
.set_color_format(ColorFormat::NV12_TWO_PLANES, {"y", "uv"})
.set_memory_type(GPU_CONFIG_KEY(BUFFER));
p_reg.input().preprocess().convert_color(ov::preprocess::ColorFormat::BGR);
p_reg.input().preprocess().convert_color(ColorFormat::BGR);
p_reg.input().model().set_layout("NCHW");
auto function_regular = p_reg.build();
@ -933,7 +933,97 @@ TEST_F(OVRemoteTensor_Test, NV12toBGR_image_ConvertTranspose) {
FuncTestUtils::compare_tensor(output_tensor_shared, output_tensor_regular, thr);
}
TEST_F(OVRemoteTensor_Test, NV12toBGR_image) {
TEST_F(OVRemoteTensor_Test, NV12toBGR_image_single_plane) {
#if defined(ANDROID)
GTEST_SKIP();
#endif
const int height = 16;
const int width = 16;
// ------------------------------------------------------
// Prepare input data
ov::Tensor fake_image_data_yuv = FuncTestUtils::create_and_fill_tensor(ov::element::u8, {1, 1, height * 3 / 2, width}, 50);
auto ie = ov::Core();
// ------------------------------------------------------
// inference using remote tensor
auto fn_ptr_remote = ngraph::builder::subgraph::makeConvPoolRelu({1, 3, height, width});
using namespace ov::preprocess;
auto p = PrePostProcessor(fn_ptr_remote);
p.input().tensor().set_element_type(ov::element::u8)
.set_color_format(ColorFormat::NV12_SINGLE_PLANE)
.set_memory_type(GPU_CONFIG_KEY(SURFACE));
p.input().preprocess().convert_color(ColorFormat::BGR);
p.input().model().set_layout("NCHW");
auto function = p.build();
auto param_input_yuv = fn_ptr_remote->get_parameters().at(0);
auto exec_net_b = ie.compile_model(function, CommonTestUtils::DEVICE_GPU);
auto inf_req_remote = exec_net_b.create_infer_request();
auto cldnn_context = exec_net_b.get_context().as<ov::intel_gpu::ocl::ClContext>();
cl_context ctx = cldnn_context.get();
auto ocl_instance = std::make_shared<OpenCL>(ctx);
cl_int err;
cl_image_format image_format;
cl_image_desc image_desc = { 0 };
image_format.image_channel_order = CL_R;
image_format.image_channel_data_type = CL_UNORM_INT8;
image_desc.image_type = CL_MEM_OBJECT_IMAGE2D;
image_desc.image_width = width;
image_desc.image_height = height * 3 / 2;
cl_mem nv12_image_plane_yuv = clCreateImage(ocl_instance->_context.get(), CL_MEM_READ_WRITE, &image_format, &image_desc, NULL, &err);
ASSERT_EQ(err, 0);
size_t origin[3] = { 0, 0, 0 };
size_t yuv_region[3] = { (size_t)width, (size_t)height * 3 / 2, 1 };
err = clEnqueueWriteImage(ocl_instance->_queue.get(), nv12_image_plane_yuv,
true, origin, yuv_region, 0, 0, fake_image_data_yuv.data(), 0, NULL, NULL);
ASSERT_EQ(err, 0);
cl::Image2D img_yuv = cl::Image2D(nv12_image_plane_yuv);
auto tensor_remote_yuv = cldnn_context.create_tensor(param_input_yuv->get_element_type(), fake_image_data_yuv.get_shape(), img_yuv);
inf_req_remote.set_tensor(*param_input_yuv->output(0).get_tensor().get_names().begin(), tensor_remote_yuv);
inf_req_remote.infer();
auto output_tensor_shared = inf_req_remote.get_tensor(function->get_results().at(0));
// ------------------------------------------------------
// regular inference
auto fn_ptr_regular = ngraph::builder::subgraph::makeConvPoolRelu({1, 3, height, width});
using namespace ov::preprocess;
auto p_reg = PrePostProcessor(fn_ptr_regular);
