tf.metrics.auc与相比产生非常不同的值sklearn.metrics.roc_auc_score与某些情况
我无法确定这种情况的特殊性，但我获得一个可重现的示例：
代码和数据可从https://www.dropbox.com/s/ym2ptqaqw2qjite/minimal_program_AUC.zip?raw=1下载
代码：
import sklearn.metrics
from keras.models import Model, load_model
import tensorflow as tf
from keras import backend as K
import numpy as np
def auc(y_true, y_pred):
auc = tf.metrics.auc(y_true, y_pred, num_thresholds=65)[1]
K.get_session().run(tf.local_variables_initializer())
return auc
model = load_model('Model.hdf5', custom_objects={'auc': auc})
X = np.fromfile('X_test(65, 80, 1292, 1).txt', sep=',').reshape(65, 80, 1292, 1)
Y = np.fromfile('Y_test(65, 1).txt', sep=',').reshape(65, 1)
batchsize = 45
evaluation = model.evaluate(X, Y, batch_size=batchsize, verbose=1)
predictions = model.predict(X, batch_size=batchsize, verbose=0, steps=None)
auc = …这是我的最小示例：https://www.dropbox.com/s/7fwsr3sigb60rtw/leveling-test.zip ?dl=0
\n\n我当前的项目文件夹结构和相关的CMakeLists内容：
\n\nleveling\n\xe2\x94\x9c\xe2\x94\x80\xe2\x94\x80 CMakeLists.txt: add_subdirectory(deps) \n\xe2\x94\x94\xe2\x94\x80\xe2\x94\x80 deps\n \xe2\x94\x9c\xe2\x94\x80\xe2\x94\x80 CMakeLists.txt: add_subdirectory(xml-reading)\n \xe2\x94\x94\xe2\x94\x80\xe2\x94\x80 xml-reading\n \xe2\x94\x9c\xe2\x94\x80\xe2\x94\x80 CMakeLists.txt: add_subdirectory(deps)\n \xe2\x94\x82 add_library(xml-reading ...)\n \xe2\x94\x82 target_include_directories(xml-reading PUBLIC ${CMAKE_CURRENT_LIST_DIR}/deps/tinyxml2)\n \xe2\x94\x82 target_link_libraries(xml-reading PUBLIC tinyxml2)\n \xe2\x94\x94\xe2\x94\x80\xe2\x94\x80 deps\n \xe2\x94\x9c\xe2\x94\x80\xe2\x94\x80 CMakeLists.txt: add_subdirectory(tinyxml2)\n \xe2\x94\x94\xe2\x94\x80\xe2\x94\x80 tinyxml2\n这会生成一个 xml-reading.dll 文件。
\n\n但是调平项目链接器选项有/DYNAMICBASE "bin\\windows-32\\debug\\xml-reading.lib" "bin\\windows-32\\debug\\BOBPrimitives.lib" "bin\\windows-32\\debug\\tinyxml2d.lib"
其结果是：\n2>LINK : fatal error LNK1104: cannot open file \'bin\\windows-32\\debug\\xml-reading.lib\'
这是怎么回事？
\n\nadd_library(xml-reading SHARED ...)，但没有帮助add_library(xml-reading STATIC ...)为强制它生成 .lib 文件 - …今天，我尝试使用Arch 解决方案通过 FFMPEG 进行流式传输，但由于 FFMPEG 上的一件简单的事情，我所有的尝试都徒劳无功。它说探针大小不够大，所以我本能地越来越大地增加探针大小......现在它-probesize "500M"仍然说它不够大。这是代码片段
[x11grab @ 0x5631f846cd00] Stream #0: not enough frames to estimate rate; consider increasing probesize
Input #0, x11grab, from ':0.0':
Duration: N/A, start: 1603397505.341400, bitrate: 1007124 kb/s
Stream #0:0: Video: rawvideo (BGR[0] / 0x524742), bgr0, 1366x768, 1007124 kb/s, 30 fps, 1000k tbr, 1000k tbn, 1000k tbc
0: Input/output error
和代码
#!/bin/bash
INRES="1366x768" # input resolution
OUTRES="1366x768" # output resolution
FPS="30" # target FPS
GOP="60" # i-frame interval, should be double …我发现了一个特定的 100MB bin 文件（最小示例中的 CarveObj_k5_rgbThreshold10_triangleCameraMatches.bin），其中谷物无法加载并抛出异常“无法从输入流读取 368 字节！读取 288”
根据相同数据构建的相应 900MB XML 文件（在最小示例中为 CarveObj_k5_rgbThreshold10_triangleCameraMatches.xml）可以正常加载。
XML 文件是由
// {
// std::ofstream outFile(base + "_triangleCameraMatches.xml");
