cv.exp-CSDN博客

原创 NCNN模型分析

NCNN模型分析层类型layer type 层名称layer name 输入数量input count 输出数量output count 网络输入层名input blobs 网络输出层名output blobs 特殊参数layer specific params Convolution .

2021-03-31 19:51:27 814 4

$ sudo docker pull tensorflow/serving[sudo] admin 的密码：Using default tag: latestlatest: Pulling from tensorflow/servingd519e2592276: Pull completed22d2dfcfa9c: Pull completeb3afe92c540b: Pull completecd7dd7eb8873: Pull complete7998bd37b38c: Pull com

2021-03-21 20:01:27 172

原创 jsonnet allennlp安装

训练问题：Traceback (most recent call last):File “train.py”, line 163, in entry_point(config)File “train.py”, line 130, in entry_pointdist.destroy_process_group()File “/home/admin/R-/lib/python3.7/site-packages/torch/distributed/distributed_c10d.py”, line

2021-03-21 09:34:43 576

原创 faiss安装

~$ pip --default-time=1000 install -i https://pypi.tuna.tsinghua.edu.cn/simple faiss-gpuLooking in indexes: https://pypi.tuna.tsinghua.edu.cn/simpleCollecting faiss-gpu Downloading https://pypi.tuna.tsinghua.edu.cn/packages/******/faiss_gpu-1.7.0-cp38-.

2021-02-19 21:25:24 466

原创 VULKAN-04-init_command_buffer 研究[完善中doing]

/*VULKAN_SAMPLE_SHORT_DESCRIPTIONCreate Vulkan command buffer*//* This is part of the draw cube progression */#include <util_init.hpp>#include <assert.h>#include <cstdlib>int sample_main(int argc, char *argv[]) { VkResult .

2021-01-11 21:39:22 119

原创 VULKAN-03-init_device 研究[doing]

#include <iostream>#include <cassert>#include <cstdlib>#include <util_init.hpp>int sample_main(int argc, char *argv[]) { struct sample_info info = {}; init_global_layer_properties(info); init_instance(info, "vulkans...

2021-01-10 10:37:44 177 1

原创 VULKAN-02-enumerate_devices 研究[doing]

#include <cassert>#include <cstdlib>#include <util_init.hpp>int sample_main(int argc, char *argv[]) { struct sample_info info = {}; init_global_layer_properties(info); init_instance(info, "vulkansamples_enumerate"); /...

2021-01-09 21:08:38 82

原创 VULKAN-01-init_instance 研究[doing]

#include <iostream>#include <cstdlib>#include <util_init.hpp>#define APP_SHORT_NAME "vulkansamples_instance"int sample_main(int argc, char *argv[]) { struct sample_info info = {}; init_global_layer_properties(info); /* ...

2021-01-09 11:19:55 151

原创 ImportError: cannot import name ‘deform_conv_cuda‘ from partially initialized module ‘assets.ops.dcn

ImportError: cannot import name 'deform_conv_cuda' from partially initialized module 'assets.ops.dcn'

2020-12-26 21:14:48 5288

原创怎么评价一个深度学习解决方案

一个好的深度学习的解决方案，应该做到以下几点：1、完整性，包括数据、代码（包括训练、部署、测试），其中训练部分，更加需要完整性；2、可持续性，希望这个方案，是不断的可以优化提升，而不是，将来被彻底的推倒重来；3、维护成本低，包含文档的支持、操作流程化等...

2020-09-13 16:29:55 260

原创 Caffe-MobileNetSSD下ncnn推理实现-6

(py365) user@ubuntu:~/caffe_ssd/caffe/examples$ git clone https://github.com/chuanqi305/MobileNet-SSD.git(py365) user@ubuntu:~/caffe_ssd/caffe/examples/MobileNet-SSD$ ./gen_model.sh 31接下来可以参考下面的网络资料：【Caffe】caffe-ssd训练自己的数据集，https://blog.csdn.net/heih

2020-09-10 07:08:55 179

原创 Caffe-MobileNetSSD下ncnn推理实现-5

安装caffe python接口(py365) user@ubuntu:~/caffe_ssd/caffe$ pip install -r python/requirements.txtpip3 install matplotlibmake pycaffe -j8出现问题：(py365) user@ubuntu:~/caffe_ssd/caffe$ make pycaffe -j8touch python/caffe/proto/__init__.pyCXX/LD -o pyth

