Move all eigen serialization to basalt-headers.

**Note:** Binary representation of eigen matrices has changed!
This commit is contained in:
Nikolaus Demmel 2019-08-27 15:19:33 +00:00 committed by Vladyslav Usenko
parent 77100b99f2
commit cae794d6af
5 changed files with 1 additions and 855 deletions

View File

@ -155,39 +155,6 @@ class DatasetIoFactory {
namespace cereal {
template <class Archive, class _Scalar, int _Rows, int _Cols, int _Options,
int _MaxRows, int _MaxCols>
static inline
typename std::enable_if<_Rows == Eigen::Dynamic || _Cols == Eigen::Dynamic,
void>::type
save(Archive &ar, const Eigen::Matrix<_Scalar, _Rows, _Cols, _Options,
_MaxRows, _MaxCols> &matrix) {
const std::int32_t rows = static_cast<std::int32_t>(matrix.rows());
const std::int32_t cols = static_cast<std::int32_t>(matrix.cols());
ar(rows);
ar(cols);
ar(binary_data(matrix.data(), rows * cols * sizeof(_Scalar)));
};
template <class Archive, class _Scalar, int _Rows, int _Cols, int _Options,
int _MaxRows, int _MaxCols>
static inline
typename std::enable_if<_Rows == Eigen::Dynamic || _Cols == Eigen::Dynamic,
void>::type
load(Archive &ar,
Eigen::Matrix<_Scalar, _Rows, _Cols, _Options, _MaxRows, _MaxCols>
&matrix) {
std::int32_t rows;
std::int32_t cols;
ar(rows);
ar(cols);
matrix.resize(rows, cols);
ar(binary_data(matrix.data(),
static_cast<std::size_t>(rows * cols * sizeof(_Scalar))));
};
template <class Archive>
void serialize(Archive &archive, basalt::ManagedImage<uint8_t> &m) {
archive(m.w);

View File

@ -1,688 +0,0 @@
/**
BSD 3-Clause License
This file is part of the Basalt project.
https://gitlab.com/VladyslavUsenko/basalt.git
Copyright (c) 2019, Vladyslav Usenko and Nikolaus Demmel.
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright notice, this
list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright notice,
this list of conditions and the following disclaimer in the documentation
and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the names of its
contributors may be used to endorse or promote products derived from
this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
// This file is adapted from Pangolin. Original license:
/* This file is part of the Pangolin Project.
* http://github.com/stevenlovegrove/Pangolin
*
* Copyright (c) 2011 Steven Lovegrove
*
* Permission is hereby granted, free of charge, to any person
* obtaining a copy of this software and associated documentation
* files (the "Software"), to deal in the Software without
* restriction, including without limitation the rights to use,
* copy, modify, merge, publish, distribute, sublicense, and/or sell
* copies of the Software, and to permit persons to whom the
* Software is furnished to do so, subject to the following
* conditions:
*
* The above copyright notice and this permission notice shall be
* included in all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
* OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
* HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
* WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
* FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
* OTHER DEALINGS IN THE SOFTWARE.
