- 博客(62)
- 资源 (1)
- 收藏
- 关注
原创 病中梦记
恍惚梦间,Q3即期,推全未果,急寻dn与han姐议计。久相持,han姐恼,曰尔等自便,起身离席。谨慎开卷,伏案观之,上有一计,似可解燃眉之急。忽有贼破门而入,欲夺此卷,han姐与之相争,势渐下。酣战时,惊坐起,寻爪机,方寅时,思胜负未明,岂言丈夫。再入梦,欲擒之,贼与卷皆不得寻,呜呼哀哉。上周火锅毕,逞一时之快,赤卧空调下,感伤风寒,咳喘至今,虽白嫖,犹甚亏。昨日萎靡不振,告假半日,寻郎中抓药两副,返屋服药,颓然入睡,三觉三醒,怪梦连篇。夏秋换季,天象多变,诚望诸君,以吾为鉴,各安尊体,长命百岁。...
2022-08-28 00:38:27
188
原创 352记--大师本纪
一日,有人见大师架置一书,名《道德经》,甚是好奇,遂往讨教。大师为人低调,不愿详言,然隔墙有耳,众人已闻,皆聚大师屋内。大师亦不推脱,连卜数卦,竟十中八九,在座瞪目结舌,无不大惊,便问个中缘由。大师略阐卦象,众人不知所云,以为妙绝,遂赠号“大师”,以表尊崇。众议纷纷,欲寻此人。唯有一人,手捋飘髯,侧目微笑,泰然自若,闲看花开花落之态,漫随云卷云舒之姿。大师者,352室始长,中州刘氏,名某俊,生卒年不祥,异人也。自此,352建制初成,大师开立大统,称寝始皇,独揽财政,雷厉风行,权倾全寝,普度众生。...
2022-08-28 00:36:23
152
原创 Feature
tf.train.Feature.https://www.tensorflow.org/api_docs/python/tf/train/Featuretf.train.Examplehttps://www.tensorflow.org/api_docs/python/tf/train/Exampletf.io.parse_exampletf.io.parse_example | TensorFlow Core v2.8.0tf.RaggedTensorint_fe...
2022-05-08 21:25:28
395
原创 windows 配置
Git 右键添加Git Bashhttps://www.cnblogs.com/mythdoraemon/p/9865567.html 增加和删除win右键的Open Folder as PyCharm Project,Jetbrains系列https://blog.csdn.net/qq_37504892/article/details/107705527
2020-08-26 14:38:25
202
原创 python 笔记
s = 'aaabbc'set(s)# {'a', 'b', 'c'}sorted((s.count(x), x) for x in set(s))# [(1, 'c'), (2, 'b'), (3, 'a')]import collectionsdeque = collections.deque([1,2,3])deque.append(4)a = deque.popleft()
2020-07-08 20:08:38
174
原创 pytorch笔记
trainloader = torch.utils.data.DataLoader(trainset, batch_size=batch_size, shuffle=True, num_workers=4, pin_memory=True)valloader = torch.utils.data.DataLoader(valset, batch_size=4, shuffle=False, pin_memory=True)optimizer = torch.optim.AdamW(model.par.
2020-07-07 21:47:50
582
原创 ubuntu 常用命令
解压缩# 解压 tar包tar -xvf file.tar # 解压tar.gztar -xzvf file.tar.gz # 解压 tar.bz2tar -xjvf file.tar.bz2# 解压tar.Ztar -xZvf file.tar.Z # 解压rarunrar e file.rar# 解压zipunzip file.zip查看硬盘# 路径文件夹大小du --max-depth=1 -h [path]# 路径总大小du -sh [path]# 非隐
2020-06-20 16:53:00
251
原创 salient object detection
A Simple Pooling-Based Design for Real-Time Salient Object Detection (2019 CVPR)Global Guidance Module (GGM) : PPM feature 与 FPN 每个level 融合。增加RF,缓解obj预测不完整。Feature Aggregation Module (FAM): 类似PPM,add代替concate。 1、reducing the aliasing effect of ups...
2020-06-04 18:18:00
632
原创 mmdet cpu版安装
conda create -n open-mmlab python=3.6conda activate open-mmlabconda install pytorch=1.3.1 torchvision=0.4.2 cpuonly -c pytorchpip install cython=0.29.19 git clone https://github.com/cocodataset/cocoapi.gitcd cocoapi/PythonAPImakecd ../..python c.