2020-09-10 07:07:57 293

原创 Caffe-MobileNetSSD下ncnn推理实现-4

安装caffegit clone https://github.com/BVLC/caffe或者git clone https://gitee.com/mirrors/caffe.gitcp Makefile.config.example Makefile.configMakefile.config文件：## Refer to http://caffe.berkeleyvision.org/installation.html# Contributions simplifying and

2020-09-10 07:05:13 191 1

原创 Caffe-MobileNetSSD下ncnn推理实现-3

安装opencvsudo apt-get install build-essentialsudo apt-get install cmake git libgtk2.0-dev pkg-config libavcodec-dev libavformat-dev libswscale-devsudo apt-get install python-dev python-numpy libtbb2 libtbb-dev libjpeg-dev libpng-dev libtiff-dev libjas

2020-09-10 07:02:04 184

原创 Caffe-MobileNetSSD下ncnn推理实现-2

https://github.com/protocolbuffers/protobuf/releases/https://github.com/protocolbuffers/protobuf/releases/download/v3.11.4/protobuf-all-3.11.4.tar.gz(py365) user@ubuntu:~/caffe_ssd/protobuf$sudo apt-get install autoconf automake libtool curl mak.

2020-09-10 07:00:29 136

原创 Caffe-MobileNetSSD下ncnn推理实现-1

首先，使用命令conda create --name py365 python=3.6.5创建python3.6.5虚拟环境；conda activate conda activate py365（conda deactivate）。1 安装ubuntu依赖包sudo apt-get install libprotobuf-dev libleveldb-dev libsnappy-dev libhdf5-serial-dev protobuf-compilersudo apt-get i..

2020-09-10 06:57:50 202

原创 ESP32 开发交叉编译工具的问题解决

$ make menuconfigmconf-idf is not required on this platformmconf-idf is not required on this platformThe following Python requirements are not satisfied:https://blog.csdn.net/weixin_37127273/article/details/84790079Please follow the instructions found

2020-07-04 20:50:48 3665 6

原创 kpu.h文件研究（完善中）

kpu.h文件研究/* Copyright 2018 Canaan Inc. * * 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/l

2020-06-28 13:53:09 645 1

原创关于深度学习编译器

深度学习框架推理框架然后，我们需要研究深度学习编译器，感觉这一块会很有价值。数据标注、模型训练，然后编译器直接适配我们的平台，感觉很有潜力，重点是深度学习编译器。移动端推理框架，目的是模型优化，能够适配移动端。深度学习编译器应该会是更加期待的方案。...

2020-06-13 18:58:42 367

原创 Android JNI OpenCV ROI问题探究

一、Rect boundRect(0,0,0,0);Mat im_boundRect(im_bgr.rows, im_bgr.cols, CV_8UC3);im_bgr(boundRect).copyTo(im_boundRect);二、Mat rot_image = im_bgr(boundRect);在jni中我选择一，问题解决，使用二结果为空

2020-06-13 00:30:09 243

原创腾讯开源推理框架TNN学习笔记（进行中）

6月10日之前期待1、推理框架TNN2、腾讯开源新一代3、和NCNN有什么不同6月10日

2020-06-10 07:50:42 4268 1

原创 psenet字符定位和warpAffine变换

psenet身份证定位效果

2020-06-06 13:39:08 300

原创 K210 camera开发

https://dl.sipeed.com/MAIX/MaixPy/release/master/maixpy_v0.5.0_88_gb7f0c13>>> import sensor>>> import lcd>>>>>> lcd.init()>>>>>> sensor.reset()init i2c2[MAIXPY]: find ov2640>>> sen

2020-06-02 22:27:26 1733

原创 K210深度学习C语言开发前期准备工作指南

k210$ wget https://s3.cn-north-1.amazonaws.com.cn/dl.kendryte.com/documents/kendryte-standalone-sdk-0.5.2.zip--2020-05-30 14:10:55-- https://s3.cn-north-1.amazonaws.com.cn/dl.kendryte.com/documents/kendryte-standalone-sdk-0.5.2.zip正在解析主机 s3.cn-north-1..

2020-05-30 15:24:18 4198 1

原创 AIOT K210开发环境起步

出现问题：kflash_gui$ python kflash_gui.pyTraceback (most recent call last): File "kflash_gui.py", line 25, in <module> from kflash_py.kflash import KFlashModuleNotFoundError: No module named 'kflash_py.kflash'参考：https://github.com/sipeed/kf...