// cereal::XMLOutputArchive oarchive(outFile);
// oarchive(m_triangleCameraMatches);
// }
二进制版本是由
// {
// std::ofstream outFile(base + "_triangleCameraMatches.bin");
// cereal::BinaryOutputArchive oarchive(outFile);
// oarchive(m_triangleCameraMatches);
// }
最小示例：https://www.dropbox.com/sh/fu9e8km0mwbhxvu/AAAfrbqn_9Tnokj4BVXB8miea?dl= 0
使用的谷物版本：1.3.0
2017年MSVS
Windows 10
这是一个错误吗？我错过了一些明显的东西吗？同时创建了一个错误报告：https ://github.com/USCiLab/cereal/issues/607
ORB 演示代码位于https://opencv-python-tutroals.readthedocs.io/en/latest/py_tutorials/py_feature2d/py_orb/py_orb.html
import numpy as np
import cv2
from matplotlib import pyplot as plt
img = cv2.imread('simple.jpg',0)
# Initiate STAR detector
orb = cv2.ORB()
# find the keypoints with ORB
kp = orb.detect(img,None)
# compute the descriptors with ORB
kp, des = orb.compute(img, kp)
# draw only keypoints location,not size and orientation
img2 = cv2.drawKeypoints(img,kp,color=(0,255,0), flags=0)
plt.imshow(img2),plt.show()
在kp = orb.detect(img,None)
[WinError 10054] An existing connection was forcibly closed by the remote hostcv2.error: Unknown C++ exception …我不明白编译错误 C2676
对于下面的代码
#ifndef __VEC_3D_H__
#define __VEC_3D_H__
#include <vector>
#include <cmath>
namespace Internal
{
/** very simple 3D vector/ point */
class Vec3D
{
public:
float mX;
float mY;
float mZ;
/// null constructor
Vec3D(void) {}
/// construct from data
Vec3D(float x, float y, float z) : mX(x), mY(y), mZ(z) {}
inline friend std::ostream& operator<< (std::ostream& os, const Vec3D& v)
{
os << "(" << v.mX << ", " << v.mY << ", " << v.mZ << ")";
return os; …git clone my-new-repo-url
mkdir deps
cd deps
git submodule add -b 6.2.0 https://github.com/leethomason/tinyxml2.git
产量
fatal: 'origin/6.2.0' is not a commit and a branch '6.2.0' cannot be created from it
Unable to checkout submodule 'deps/tinyxml2'
不填充 .gitmodules，而是创建一个文件夹.git/modules/deps/tinyxml2并在其中添加一个 repodeps/tinyxml2
我以为我以前这样做过，它会填充 .gitmodules
[submodule "deps/tinyxml2"]
path = deps/tinyxml2
url = https://github.com/leethomason/tinyxml2.git
branch = 6.2.0
但它现在不起作用怎么办？
我需要向量中值的最小值和均值。
我用以下内容分别计算它们：
template <class T>
T Minimum(std::vector<T> & v){ return *min_element(begin(v), end(v)); }
template <class T>
T Mean(std::vector<T> & v)
{
T sum = std::accumulate(v.begin(), v.end(), static_cast<T>(0));
T mean = sum / v.size();
return mean;
}
这两个都必须扫描向量。
是否有一种std 有效的方法来计算向量的最小值和均值，将其仅扫描一次？
我\xe2\x80\x99m很困惑，请帮我理解：
\n\n我有这个错误：
\n\n img_array = cv2.resize(img_array, dsize=(299, 299, 3), interpolation=cv2.INTER_CUBIC) \nTypeError: function takes exactly 2 arguments (3 given) \n什么时候
\n\nprint(img_array): <class \'numpy.ndarray\'>\nimg_array.shape: (240, 320, 3)\nprint(cv2.__version__): 4.2.0 \nVSCode 中将鼠标悬停在调整大小上会显示
\n\ndef resize(src, dsize, dst=None, fx=None, fy=None, interpolation=None)\nresize(src, dsize[, dst[, fx[, fy[, interpolation]]]]) -> dst . @brief Resizes an image. .\n. The function resize resizes the image src down to or up to the specified size. Note that the . initial dst type or size are …