*/
#pragma once
#include <memory>
#include <basalt/utils/assert.h>
// Renamed Pangoling defines to avoid clash
#define BASALT_HOST_DEVICE
#define BASALT_EXTENSION_IMAGE
#ifdef BASALT_ENABLE_BOUNDS_CHECKS
#define BASALT_BOUNDS_ASSERT(...) BASALT_ASSERT(##__VA_ARGS__)
#else
#define BASALT_BOUNDS_ASSERT(...) ((void)0)
#endif
namespace basalt {
template <typename T>
struct CopyObject {
CopyObject(const T& obj) : obj(obj) {}
const T& obj;
};
inline void PitchedCopy(char* dst, unsigned int dst_pitch_bytes,
const char* src, unsigned int src_pitch_bytes,
unsigned int width_bytes, unsigned int height) {
if (dst_pitch_bytes == width_bytes && src_pitch_bytes == width_bytes) {
std::memcpy(dst, src, height * width_bytes);
} else {
for (unsigned int row = 0; row < height; ++row) {
std::memcpy(dst, src, width_bytes);
dst += dst_pitch_bytes;
src += src_pitch_bytes;
}
}
}
template <typename T>
struct Image {
inline Image() : pitch(0), ptr(0), w(0), h(0) {}
inline Image(T* ptr, size_t w, size_t h, size_t pitch)
: pitch(pitch), ptr(ptr), w(w), h(h) {}
BASALT_HOST_DEVICE inline size_t SizeBytes() const { return pitch * h; }
BASALT_HOST_DEVICE inline size_t Area() const { return w * h; }
BASALT_HOST_DEVICE inline bool IsValid() const { return ptr != 0; }
BASALT_HOST_DEVICE inline bool IsContiguous() const {
return w * sizeof(T) == pitch;
}
//////////////////////////////////////////////////////
// Iterators
//////////////////////////////////////////////////////
BASALT_HOST_DEVICE inline T* begin() { return ptr; }
BASALT_HOST_DEVICE inline T* end() { return RowPtr(h - 1) + w; }
BASALT_HOST_DEVICE inline const T* begin() const { return ptr; }
BASALT_HOST_DEVICE inline const T* end() const { return RowPtr(h - 1) + w; }
BASALT_HOST_DEVICE inline size_t size() const { return w * h; }
//////////////////////////////////////////////////////
// Image transforms
//////////////////////////////////////////////////////
template <typename UnaryOperation>
BASALT_HOST_DEVICE inline void Transform(UnaryOperation unary_op) {
BASALT_ASSERT(IsValid());
for (size_t y = 0; y < h; ++y) {
T* el = RowPtr(y);
const T* el_end = el + w;
for (; el != el_end; ++el) {
*el = unary_op(*el);
}
}
}
BASALT_HOST_DEVICE inline void Fill(const T& val) {
Transform([&](const T&) { return val; });
}
BASALT_HOST_DEVICE inline void Replace(const T& oldval, const T& newval) {
Transform([&](const T& val) { return (val == oldval) ? newval : val; });
}
inline void Memset(unsigned char v = 0) {
BASALT_ASSERT(IsValid());
if (IsContiguous()) {
std::memset((char*)ptr, v, pitch * h);
} else {
for (size_t y = 0; y < h; ++y) {
std::memset((char*)RowPtr(y), v, pitch);
}
}
}
inline void CopyFrom(const Image<T>& img) {
if (IsValid() && img.IsValid()) {
BASALT_ASSERT(w >= img.w && h >= img.h);
PitchedCopy((char*)ptr, pitch, (char*)img.ptr, img.pitch,
std::min(img.w, w) * sizeof(T), std::min(img.h, h));
} else if (img.IsValid() != IsValid()) {
BASALT_ASSERT(false && "Cannot copy from / to an unasigned image.");
}
}
//////////////////////////////////////////////////////
// Reductions
//////////////////////////////////////////////////////
template <typename BinaryOperation>
BASALT_HOST_DEVICE inline T Accumulate(const T init,
BinaryOperation binary_op) {
BASALT_ASSERT(IsValid());
T val = init;