2020-06-03 12:31:23
1443
2
原创 mmdetection训练VOC数据
训练自己数据 VOC格式目录1、加入类别name定义2、定义dataset3、修改config.py1、加入类别name定义# mmdetection\mmdet\core\evaluation\class_names.pydef voc_trash_classes(): return [ '一次性快餐盒', '书籍纸张', '充电宝', '剩饭剩菜', '包', '垃圾桶', '塑料器皿', '塑料玩具', '塑料衣架', '大骨头',
2020-05-30 22:47:17
615
原创 Knowledge Distillation for Segmentation 笔记
目录Inter-Region Affinity Distillation for Road Marking Segmentation (2020.04)Learning Lightweight Lane Detection CNNs by Self Attention Distillation (2019.08)Knowledge Adaptation for Efficient Semantic Segmentation(CVPR 2019)Inter-Region Affinity Di
2020-05-22 17:15:04
719
1
原创 Knowledge Distillation 笔记
Inter-Region Affinity Distillation for Road Marking Segmentation (2020.04)用 inter-region affinity graph 描述structural relationship。每个node表示一种class的 areas of interest (AOI)(同类算一个 or instance算一个?),...
2020-04-15 16:33:49
1438
原创 轻量化backbone
MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications (2017.04)1、使用 depthwiseconv + pointwise conv代替 standard conv,计算量减少比例:计算量和参数量主要集中在 pointwise conv上,pointwise...
2020-04-09 11:56:53
936
原创 one/few-shot segmentation 笔记
数据集pascal-5i每个sub-dataset中,trainset来自pascal trainset,testset来自pascal valsetOne-Shot Learning for Semantic Segmentation (Sep 2017)使用dataset:pascal-5itrain时,support 和 query 均从Dtrain中抽样,cla...
2020-03-18 17:33:07
2160
3
原创 deep learning 知识点
优化都9102年了,别再用Adam + L2 regularization了AdamW and Super-convergence is now the fastest way to train neural nets
2020-02-09 22:52:18
255
原创 pytorch资源
系列pytorch学习笔记Optimizerhttps://pytorch.org/docs/stable/optim.html【PyTorch】优化器 torch.optim.Optimizer初始化pytorch学习(8) layer的抽取pytorch学习(9) 提取参数及自定义初始化模型SAVING AND LOADING MODELS详解Pytorch中的网络...
2020-02-09 09:45:02
199
原创 DUC HDC 笔记
Understanding Convolution for Semantic Segmentation1 Jun 2018thought1、增加的可学习参数形式上不一定对应,能变换回来就行2、直觉上的不合理要仔细分析motivation1、上采样更好地恢复信息,bilinear upsampling is not learnable, deconvolution zeros have...
2020-02-02 20:41:39
3491
原创 deeplab笔记
deeplab V1backbone:VGG-16 + CRF最后两层max pooling (pool4, pool5) 的 stride = 1最后三个conv层(conv5_1、conv5_2、conv5_3)的dilate rate设置为2,第一个全连接层的dilate rate设置为4during trainingtarget 为 GT label subsampled by...
2020-02-01 19:46:10
206
原创 RefineNet笔记
RefineNet: Multi-Path Refinement Networks for High-Resolution Semantic Segmentationhttps://arxiv.org/pdf/1611.06612.pdfThought1、多利用residual 结构, 包括广义的,利于优化2、与 FPN 同时间出现,当时流行特征融合institution1、subsa...
2020-01-30 13:54:23
318
原创 FDNet笔记
Learning Fully Dense Neural Networks for Image Semantic Segmentation1、融合feature map时,concate所有前层feature map,encoder network is based on the DenseNet-2642、boundary-aware loss:边界像素作为hard exampleArchi...
2020-01-29 18:26:27
1084
原创 U-Net笔记
U-Net: Convolutional Networks for Biomedical Image Segmentation18 May 2015Network ArchitectureThe architecture consists of a contracting path to capture context and a symmetric expanding path that ...
2020-01-29 17:32:23
250
原创 numpy操作
np.expand_dims(A, axis=0)np.argmax(E, axis=1)np.newaxis>> type(np.newaxis)NoneType>> np.newaxis == NoneTrue>> x = np.arange(3)>> xarray([0, 1, 2])>> x.shap...
2020-01-12 10:21:42
166
原创 Labelme Anaconda Windows安装教程
1、安装anaconda下载地址:https://mirrors.tuna.tsinghua.edu.cn/anaconda/archive/Anaconda3-5.3.1-Windows-x86_64.exe复制地址到浏览器直接下载下载完成后双击打开,如下图依次操作点击"install",开始安装点击“Finish”,安装完成。弹出的网页可直接关闭...
2019-12-26 11:43:44
448
原创 文件操作
csvimport pandas as pddata = pd.read_csv(path) # image_path, labellabel = data['label']image_path = data['image_path']csvFile = open("csvData.csv", "w") #创建csv文件writer = csv.wri...
2019-12-22 20:53:56
146
原创 Understanding the Effective Receptive Field in Deep Convolutional Neural Networks
相关参考https://zhuanlan.zhihu.com/p/86599013ERF比理论感受野小近似成高斯分布subsampling、diation conv 可增大ERF训练过程中ERF不断增大
2019-11-20 13:39:19
348
原创 生成二维高斯分布热力图
http://muyaan.com/2018/10/25/%E7%94%9F%E6%88%90%E4%BA%8C%E7%BB%B4%E9%AB%98%E6%96%AF%E5%88%86%E5%B8%83%E7%83%AD%E5%8A%9B%E5%9B%BE/
2019-11-13 20:29:22
5082
1
原创 Caffe VGG训练图片分类任务
准备数据集train.txt test.txt每行一个图片绝对路径+空格+类别标号,eg “绝对path/001.jpg 0”构建modelVGG主页http://www.robots.ox.ac.uk/~vgg/research/very_deep/下载weights(VGG_ILSVRC_16_layers.caffemodel)和网络prototxt(VGG_ILS...
