mirror of https://github.com/alibaba/MNN.git
83 lines
2.9 KiB
C++
83 lines
2.9 KiB
C++
//
|
|
// ShapeResize.cpp
|
|
// MNN
|
|
//
|
|
// Created by MNN on 2019/01/10.
|
|
// Copyright © 2018, Alibaba Group Holding Limited
|
|
//
|
|
|
|
#include "shape/SizeComputer.hpp"
|
|
#include "core/Macro.h"
|
|
|
|
namespace MNN {
|
|
// Size Computer
|
|
class ResizeComputer : public SizeComputer {
|
|
virtual bool onComputeSize(const MNN::Op *op, const std::vector<Tensor *> &inputs,
|
|
const std::vector<Tensor *> &outputs) const override {
|
|
MNN_ASSERT(1 == inputs.size());
|
|
MNN_ASSERT(1 == outputs.size());
|
|
|
|
// copy dims
|
|
auto resize = op->main_as_Resize();
|
|
auto &input = inputs[0]->buffer();
|
|
auto &output = outputs[0]->buffer();
|
|
TensorUtils::copyShape(inputs[0], outputs[0], true);
|
|
|
|
// set dims
|
|
output.dim[3].extent = input.dim[3].extent * resize->xScale();
|
|
output.dim[2].extent = input.dim[2].extent * resize->yScale();
|
|
output.type = inputs[0]->getType();
|
|
|
|
return true;
|
|
}
|
|
virtual float onComputeFlops(const MNN::Op *op, const std::vector<Tensor *> &inputs,
|
|
const std::vector<Tensor *> &outputs) const override {
|
|
return (float)outputs[0]->elementSize() / 1024.0f / 1024.0f * 4;
|
|
}
|
|
};
|
|
|
|
class ImageProcessComputer : public SizeComputer {
|
|
virtual bool onComputeSize(const MNN::Op *op, const std::vector<Tensor *> &inputs,
|
|
const std::vector<Tensor *> &outputs) const override {
|
|
MNN_ASSERT(1 == inputs.size() || inputs.size() == 3);
|
|
MNN_ASSERT(1 == outputs.size());
|
|
if (inputs.size() == 3) {
|
|
auto &output = outputs[0]->buffer();
|
|
output.dimensions = 1;
|
|
output.dim[0].extent = 1;
|
|
return true;
|
|
}
|
|
|
|
// copy dims
|
|
auto &input = inputs[0]->buffer();
|
|
auto &output = outputs[0]->buffer();
|
|
TensorUtils::copyShape(inputs[0], outputs[0], true);
|
|
|
|
// set dims
|
|
auto process = op->main_as_ImageProcessParam();
|
|
int c = process->shape()->Get(1);
|
|
int h = process->shape()->Get(2);
|
|
int w = process->shape()->Get(3);
|
|
if (MNN_DATA_FORMAT_NHWC == TensorUtils::getDescribe(inputs[0])->dimensionFormat) {
|
|
output.dim[1].extent = h;
|
|
output.dim[2].extent = w;
|
|
output.dim[3].extent = c;
|
|
} else {
|
|
output.dim[1].extent = c;
|
|
output.dim[2].extent = h;
|
|
output.dim[3].extent = w;
|
|
}
|
|
// set dtype
|
|
outputs[0]->setType(process->outputType());
|
|
return true;
|
|
}
|
|
virtual float onComputeFlops(const MNN::Op *op, const std::vector<Tensor *> &inputs,
|
|
const std::vector<Tensor *> &outputs) const override {
|
|
return (float)outputs[0]->elementSize() / 1024.0f / 1024.0f * 4;
|
|
}
|
|
};
|
|
|
|
REGISTER_SHAPE(ResizeComputer, OpType_Resize);
|
|
REGISTER_SHAPE(ImageProcessComputer, OpType_ImageProcess);
|
|
} // namespace MNN
|