for (size_t y = 0; y < h; ++y) {
T* el = RowPtr(y);
const T* el_end = el + w;
for (; el != el_end; ++el) {
val = binary_op(val, *el);
}
}
return val;
}
std::pair<T, T> MinMax() const {
BASALT_ASSERT(IsValid());
std::pair<T, T> minmax(std::numeric_limits<T>::max(),
std::numeric_limits<T>::lowest());
for (size_t r = 0; r < h; ++r) {
const T* ptr = RowPtr(r);
const T* end = ptr + w;
while (ptr != end) {
minmax.first = std::min(*ptr, minmax.first);
minmax.second = std::max(*ptr, minmax.second);
++ptr;
}
}
return minmax;
}
template <typename Tout = T>
Tout Sum() const {
return Accumulate((T)0,
[](const T& lhs, const T& rhs) { return lhs + rhs; });
}
template <typename Tout = T>
Tout Mean() const {
return Sum<Tout>() / Area();
}
//////////////////////////////////////////////////////
// Direct Pixel Access
//////////////////////////////////////////////////////
BASALT_HOST_DEVICE inline T* RowPtr(size_t y) {
return (T*)((unsigned char*)(ptr) + y * pitch);
}
BASALT_HOST_DEVICE inline const T* RowPtr(size_t y) const {
return (T*)((unsigned char*)(ptr) + y * pitch);
}
BASALT_HOST_DEVICE inline T& operator()(size_t x, size_t y) {
BASALT_BOUNDS_ASSERT(InBounds(x, y));
return RowPtr(y)[x];
}
BASALT_HOST_DEVICE inline const T& operator()(size_t x, size_t y) const {
BASALT_BOUNDS_ASSERT(InBounds(x, y));
return RowPtr(y)[x];
}
template <typename TVec>
BASALT_HOST_DEVICE inline T& operator()(const TVec& p) {
BASALT_BOUNDS_ASSERT(InBounds(p[0], p[1]));
return RowPtr(p[1])[p[0]];
}
template <typename TVec>
BASALT_HOST_DEVICE inline const T& operator()(const TVec& p) const {
BASALT_BOUNDS_ASSERT(InBounds(p[0], p[1]));
return RowPtr(p[1])[p[0]];
}
BASALT_HOST_DEVICE inline T& operator[](size_t ix) {
BASALT_BOUNDS_ASSERT(InImage(ptr + ix));
return ptr[ix];
}
BASALT_HOST_DEVICE inline const T& operator[](size_t ix) const {
BASALT_BOUNDS_ASSERT(InImage(ptr + ix));
return ptr[ix];
}
//////////////////////////////////////////////////////
// Interpolated Pixel Access
//////////////////////////////////////////////////////
template <typename S>
inline S interp(const Eigen::Matrix<S, 2, 1>& p) const {
return interp<S>(p[0], p[1]);
}
template <typename S>
inline Eigen::Matrix<S, 3, 1> interpGrad(
const Eigen::Matrix<S, 2, 1>& p) const {
return interpGrad<S>(p[0], p[1]);
}
template <typename S>
inline float interp(S x, S y) const {
int ix = x;
int iy = y;
S dx = x - ix;
S dy = y - iy;
S ddx = 1.0f - dx;
S ddy = 1.0f - dy;
return ddx * ddy * (*this)(ix, iy) + ddx * dy * (*this)(ix, iy + 1) +
dx * ddy * (*this)(ix + 1, iy) + dx * dy * (*this)(ix + 1, iy + 1);
}
template <typename S>
inline Eigen::Matrix<S, 3, 1> interpGrad(S x, S y) const {
int ix = x;
int iy = y;
S dx = x - ix;
S dy = y - iy;
S ddx = 1.0f - dx;
S ddy = 1.0f - dy;
Eigen::Matrix<S, 3, 1> res;
const T& px0y0 = (*this)(ix, iy);
const T& px1y0 = (*this)(ix + 1, iy);
const T& px0y1 = (*this)(ix, iy + 1);
const T& px1y1 = (*this)(ix + 1, iy + 1);
res[0] = ddx * ddy * px0y0 + ddx * dy * px0y1 + dx * ddy * px1y0 +
dx * dy * px1y1;
const T& pxm1y0 = (*this)(ix - 1, iy);
const T& pxm1y1 = (*this)(ix - 1, iy + 1);
S res_mx = ddx * ddy * pxm1y0 + ddx * dy * pxm1y1 + dx * ddy * px0y0 +
dx * dy * px0y1;
const T& px2y0 = (*this)(ix + 2, iy);
const T& px2y1 = (*this)(ix + 2, iy + 1);