p_reg.input().tensor().set_element_type(ov::element::u8)
.set_color_format(ColorFormat::NV12_SINGLE_PLANE)
.set_memory_type(GPU_CONFIG_KEY(BUFFER));
p_reg.input().preprocess().convert_color(ColorFormat::BGR);
p_reg.input().model().set_layout("NCHW");
auto function_regular = p_reg.build();
auto exec_net_regular = ie.compile_model(function_regular, CommonTestUtils::DEVICE_GPU);
auto inf_req_regular = exec_net_regular.create_infer_request();
inf_req_regular.set_tensor(param_input_yuv, fake_image_data_yuv);
inf_req_regular.infer();
auto output_tensor_regular = inf_req_regular.get_tensor(exec_net_regular.output());
// ------------------------------------------------------
// compare results
ASSERT_EQ(output_tensor_regular.get_size(), output_tensor_shared.get_size());
ASSERT_NO_THROW(output_tensor_regular.data());
ASSERT_NO_THROW(output_tensor_shared.data());
float thr = 0.1;
FuncTestUtils::compare_tensor(output_tensor_shared, output_tensor_regular, thr);
}
TEST_F(OVRemoteTensor_Test, NV12toBGR_image_two_planes) {
#if defined(ANDROID)
GTEST_SKIP();
#endif
@ -954,9 +1044,9 @@ TEST_F(OVRemoteTensor_Test, NV12toBGR_image) {
using namespace ov::preprocess;
auto p = PrePostProcessor(fn_ptr_remote);
p.input().tensor().set_element_type(ov::element::u8)
.set_color_format(ov::preprocess::ColorFormat::NV12_TWO_PLANES, {"y", "uv"})
.set_color_format(ColorFormat::NV12_TWO_PLANES, {"y", "uv"})
.set_memory_type(GPU_CONFIG_KEY(SURFACE));
p.input().preprocess().convert_color(ov::preprocess::ColorFormat::BGR);
p.input().preprocess().convert_color(ColorFormat::BGR);
p.input().model().set_layout("NCHW");
auto function = p.build();
@ -1019,9 +1109,9 @@ TEST_F(OVRemoteTensor_Test, NV12toBGR_image) {
using namespace ov::preprocess;
auto p_reg = PrePostProcessor(fn_ptr_regular);
p_reg.input().tensor().set_element_type(ov::element::u8)
.set_color_format(ov::preprocess::ColorFormat::NV12_TWO_PLANES, {"y", "uv"})
.set_color_format(ColorFormat::NV12_TWO_PLANES, {"y", "uv"})
.set_memory_type(GPU_CONFIG_KEY(BUFFER));
p_reg.input().preprocess().convert_color(ov::preprocess::ColorFormat::BGR);
p_reg.input().preprocess().convert_color(ColorFormat::BGR);
p_reg.input().model().set_layout("NCHW");
auto function_regular = p_reg.build();
@ -1061,9 +1151,9 @@ TEST_F(OVRemoteTensor_Test, NV12toBGR_buffer) {
using namespace ov::preprocess;
auto p = PrePostProcessor(fn_ptr_remote);
p.input().tensor().set_element_type(ov::element::u8)
.set_color_format(ov::preprocess::ColorFormat::NV12_TWO_PLANES, {"y", "uv"})
.set_color_format(ColorFormat::NV12_TWO_PLANES, {"y", "uv"})
.set_memory_type(GPU_CONFIG_KEY(BUFFER));
p.input().preprocess().convert_color(ov::preprocess::ColorFormat::BGR);
p.input().preprocess().convert_color(ColorFormat::BGR);
p.input().model().set_layout("NCHW");
auto function = p.build();
@ -1142,7 +1232,113 @@ protected:
std::vector<std::shared_ptr<ngraph::Function>> fn_ptrs;
};
TEST_P(OVRemoteTensorBatched_Test, NV12toBGR_image) {
TEST_P(OVRemoteTensorBatched_Test, NV12toBGR_image_single_plane) {
#if defined(ANDROID)
GTEST_SKIP();
#endif
const int height = 16;
const int width = 16;
// ------------------------------------------------------
// Prepare input data
