开源 企业版 高校版 私有云 模力方舟 AI 队友
代码拉取完成,页面将自动刷新
加入 Gitee
与超过 1400万 开发者一起发现、参与优秀开源项目,私有仓库也完全免费 :)
免费加入
已有帐号? 立即登录
文件
master
分支 (8)
标签 (20)
master
cadgit
devel
bfm2017-xcode
landmarks-weighting
multi_image_fit_devel
gh-pages
perspective-camera-estimation
v0.13.0
v0.12.2
v0.12.1
v0.12.0
v0.11.1
v0.11.0
v0.10.1
v0.10.0
v0.9.1
v0.9.0
v0.8.0
v0.7.1
v0.7.0
v0.6.1
v0.6.0
v0.5.0
v0.4.0
v0.3.0
v0.2.0
v0.1.0
master
分支 (8)
标签 (20)
master
cadgit
devel
bfm2017-xcode
landmarks-weighting
multi_image_fit_devel
gh-pages
perspective-camera-estimation
v0.13.0
v0.12.2
v0.12.1
v0.12.0
v0.11.1
v0.11.0
v0.10.1
v0.10.0
v0.9.1
v0.9.0
v0.8.0
v0.7.1
v0.7.0
v0.6.1
v0.6.0
v0.5.0
v0.4.0
v0.3.0
v0.2.0
v0.1.0
克隆/下载
克隆/下载
提示
下载代码请复制以下命令到终端执行
为确保你提交的代码身份被 Gitee 正确识别,请执行以下命令完成配置
初次使用 SSH 协议进行代码克隆、推送等操作时,需按下述提示完成 SSH 配置
1 生成 RSA 密钥
2 获取 RSA 公钥内容,并配置到 SSH公钥
在 Gitee 上使用 SVN,请访问 使用指南
使用 HTTPS 协议时,命令行会出现如下账号密码验证步骤。基于安全考虑,Gitee 建议 配置并使用私人令牌 替代登录密码进行克隆、推送等操作
Username for 'https://gitee.com': userName
Password for 'https://userName@gitee.com': # 私人令牌
master
分支 (8)
标签 (20)
master
cadgit
devel
bfm2017-xcode
landmarks-weighting
multi_image_fit_devel
gh-pages
perspective-camera-estimation
v0.13.0
v0.12.2
v0.12.1
v0.12.0
v0.11.1
v0.11.0
v0.10.1
v0.10.0
v0.9.1
v0.9.0
v0.8.0
v0.7.1
v0.7.0
v0.6.1
v0.6.0
v0.5.0
v0.4.0
v0.3.0
v0.2.0
v0.1.0
eos
/
python
/
pybind11_Image.hpp
eos
/
python
/
pybind11_Image.hpp
pybind11_Image.hpp 6.08 KB
一键复制 编辑 原始数据 按行查看 历史
Patrik Huber 提交于 2017年06月30日 02:39 +08:00 . Split type casters into Image3u and Image4u
/*
* eos - A 3D Morphable Model fitting library written in modern C++11/14.
*
* File: python/pybind11_Image.hpp
*
* Copyright 2017 Patrik Huber
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#pragma once
#include "pybind11/numpy.h"
#include "Eigen/Core"
#include <cstddef>
#include <vector>
NAMESPACE_BEGIN(pybind11)
NAMESPACE_BEGIN(detail)
/**
* @file python/pybind11_Image.hpp
* @brief Transparent conversion to and from Python for eos::core::Image.
*
* Numpy uses row-major storage order by default.
* eos::core::Image uses col-major storage (like Eigen).
*
* If given non-standard strides or something from numpy, probably doesn't work.
* May need to .clone()? in numpy before passing to the C++ function.
*/
/**
* @brief Transparent conversion for eos::core::Image3u to and from Python.
*
* Converts an eos::core::Image3u to and from Python. Can construct a eos::core::Image3u from numpy arrays,
* as well as potentially other Python array types.
*
* Note: Not sure what happens if the given numpy array is not contiguous, not in default (row-major) storage
* order, or has non-standard strides. It may or may not work.
*/
template<>
struct type_caster<eos::core::Image3u>
{
bool load(handle src, bool)
{
auto buf = pybind11::array::ensure(src);
if (!buf)
return false;
// Todo: We should probably check that buf.strides(i) is "default", by dividing it by the Scalar type or something.
if (!pybind11::isinstance<pybind11::array_t<std::uint8_t>>(buf))
{
return false; // we only convert uint8_t for now.
}
if (buf.ndim() != 3) {
return false; // we expected a numpy array with 3 dimensions.
}
// We got something with 3 dimensions, i.e. an image with 2, 3 or 4 channels (or 'k' for that matter):
if (buf.shape(2) != 3) {
return false; // We expected a 3-channel image.
}
// Note: If our Image class had support for col/row major, we could just map buf.mutable_data().
// Like with OpenCV: value = cv::Mat(buf.shape(0), buf.shape(1), opencv_type, buf.mutable_data());
// But since it doesn't, we just copy the data for now:
value = eos::core::Image3u(buf.shape(0), buf.shape(1));
array_t<std::uint8_t> buf_as_array(buf);
for (int r = 0; r < buf.shape(0); ++r) {
for (int c = 0; c < buf.shape(1); ++c) {
value(r, c)[0] = buf_as_array.at(r, c, 0);