2019-09-01 21:07:37
690
原创 《人生》
《人生》是路遥的早期作品,《平凡的世界》风格与其一致,都以同一个时代为背景。前者像是后者的阉版。两部作品的男主人公有很多相似之处:1. 农民出身,家境不好。2. 爱读书,有文化,有自己的想法。3.都有妹子倒追,但最后都没成(难道妹子都喜欢文艺范的男生?看来要多读点书了,哈哈)。《人生》的男主高加林在社会面前,还是太嫩。1. 得到了个写稿子的差事后,有点飘了,没有正确给自己定位。走后...
2019-08-06 11:42:23
256
原创 Debug Caffe
ImportError: cannot import name symbol_database>>> import caffeTraceback (most recent call last): File "<stdin>", line 1, in <module> File "/opt/caffe/python/caffe/__init...
2019-07-31 16:37:10
137
原创 重游十二小
己亥年庚午月丁丑日,端午假末,随母省亲,归时适过油田十二小。见正门布告曰,校欲新迁,招生已止,学期末校舍交移兰考礼部。余尝归校数次,然校门皆闭,不得入,今日门大开,似料吾将至,迎吾之归。大喜,忧时日不多,恐良机难逢,遂欣然入园。既入,环视四周,甚惊,竟未尝有变,往事历目。“希望之星”坐南朝北,正对大门,寄学子以厚望,毕业临行时,尝与友留影于此,然年久遗失,悲哉。如有留副本者,劳请告吾,感激不尽......
2019-06-21 08:47:26
219
转载 Ubuntu删除多余内核
Ubuntu删除多余内核 原文 https://www.jianshu.com/p/f284bc90944f 第一步:查看当前内核rew $ uname -aLinux rew 4.15.0-42-generic #45~16.04.1-Ubuntu SMP Mon Nov 19 13:02:27 UTC 2018 x86_64 x...
2019-06-07 09:55:02
2840
原创 Google Colaboratory使用
https://colab.research.google.comhttps://drive.google.com/在ipython中执行linux命令行:!命令挂载google drivefrom google.colab import drivedrive.mount('/content/drive/')# 点击链接登录google账户,获取认证码,输入认证码cd /con...
2019-05-17 16:11:39
207
原创 Debug RefineDet pytorch
https://github.com/lzx1413/PytorchSSD#pytorch-41-is-suppoted-on-branch-04-now环境配置问题:libstdc++.so.6: version `GLIBCXX_3.4.22' not found解决:https://blog.csdn.net/pursuit_zhangyu/article/details/79...
2019-05-15 17:58:01
946
原创 杨振宁先生UCAS讲座 感悟
直觉需要随时修正,直觉与书本知识的冲突是很好的机会。研究生最重要的不是学多少技术,而是进入一个未来5-10年会大发展的领域。这样占便宜推演法:理论->现象归纳法:现象->理论。注重新现象、新方法,少注重书本知识。遇到新现象,纳入已有理论,或修改已有理论。(杨老在西南联大推演法学的很好,在芝加哥大学吸收了归纳法,并结合二者。)研究生找题目感到沮丧是极普遍的现象。(杨老做20...
2019-04-30 13:24:01
292
原创 M2Det 阅读
@InProceedings{author = {Qijie Zhao, Tao Sheng1,Yongtao Wang, Zhi Tang, Ying Chen, Ling Cai and Haibin Ling},title = {M2Det: A Single-Shot Object Detector based on Multi-Level Feature Pyramid Networ...
2019-04-28 16:08:11
858
2
原创 Ubuntu nvidia-docker2 安装&使用
git地址:https://github.com/NVIDIA/nvidia-docker阅读readmeMake sure you have installed theNVIDIA driverand asupported versionofDockerfor your distribution (seeprerequisites).installtheNVIDIA ...
2019-04-24 11:28:38
1043
原创 FSAF 阅读
@InProceedings{author = {Chenchen Zhu, Yihui He, Marios Savvides},title = {Feature Selective Anchor-Free Module for Single-Shot Object Detection},booktitle = {The IEEE Conference on Computer Vision...
2019-04-22 21:24:09
2039
2
原创 Debug ubuntu
gcc: error trying to exec 'cc1plus' : execvp: No such file or directorysolve: https://askubuntu.com/questions/637014/gcc-error-trying-to-exec-cc1plus-execvp-no-such-file-or-directorysudo apt-get ...
2019-04-22 10:36:22
261
空空如也
TA创建的收藏夹 TA关注的收藏夹
TA关注的人