S res_px = ddx * ddy * px1y0 + ddx * dy * px1y1 + dx * ddy * px2y0 +
dx * dy * px2y1;
res[1] = 0.5 * (res_px - res_mx);
const T& px0ym1 = (*this)(ix, iy - 1);
const T& px1ym1 = (*this)(ix + 1, iy - 1);
S res_my = ddx * ddy * px0ym1 + ddx * dy * px0y0 + dx * ddy * px1ym1 +
dx * dy * px1y0;
const T& px0y2 = (*this)(ix, iy + 2);
const T& px1y2 = (*this)(ix + 1, iy + 2);
S res_py = ddx * ddy * px0y1 + ddx * dy * px0y2 + dx * ddy * px1y1 +
dx * dy * px1y2;
res[2] = 0.5 * (res_py - res_my);
return res;
}
//////////////////////////////////////////////////////
// Bounds Checking
//////////////////////////////////////////////////////
BASALT_HOST_DEVICE
bool InImage(const T* ptest) const {
return ptr <= ptest && ptest < RowPtr(h);
}
BASALT_HOST_DEVICE inline bool InBounds(int x, int y) const {
return 0 <= x && x < (int)w && 0 <= y && y < (int)h;
}
BASALT_HOST_DEVICE inline bool InBounds(float x, float y,
float border) const {
return border <= x && x < (w - border) && border <= y && y < (h - border);
}
template <typename TVec, typename TBorder>
BASALT_HOST_DEVICE inline bool InBounds(
const TVec& p, const TBorder border = (TBorder)0) const {
return border <= p[0] && p[0] < ((int)w - border) && border <= p[1] &&
p[1] < ((int)h - border);
}
//////////////////////////////////////////////////////
// Obtain slices / subimages
//////////////////////////////////////////////////////
BASALT_HOST_DEVICE inline const Image<const T> SubImage(size_t x, size_t y,
size_t width,
size_t height) const {
BASALT_ASSERT((x + width) <= w && (y + height) <= h);
return Image<const T>(RowPtr(y) + x, width, height, pitch);
}
BASALT_HOST_DEVICE inline Image<T> SubImage(size_t x, size_t y, size_t width,
size_t height) {
BASALT_ASSERT((x + width) <= w && (y + height) <= h);
return Image<T>(RowPtr(y) + x, width, height, pitch);
}
BASALT_HOST_DEVICE inline Image<T> Row(int y) const {
return SubImage(0, y, w, 1);
}
BASALT_HOST_DEVICE inline Image<T> Col(int x) const {
return SubImage(x, 0, 1, h);
}
//////////////////////////////////////////////////////
// Data mangling
//////////////////////////////////////////////////////
template <typename TRecast>
BASALT_HOST_DEVICE inline Image<TRecast> Reinterpret() const {
BASALT_ASSERT_STREAM(sizeof(TRecast) == sizeof(T),
"sizeof(TRecast) must match sizeof(T): "
<< sizeof(TRecast) << " != " << sizeof(T));
return UnsafeReinterpret<TRecast>();
}
template <typename TRecast>
BASALT_HOST_DEVICE inline Image<TRecast> UnsafeReinterpret() const {
return Image<TRecast>((TRecast*)ptr, w, h, pitch);
}
//////////////////////////////////////////////////////
// Deprecated methods
//////////////////////////////////////////////////////
// PANGOLIN_DEPRECATED inline
Image(size_t w, size_t h, size_t pitch, T* ptr)
: pitch(pitch), ptr(ptr), w(w), h(h) {}
// Use RAII/move aware pangolin::ManagedImage instead
// PANGOLIN_DEPRECATED inline
void Dealloc() {
if (ptr) {
::operator delete(ptr);
ptr = nullptr;
}
}
// Use RAII/move aware pangolin::ManagedImage instead
// PANGOLIN_DEPRECATED inline
void Alloc(size_t w, size_t h, size_t pitch) {
Dealloc();
this->w = w;
this->h = h;
this->pitch = pitch;
this->ptr = (T*)::operator new(h* pitch);
}
//////////////////////////////////////////////////////
// Data members
//////////////////////////////////////////////////////