std::vector<ov::Tensor> fake_image_data_yuv;
for (int i = 0; i < num_batch; i++) {
fake_image_data_yuv.push_back(FuncTestUtils::create_and_fill_tensor(ov::element::u8, {1, 1, height * 3 / 2, width}, 50, 0, 1, i));
}
auto ie = ov::Core();
// ------------------------------------------------------
// inference using remote tensor
auto fn_ptr_remote = ngraph::builder::subgraph::makeConvPoolRelu({num_batch, 3, height, width});
using namespace ov::preprocess;
auto p = PrePostProcessor(fn_ptr_remote);
p.input().tensor().set_element_type(ov::element::u8)
.set_color_format(ColorFormat::NV12_SINGLE_PLANE)
.set_memory_type(GPU_CONFIG_KEY(SURFACE));
p.input().preprocess().convert_color(ColorFormat::BGR);
p.input().model().set_layout("NCHW");
auto function = p.build();
auto param_input_yuv = fn_ptr_remote->get_parameters().at(0);
auto exec_net_b = ie.compile_model(function, CommonTestUtils::DEVICE_GPU);
auto inf_req_remote = exec_net_b.create_infer_request();
auto cldnn_context = exec_net_b.get_context().as<ov::intel_gpu::ocl::ClContext>();
cl_context ctx = cldnn_context.get();
auto ocl_instance = std::make_shared<OpenCL>(ctx);
cl_int err;
std::vector<cl_mem> nv12_image_plane_yuv;
std::vector<cl::Image2D> img_yuv;
std::vector<ov::Tensor> tensor_remote_yuv;
for (size_t i = 0; i < num_batch; ++i) {
cl_image_format image_format;
cl_image_desc image_desc = { 0 };
image_format.image_channel_order = CL_R;
image_format.image_channel_data_type = CL_UNORM_INT8;
image_desc.image_type = CL_MEM_OBJECT_IMAGE2D;
image_desc.image_width = width;
image_desc.image_height = height * 3 / 2;
nv12_image_plane_yuv.emplace_back(clCreateImage(ocl_instance->_context.get(), CL_MEM_READ_WRITE, &image_format, &image_desc, NULL, &err));
ASSERT_EQ(err, 0);
size_t origin[3] = { 0, 0, 0 };
size_t yuv_region[3] = { (size_t)width, (size_t)height * 3 / 2, 1 };
err = clEnqueueWriteImage(ocl_instance->_queue.get(), nv12_image_plane_yuv[i],
true, origin, yuv_region, 0, 0, fake_image_data_yuv[i].data(), 0, NULL, NULL);
ASSERT_EQ(err, 0);
img_yuv.emplace_back(nv12_image_plane_yuv[i]);
tensor_remote_yuv.emplace_back(cldnn_context.create_tensor(param_input_yuv->get_element_type(), fake_image_data_yuv[i].get_shape(), img_yuv[i]));
}
for (size_t i = 0; i < 5; ++i) { // to test repeating set_tensors/infer functionality
inf_req_remote.set_tensors(*param_input_yuv->output(0).get_tensor().get_names().begin(), tensor_remote_yuv);
inf_req_remote.infer();
}
auto output_tensor_shared = inf_req_remote.get_tensor(function->get_results().at(0));
ASSERT_NO_THROW(output_tensor_shared.data());
// ------------------------------------------------------
// regular inference
auto fn_ptr_regular = ngraph::builder::subgraph::makeConvPoolRelu({1, 3, height, width});
using namespace ov::preprocess;
auto p_reg = PrePostProcessor(fn_ptr_regular);
p_reg.input().tensor().set_element_type(ov::element::u8)
.set_color_format(ColorFormat::NV12_SINGLE_PLANE)
.set_memory_type(GPU_CONFIG_KEY(BUFFER));
p_reg.input().preprocess().convert_color(ColorFormat::BGR);
p_reg.input().model().set_layout("NCHW");
auto function_regular = p_reg.build();
auto param_input_yuv_reg = fn_ptr_regular->get_parameters().at(0);