2020-05-29 08:13:57 1998 2

原创 faster-rcnn.pytorch问题汇总和解决

1、问题：ImportError: cannot import name '_mask' from 'pycocotools' (/media/admin1/f7bb2a6e-7469-413e-89a5-3d4b2a04b76f/alg/iuloop-ocr/faster-rcnn.pytorch/lib/pycocotools/__init__.py)解决：https://github.com/jwyang/faster-rcnn.pytorch/issues/6112、问题： f...

2020-05-10 13:43:33 2809 4

原创 OCR PSEnet定位

测试图片来源百度

2020-05-03 09:34:40 411

原创问题：g++: error: unrecognized command line option ‘-fno-plt

问题：make后出现问题g++: error: unrecognized command line option ‘-fno-plt解决：makefile文件CXXFLAGS = -I include -std=c++11 -O3 $(shell python3-config --cflags)LDFLAGS = $(shell python3-config --ldfla...

2020-05-01 22:06:15 7050 2

原创 MNN使用问题和解决汇总

问题：../../../../includes\MNN/Matrix.h:147: error: undefined reference to 'MNN::CV::Matrix::computeTypeMask() const'解决：

2020-04-25 23:17:59 815 1

原创 These Op Not Support: Tensorflow::FusedBatchNorm 问题

~/mnn/build$ ./MNNConvert -f TF --modelFile 1frozen_model.pb --MNNModel 1frozen_model.mnn --bizCode bizMNNConverter Version: 0.2.1.5git - MNN @ 2018Start to Convert Other Model Format To MNN Mode...

2020-04-25 14:35:57 1168 5

原创 libmnncore.so问题，解决思路

ninja: error: '../../../../libs/x86_64/libMNN.so', needed by '../../../../build/intermediates/cmake/debug/obj/x86_64/libmnncore.so', missing and no known rule to make it at com.android.build.grad...

2020-04-25 10:55:50 733 1

原创超大图片加载

超大图片加载pip install openslide-python；加载超大图片出现问题：PIL.Image.DecompressionBombError:Imagesize(***67***314pixels)exceedslimitof***89***70pixels,couldbedecompressionbombDOSattack. ...

2020-04-18 09:02:22 1398

原创 Failed to recurse into submodule path 'third_party/onnx'

user@ubuntu:~/pytorch$ git submodule update --init --recursivefatal: Needed a single revisionUnable to find current revision in submodule path 'third_party/onnx/third_party/benchmark'Failed to recu...

2018-11-13 18:44:14 3855

原创安装dlib问题解决

user@ubuntu:~/aitool/project$ sudo pip install dlibThe directory '/home/user/.cache/pip/http' or its parent directory is not owned by the current user and the cache has been disabled. Please check th...

2018-11-08 11:17:00 7780 2

转载主流ocr算法：CNN+BLSTM+CTC架构

ocr算法研究实验性的项目，目前实现了CNN+BLSTM+CTC架构链接：https://github.com/senlinuc/caffe_ocr主流ocr算法研究实验性的项目，目前实现了CNN+BLSTM+CTC架构ocrlstmctcres-lstmdensenet 41 commits 1 branch 0

2017-11-18 09:26:01 45259 18

原创人脸识别研究总结

人脸识别算法思路

2017-11-18 09:22:36 3241 3

原创 Faster RCNN训练（Matlab版本）结果

Faster RCNN训练（Matlab版本）结果：

2017-11-14 06:34:05 3336

原创 Faster RCNN训练出现问题：Matlab版本训练

I1113 20:56:49.405777 23834 net.cpp:42] Initializing net from parameters: name: "Zeiler_conv5"input: "data"input_dim: 1input_dim: 3input_dim: 224input_dim: 224state { phase: TEST}layer {

2017-11-13 21:00:58 2340

原创 Faster RCNN训练出现问题：generated_database_->Add(encoded_file_descriptor, size)

I1112 18:49:04.704902 18730 net.cpp:743] Ignoring source layer relu3I1112 18:49:04.704903 18730 net.cpp:743] Ignoring source layer conv4I1112 18:49:04.704905 18730 net.cpp:743] Ignoring source layer

2017-11-12 21:39:35 2225

原创 Faster RCNN训练出现问题：smooth_L1_loss_layer.cpp:28] Check failed: bottom[0]->channels() == bottom[1]->cha

I1030 20:49:01.633868 24921 net.cpp:157] Top shape: 1 4096 (4096)I1030 20:49:01.633885 24921 net.cpp:165] Memory required for data: 280497656I1030 20:49:01.633891 24921 layer_factory.hpp:77] Creatin

2017-10-30 21:31:13 2508 3