value(r, c)[1] = buf_as_array.at(r, c, 1);
value(r, c)[2] = buf_as_array.at(r, c, 2);
}
}
return true;
};
static handle cast(const eos::core::Image3u& src, return_value_policy /* policy */, handle /* parent */)
{
const std::size_t num_channels = 3;
std::vector<std::size_t> shape = { src.rows, src.cols, num_channels };
// (2048, 4, 1) is default which results in transposed image
// Below line works now. In numpy the strides are (2048, 4, 1) though. I think a copy gets created nevertheless?
std::vector<std::size_t> strides = { num_channels, num_channels * src.rows, 1 }; // might be cols or rows...? I think rows?
return array(pybind11::dtype::of<std::uint8_t>(), shape, strides, &src.data[0]).release();
};
PYBIND11_TYPE_CASTER(eos::core::Image3u, _("numpy.ndarray[uint8[m, n, 3]]"));
};
/**
* @brief Transparent conversion for eos::core::Image4u to and from Python.
*
* Converts an eos::core::Image4u to and from Python. Can construct a eos::core::Image4u from numpy arrays,
* as well as potentially other Python array types.
*
* Note: Not sure what happens if the given numpy array is not contiguous, not in default (row-major) storage
* order, or has non-standard strides. It may or may not work.
*/
template<>
struct type_caster<eos::core::Image4u>
{
bool load(handle src, bool)
{
auto buf = pybind11::array::ensure(src);
if (!buf)
return false;
// Todo: We should probably check that buf.strides(i) is "default", by dividing it by the Scalar type or something.
if (!pybind11::isinstance<pybind11::array_t<std::uint8_t>>(buf))
{
return false; // we only convert uint8_t for now.
}
if (buf.ndim() != 3) {
return false; // we expected a numpy array with 3 dimensions.
}
// We got something with 3 dimensions, i.e. an image with 2, 3 or 4 channels (or 'k' for that matter):
if (buf.shape(2) != 4) {
return false; // We expected a 4-channel image.
}
// Note: If our Image class had support for col/row major, we could just map buf.mutable_data().
// Like with OpenCV: value = cv::Mat(buf.shape(0), buf.shape(1), opencv_type, buf.mutable_data());
// But since it doesn't, we just copy the data for now:
value = eos::core::Image4u(buf.shape(0), buf.shape(1));
array_t<std::uint8_t> buf_as_array(buf);
for (int r = 0; r < buf.shape(0); ++r) {
for (int c = 0; c < buf.shape(1); ++c) {
value(r, c)[0] = buf_as_array.at(r, c, 0);
value(r, c)[1] = buf_as_array.at(r, c, 1);
value(r, c)[2] = buf_as_array.at(r, c, 2);
value(r, c)[3] = buf_as_array.at(r, c, 3);
}
}
return true;
};
static handle cast(const eos::core::Image4u& src, return_value_policy /* policy */, handle /* parent */)
{
const std::size_t num_chanels = 4;
std::vector<std::size_t> shape;
shape = { src.rows, src.cols, num_chanels };
// (2048, 4, 1) is default which results in transposed image
// Below line works now. In numpy the strides are (2048, 4, 1) though. I think a copy gets created nevertheless?
std::vector<size_t> strides = { num_chanels, num_chanels * src.rows, 1 }; // might be cols or rows...? I think rows?
return array(pybind11::dtype::of<std::uint8_t>(), shape, strides, &src.data[0]).release();
};
PYBIND11_TYPE_CASTER(eos::core::Image4u, _("numpy.ndarray[uint8[m, n, 4]]"));
};
NAMESPACE_END(detail)
NAMESPACE_END(pybind11)
Loading...
举报
举报成功
我们将于2个工作日内通过站内信反馈结果给你!
请认真填写举报原因,尽可能描述详细。
请选择举报类型
取消
发送
误判申诉

此处可能存在不合适展示的内容,页面不予展示。您可通过相关编辑功能自查并修改。

如您确认内容无涉及 不当用语 / 纯广告导流 / 暴力 / 低俗色情 / 侵权 / 盗版 / 虚假 / 无价值内容或违法国家有关法律法规的内容,可点击提交进行申诉,我们将尽快为您处理。

取消
提交

简介

暂无描述
暂无标签
Apache-2.0
使用 Apache-2.0 开源许可协议
取消

发行版

暂无发行版

贡献者

全部

近期动态

不能加载更多了
编辑仓库简介
简介内容
主页
马建仓 AI 助手
尝试更多
代码解读
代码找茬
代码优化
1
https://gitee.com/cadgit-dev/eos.git
git@gitee.com:cadgit-dev/eos.git
cadgit-dev
eos
eos
master
点此查找更多帮助

搜索帮助

评论
仓库举报
回到顶部
登录提示
该操作需登录 Gitee 帐号,请先登录后再操作。
立即登录
没有帐号,去注册

AltStyle によって変換されたページ (->オリジナル) /