auto exec_net_regular = ie.compile_model(function_regular, CommonTestUtils::DEVICE_GPU);
auto inf_req_regular = exec_net_regular.create_infer_request();
for (size_t i = 0; i < num_batch; ++i) {
inf_req_regular.set_tensor(param_input_yuv_reg, fake_image_data_yuv[i]);
inf_req_regular.infer();
auto output_tensor_regular = inf_req_regular.get_tensor(exec_net_regular.output());
ASSERT_EQ(output_tensor_regular.get_size() * num_batch, output_tensor_shared.get_size());
float thr = 0.1;
FuncTestUtils::compareRawBuffers<float>(static_cast<float*>(output_tensor_shared.data()) + i * output_tensor_regular.get_size(),
static_cast<float*>(output_tensor_regular.data()),
output_tensor_regular.get_size(), output_tensor_regular.get_size(), thr);
}
}
TEST_P(OVRemoteTensorBatched_Test, NV12toBGR_image_two_planes) {
#if defined(ANDROID)
GTEST_SKIP();
#endif
@ -1166,9 +1362,9 @@ TEST_P(OVRemoteTensorBatched_Test, NV12toBGR_image) {
using namespace ov::preprocess;
auto p = PrePostProcessor(fn_ptr_remote);
p.input().tensor().set_element_type(ov::element::u8)
.set_color_format(ov::preprocess::ColorFormat::NV12_TWO_PLANES, {"y", "uv"})
.set_color_format(ColorFormat::NV12_TWO_PLANES, {"y", "uv"})
.set_memory_type(GPU_CONFIG_KEY(SURFACE));
p.input().preprocess().convert_color(ov::preprocess::ColorFormat::BGR);
p.input().preprocess().convert_color(ColorFormat::BGR);
p.input().model().set_layout("NCHW");
auto function = p.build();
@ -1240,9 +1436,9 @@ TEST_P(OVRemoteTensorBatched_Test, NV12toBGR_image) {
using namespace ov::preprocess;
auto p_reg = PrePostProcessor(fn_ptr_regular);
p_reg.input().tensor().set_element_type(ov::element::u8)
.set_color_format(ov::preprocess::ColorFormat::NV12_TWO_PLANES, {"y", "uv"})
.set_color_format(ColorFormat::NV12_TWO_PLANES, {"y", "uv"})
.set_memory_type(GPU_CONFIG_KEY(BUFFER));
p_reg.input().preprocess().convert_color(ov::preprocess::ColorFormat::BGR);
p_reg.input().preprocess().convert_color(ColorFormat::BGR);
p_reg.input().model().set_layout("NCHW");
auto function_regular = p_reg.build();
@ -1409,9 +1605,9 @@ TEST_P(OVRemoteTensorBatched_Test, NV12toBGR_buffer) {
using namespace ov::preprocess;
auto p = PrePostProcessor(fn_ptr_remote);
p.input().tensor().set_element_type(ov::element::u8)
.set_color_format(ov::preprocess::ColorFormat::NV12_TWO_PLANES, {"y", "uv"})
.set_color_format(ColorFormat::NV12_TWO_PLANES, {"y", "uv"})
.set_memory_type(GPU_CONFIG_KEY(BUFFER));
p.input().preprocess().convert_color(ov::preprocess::ColorFormat::BGR);
p.input().preprocess().convert_color(ColorFormat::BGR);
p.input().model().set_layout("NCHW");
auto function = p.build();
@ -1511,9 +1707,9 @@ TEST_P(OVRemoteTensorBatched_Test, NV12toBGR_buffer) {
using namespace ov::preprocess;
auto p_reg = PrePostProcessor(fn_ptr_regular);
p_reg.input().tensor().set_element_type(ov::element::u8)
.set_color_format(ov::preprocess::ColorFormat::NV12_TWO_PLANES, {"y", "uv"})
.set_color_format(ColorFormat::NV12_TWO_PLANES, {"y", "uv"})
.set_memory_type(GPU_CONFIG_KEY(BUFFER));
p_reg.input().preprocess().convert_color(ov::preprocess::ColorFormat::BGR);
p_reg.input().preprocess().convert_color(ColorFormat::BGR);
p_reg.input().model().set_layout("NCHW");
auto function_regular = p_reg.build();