size_t pitch;
T* ptr;
size_t w;
size_t h;
BASALT_EXTENSION_IMAGE
};
template <class T>
using DefaultImageAllocator = std::allocator<T>;
// Image that manages it's own memory, storing a strong pointer to it's memory
template <typename T, class Allocator = DefaultImageAllocator<T>>
class ManagedImage : public Image<T> {
public:
typedef std::shared_ptr<ManagedImage<T>> Ptr;
// Destructor
inline ~ManagedImage() { Deallocate(); }
// Null image
inline ManagedImage() {}
// Row image
inline ManagedImage(size_t w)
: Image<T>(Allocator().allocate(w), w, 1, w * sizeof(T)) {}
inline ManagedImage(size_t w, size_t h)
: Image<T>(Allocator().allocate(w * h), w, h, w * sizeof(T)) {}
inline ManagedImage(size_t w, size_t h, size_t pitch_bytes)
: Image<T>(Allocator().allocate((h * pitch_bytes) / sizeof(T) + 1), w, h,
pitch_bytes) {}
// Not copy constructable
inline ManagedImage(const ManagedImage<T>& other) = delete;
// Move constructor
inline ManagedImage(ManagedImage<T, Allocator>&& img) {
*this = std::move(img);
}
// Move asignment
inline void operator=(ManagedImage<T, Allocator>&& img) {
Deallocate();
Image<T>::pitch = img.pitch;
Image<T>::ptr = img.ptr;
Image<T>::w = img.w;
Image<T>::h = img.h;
img.ptr = nullptr;
}
// Explicit copy constructor
template <typename TOther>
ManagedImage(const CopyObject<TOther>& other) {
CopyFrom(other.obj);
}
// Explicit copy assignment
template <typename TOther>
void operator=(const CopyObject<TOther>& other) {
CopyFrom(other.obj);
}
inline void Swap(ManagedImage<T>& img) {
std::swap(img.pitch, Image<T>::pitch);
std::swap(img.ptr, Image<T>::ptr);
std::swap(img.w, Image<T>::w);
std::swap(img.h, Image<T>::h);
}
inline void CopyFrom(const Image<T>& img) {
if (!Image<T>::IsValid() || Image<T>::w != img.w || Image<T>::h != img.h) {
Reinitialise(img.w, img.h);
}
Image<T>::CopyFrom(img);
}
inline void Reinitialise(size_t w, size_t h) {
if (!Image<T>::ptr || Image<T>::w != w || Image<T>::h != h) {
*this = ManagedImage<T, Allocator>(w, h);
}
}
inline void Reinitialise(size_t w, size_t h, size_t pitch) {
if (!Image<T>::ptr || Image<T>::w != w || Image<T>::h != h ||
Image<T>::pitch != pitch) {
*this = ManagedImage<T, Allocator>(w, h, pitch);
}
}
inline void Deallocate() {
if (Image<T>::ptr) {
Allocator().deallocate(Image<T>::ptr,
(Image<T>::h * Image<T>::pitch) / sizeof(T));
Image<T>::ptr = nullptr;
}
}
// Move asignment
template <typename TOther, typename AllocOther>
inline void OwnAndReinterpret(ManagedImage<TOther, AllocOther>&& img) {
Deallocate();
Image<T>::pitch = img.pitch;
Image<T>::ptr = (T*)img.ptr;
Image<T>::w = img.w;
Image<T>::h = img.h;
img.ptr = nullptr;
}
template <typename T1>
inline void ConvertFrom(const ManagedImage<T1>& img) {
Reinitialise(img.w, img.h);
for (size_t j = 0; j < img.h; j++) {
T* this_row = this->RowPtr(j);
const T1* other_row = img.RowPtr(j);
for (size_t i = 0; i < img.w; i++) {
this_row[i] = T(other_row[i]);
}
}
}
inline void operator-=(const ManagedImage<T>& img) {
for (size_t j = 0; j < img.h; j++) {
T* this_row = this->RowPtr(j);
const T* other_row = img.RowPtr(j);
for (size_t i = 0; i < img.w; i++) {
this_row[i] -= other_row[i];
}
}
}
};
template <typename T, class Allocator = DefaultImageAllocator<T>>
class ManagedImagePyr {
public:
