qq_44920947的博客

编译报错：undefined reference to `cv::imshow(cv::String const&, cv::_InputArray const&)'解决办法：CMakeLists.txt中加入find_package(OpenCV 3 QUIET)target_link_libraries(program_name ${OpenCV_LIBS})OpenCV 3 是安装OpenCV的版本号，如果装的4.*就写OpenCV 4program_name是可执行

Pangolin安装报错：make: *** 没有规则可制作目标“pypangolin_pip_install”。 停止。按照github上的步骤安装，到$ cmake --build . -t pypangolin_pip_install这步就报错make: *** 没有规则可制作目标“pypangolin_pip_install”。 停止。

No module named 'rosdep2'报错按照wiki.ros.org上的教程，执行$ rospack depends1 beginner_tuto就报错了：ModuleNotFoundError: No module named 'rosdep2'ModuleNotFoundError: No module named 'rosdep2'[rospack] Error: could not find python module 'rosdep2.rospack'. is rosdep

按照wiki.ros.org上的教程安装完，创建ros工作空间，$ catkin_make它就报错了：CMake Error at /opt/ros/noetic/share/catkin/cmake/empy.cmake:30 (message): Unable to find either executable 'empy' or Python module 'em'... try installing the package 'python3-empy'最后就来一句Invoking “

本文试从概念到实例，把c++里的定义、声明和初始化这几个东西讲清楚

Papers Notes_7_ Batch Normalization--Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate ShiftIntroductionBatch NormalizationalgorithmworksbenefitReferencesIntroductionmini-batchestimate the gradient over the training

Papers Notes_6_ DCGAN--Unsupervised Representation Learning with Deep Convolutional GANApproachArchitectureApproachscale up GANs using CNNs to model imagesthe all convolutional netreplace deterministic spatial pooling functions (such as maxpooling) wi

Papers Notes_5_ GAN--Generative Adversarial NetsAdversarial NetsReferencesAdversarial Netsdiscriminative modelgenerative modelReferencesDeep Residual Learning for Image Recognition

Papers Notes_4_ ResNet--Deep Residual Learning for Image RecognitionNetwork DepthResidual MappingArchitectureNetwork Depthnetwork depth is of crucial importance, and the leading results on the challenging ImageNet dataset all exploit “very deep” models,

Papers Notes_3_ GoogLeNet--Going deeper with convolutionsArchitectureInception moduleGoogLeNetArchitectureInception moduleparallel conv. with different kernel sizevisual information should be processed at various scales and then aggregated→the next

Papers Notes_2_ VGG--Very Deep Convolutional Networks for Lage-scale Image RecognitionArchitectureTrainingArchitectureinput-224×224 RGB imagepreprocess-subtract the mean RGB value,computed on the training set,from each pixelstride-fixed to 1 pixelpad

Papers Notes_1_ AlexNet--ImageNet Classification with Deep Convolutional Neural NetworksArchitectureReLU NonlinearityLocal Response NormalizationOverlapping PoolingOverall ArchitectureReducing OverfittingData AugmentationDropoutDetails of LearningReference

FCOS论文及源码详解（七）

FCOS论文及源码详解（六）

FCOS论文及源码详解（五）

FCOS论文及源码详解（四）

FCOS论文及源码详解（三）

FCOS论文及源码详解（二）

FCOS论文及源码详解（一）FCOS论文Fully Convolutional One-Stage Object DetectorMulti-level Prediction with FPN for FCOSCenter-ness for FCOS参考文献FCOS论文以下内容摘自论文(FCOS: Fully Convolutional One-Stage Object Detection)，本人英语水平极差，翻译勉强看看就好Almostall state-of-the-art object dete

ffmpeg 压缩视频代码(Python)import threadingimport osdef zip(): compress = "ffmpeg -i C:/Users/123456/Videos/Captures/1234.mp4 -r 30 -c:v libx264 -preset veryslow \ -profile:v baseline -crf 24 -c:a copy C:/Users/123456/Videos/Captures/12.mp4" isRun =