inline ManagedImagePyr() {}
inline ManagedImagePyr(ManagedImage<T>& other, size_t num_levels) {
setFromImage(other, num_levels);
}
inline void setFromImage(const ManagedImage<T>& other, size_t num_levels) {
orig_w = other.w;
image.Reinitialise(other.w + other.w / 2, other.h);
image.Fill(0);
lvl_internal(0).CopyFrom(other);
for (size_t i = 0; i < num_levels; i++) {
const Image<const T> l = lvl(i);
Image<T> lp1 = lvl_internal(i + 1);
subsample(l, lp1);
}
}
static inline int border101(int x, int h) {
return h - 1 - std::abs(h - 1 - x);
}
static void subsample(const Image<const T>& img, Image<T>& img_sub) {
static_assert(std::is_same<T, uint16_t>::value ||
std::is_same<T, uint8_t>::value);
constexpr int kernel[5] = {1, 4, 6, 4, 1};
// accumulator
ManagedImage<int> tmp(img_sub.h, img.w);
// Vertical convolution
{
for (int r = 0; r < int(img_sub.h); r++) {
const T* row_m2 = img.RowPtr(std::abs(2 * r - 2));
const T* row_m1 = img.RowPtr(std::abs(2 * r - 1));
const T* row = img.RowPtr(2 * r);
const T* row_p1 = img.RowPtr(border101(2 * r + 1, img.h));
const T* row_p2 = img.RowPtr(border101(2 * r + 2, img.h));
for (int c = 0; c < int(img.w); c++) {
tmp(r, c) = kernel[0] * int(row_m2[c]) + kernel[1] * int(row_m1[c]) +
kernel[2] * int(row[c]) + kernel[3] * int(row_p1[c]) +
kernel[4] * int(row_p2[c]);
}
}
}
// Horizontal convolution
{
for (int c = 0; c < int(img_sub.w); c++) {
const int* row_m2 = tmp.RowPtr(std::abs(2 * c - 2));
const int* row_m1 = tmp.RowPtr(std::abs(2 * c - 1));
const int* row = tmp.RowPtr(2 * c);
const int* row_p1 = tmp.RowPtr(border101(2 * c + 1, tmp.h));
const int* row_p2 = tmp.RowPtr(border101(2 * c + 2, tmp.h));
for (int r = 0; r < int(tmp.w); r++) {
int val_int = kernel[0] * row_m2[r] + kernel[1] * row_m1[r] +
kernel[2] * row[r] + kernel[3] * row_p1[r] +
kernel[4] * row_p2[r];
T val = ((val_int + (1 << 7)) >> 8);
img_sub(c, r) = val;
}
}
}
}
inline const Image<const T> lvl(size_t lvl) const {
size_t x = (lvl == 0) ? 0 : orig_w;
size_t y = (lvl <= 1) ? 0 : (image.h - (image.h >> (lvl - 1)));
size_t width = (orig_w >> lvl);
size_t height = (image.h >> lvl);
return image.SubImage(x, y, width, height);
}
template <typename S>
inline Eigen::Matrix<S, 2, 1> lvl_offset(size_t lvl) {
size_t x = (lvl == 0) ? 0 : orig_w;
size_t y = (lvl <= 1) ? 0 : (image.h - (image.h >> (lvl - 1)));
return Eigen::Matrix<S, 2, 1>(x, y);
}
private:
inline Image<T> lvl_internal(size_t lvl) {
size_t x = (lvl == 0) ? 0 : orig_w;
size_t y = (lvl <= 1) ? 0 : (image.h - (image.h >> (lvl - 1)));
size_t width = (orig_w >> lvl);
size_t height = (image.h >> lvl);
return image.SubImage(x, y, width, height);
}
size_t orig_w;
ManagedImage<T> image;
};
} // namespace basalt

View File

@ -5,9 +5,6 @@ cmake_minimum_required(VERSION 3.10)
include_directories(../thirdparty/basalt-headers/test/include)
add_executable(test_image src/test_image.cpp)
target_link_libraries(test_image gtest gtest_main basalt)
add_executable(test_spline_opt src/test_spline_opt.cpp)
target_link_libraries(test_spline_opt gtest gtest_main basalt)
@ -22,12 +19,10 @@ enable_testing()
include(GoogleTest)
#gtest_discover_tests(test_image DISCOVERY_TIMEOUT 60)
#gtest_discover_tests(test_spline_opt DISCOVERY_TIMEOUT 60)
#gtest_discover_tests(test_vio DISCOVERY_TIMEOUT 60)
#gtest_discover_tests(test_nfr DISCOVERY_TIMEOUT 60)
gtest_add_tests(TARGET test_image AUTO)
gtest_add_tests(TARGET test_spline_opt AUTO)
gtest_add_tests(TARGET test_vio AUTO)
gtest_add_tests(TARGET test_nfr AUTO)

View File

@ -1,128 +0,0 @@
#include <Eigen/Dense>
#include <basalt/utils/image.h>
#include "gtest/gtest.h"
#include "test_utils.h"
void setImageData(uint16_t *imageArray, int size) {
double norm = RAND_MAX;
norm /= (double)std::numeric_limits<uint16_t>::max();
for (int i = 0; i < size; i++) {
imageArray[i] = (unsigned char)(rand() / norm);
}
}
TEST(Pattern, ImageInterp) {
Eigen::Vector2d offset(231.234242, 123.23424);
basalt::ManagedImage<uint16_t> img(640, 480);
setImageData(img.ptr, img.size());
Eigen::Vector3d vg = img.interpGrad(offset);
Eigen::Matrix<double, 1, 2> J = vg.tail<2>();
// std::cerr << "vg\n" << vg << std::endl;
test_jacobian(
"d_val_d_p", J,
[&](const Eigen::Vector2d &x) {
Eigen::Matrix<double, 1, 1> res;
Eigen::Vector2d p1 = offset + x;
res[0] = img.interpGrad(p1)[0];
return res;
},
Eigen::Vector2d::Zero(), 1.0);
}
TEST(Image, ImageInterpolate) {
Eigen::Vector2i offset(231, 123);
basalt::ManagedImage<uint16_t> img(640, 480);
setImageData(img.ptr, img.size());
double eps = 1e-12;
double threshold = 1e-8;
{
Eigen::Vector2i pi = offset;
Eigen::Vector2d pd = pi.cast<double>() + Eigen::Vector2d(eps, eps);
uint16_t val1 = img(pi);
double val2 = img.interp(pd);
double val3 = img.interpGrad(pd)[0];
EXPECT_LE(std::abs(val2 - val1), threshold);
EXPECT_FLOAT_EQ(val2, val3);
}
{
Eigen::Vector2i pi = offset;
Eigen::Vector2d pd = pi.cast<double>() + Eigen::Vector2d(eps, eps);
uint16_t val1 = img(pi);
double val2 = img.interp(pd);
double val3 = img.interpGrad(pd)[0];
EXPECT_LE(std::abs(val2 - val1), threshold);
EXPECT_FLOAT_EQ(val2, val3);
}
{
Eigen::Vector2i pi = offset + Eigen::Vector2i(1, 0);
Eigen::Vector2d pd = pi.cast<double>() + Eigen::Vector2d(-eps, eps);
uint16_t val1 = img(pi);
double val2 = img.interp(pd);
double val3 = img.interpGrad(pd)[0];
EXPECT_LE(std::abs(val2 - val1), threshold);
EXPECT_FLOAT_EQ(val2, val3);
}
{
Eigen::Vector2i pi = offset + Eigen::Vector2i(0, 1);
Eigen::Vector2d pd = pi.cast<double>() + Eigen::Vector2d(eps, -eps);
uint16_t val1 = img(pi);
double val2 = img.interp(pd);
double val3 = img.interpGrad(pd)[0];
EXPECT_LE(std::abs(val2 - val1), threshold);
EXPECT_FLOAT_EQ(val2, val3);
}
{
Eigen::Vector2i pi = offset + Eigen::Vector2i(1, 1);
Eigen::Vector2d pd = pi.cast<double>() + Eigen::Vector2d(-eps, -eps);
uint16_t val1 = img(pi);
double val2 = img.interp(pd);
double val3 = img.interpGrad(pd)[0];
EXPECT_LE(std::abs(val2 - val1), threshold);
EXPECT_FLOAT_EQ(val2, val3);
}
}
TEST(Image, ImageInterpolateGrad) {
Eigen::Vector2i offset(231, 123);
basalt::ManagedImage<uint16_t> img(640, 480);
setImageData(img.ptr, img.size());
Eigen::Vector2d pd = offset.cast<double>() + Eigen::Vector2d(0.4, 0.34345);
Eigen::Vector3d valGrad = img.interpGrad<double>(pd);
Eigen::Matrix<double, 1, 2> J = valGrad.tail<2>();
test_jacobian(
"d_res_d_x", J,
[&](const Eigen::Vector2d &x) {
return Eigen::Matrix<double, 1, 1>(img.interp<double>(pd + x));
},
Eigen::Vector2d::Zero(), 1);
}

@ -1 +1 @@
Subproject commit c339ea747a29881271cfcbc559937b81dcf76324
Subproject commit 792c0e9b5a2ccffac60c80f29905cfde145d4042