論文：Joint Face Detection and Alignment using Multi-task Cascaded Convolutional Networks 閱讀筆記
一、論文
(16)Joint Face Detection and Alignment using Multi-task Cascaded Convolutional Networks
https://arxiv.org/abs/1604.02878
二、論文筆記
1、背景
（1）、沒人把facedetection and alignment結合起來做，或者前面人做得效果不太好
（2）、難分樣本挖掘之前採用離線的方式，這樣太複雜
2、創新
（1）、提出了一個face detection和alignment聯合起來做的分三個階段的多工網路
訓練資料：
描述的太過粗糙，不太明白他的訓練過程以及inference過程
（2）、提出一個線上難樣本挖掘的方式
把每次訓練的mini batch裡面損失在前70%的樣本的loss進行回傳，剩下的不回傳
相關推薦
論文：Joint Face Detection and Alignment using Multi-task Cascaded Convolutional Networks 閱讀筆記
一、論文 (16)Joint Face Detection and Alignment using Multi-task Cascaded Convolutional Networks https://arxiv.org/abs/1604.02878
Image Splicing Localization Using A Multi-Task Fully Convolutional Network (MFCN)
論文解讀-Image Splicing Localization Using A Multi-Task Fully Convolutional Network (MFCN) 論文連結：TransForensics: Image Forgery Localization with Dense Self-Attention
深度學習論文翻譯解析（八）：Rich feature hierarchies for accurate object detection and semantic segmentation
論文標題：Rich feature hierarchies for accurate object detection and semantic segmentation 標題翻譯：豐富的特徵層次結構，可實現準確的目標檢測和語義分割
筆記：Joint Extraction of Entities and Overlapping Relations Using Position-Attentive Sequence Labeling
Joint Extraction of Entities and Overlapping Relations Using Position-Attentive Sequence Labeling 作者：Dai Dai et al., AAAI 2019.
深度學習論文翻譯解析（十）：Visualizing and Understanding Convolutional Networks
論文標題：Visualizing and Understanding Convolutional Networks 標題翻譯：視覺化和理解卷積網路
論文閱讀筆記《Deep Active Learning for Civil Infrastructure Defect Detection and Classification》
小樣本學習&元學習經典論文整理||持續更新 核心思想 本文提出一種基於主動學習的民用設施缺陷檢測方法，其思路主要是考慮到在樣本較少的情況下，訓練得到的網路可能不能很好的對各種型別的缺陷都進
OverFeat:Integrated Recognition, Localization and Detection using Convolutional Networks
目錄概主要內容 Sermanet P., Eigen D., Zhang X., Mathieu M., Fergus R., LeCun Y. OverFeat:integrated recognition, localization and detection using convolutional networks. In International Conference
os方向論文推薦：NrOS: Effective Replication and Sharing in an Operating System
關於論文NrOS: Effective Replication and Sharingin an Operating System的閱讀筆記 NrOS: Effective Replication and Sharing in an Operating System
筆記：Joint Extraction of Entities and Relations Based on a Novel Tagging Scheme
Joint Extraction of Entities and Relations Based on a Novel Tagging Scheme 作者：Suncong Zheng et al., ACL 2017.
筆記：Joint Type Inference on Entities and Relations via Graph Convolutional Networks
Joint Type Inference on Entities and Relations via Graph Convolutional Networks 作者：Sun et al., 2019 ACL
論文閱讀：Distantly Supervised Named Entity Recognition using Positive-Unlabeled Learning
題目：使用正-無標籤學習的遠端監督命名實體識別 論文地址：https://aclanthology.org/P19-1231/
深度學習論文翻譯解析（九）：Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition
論文標題：Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition 標題翻譯：用於視覺識別的深度卷積神經網路中的空間金字塔池
MIT 6.828 Lab 05：File system, Spawn and Shell
MIT 6.828 Lab5：File system, Spawn and Shell 目錄MIT 6.828 Lab5：File system, Spawn and Shell概括總結File system preliminariesOn-Disk File System Structure補充：inodeSectors and BlocksSuperblocksFile
CVPR2020｜Image Processing Using Multi-Code GAN Prior【論文閱讀筆記】
CVPR2020｜Image Processing Using Multi-Code GAN Prior【論文閱讀筆記】 作者相關連結前言介紹
【論文筆記（5）ECCV2020】Graph convolutional networks for learning with few clean and many noisy labels
Graph convolutional networks for learning with few clean and many noisy labels AbstractIntroductionRelated WrokProblem formulationCleaning with graph convolutional networksLearning a classi
FCN論文解讀：FCN-Fully Convolutional Networks for Semantic Segmentation
FCN原文作為語義分割領域的開山之作，對其進行研究和閱讀幾乎是入門語義分割領域的基礎，這篇部落格整理了自己閱讀該論文的一些心得感悟和收穫。
論文閱讀筆記：《SegNet: A Deep Convolutional Encoder-Decoder Architecture for Image Segmentation》
論文閱讀筆記：《SegNet: A Deep Convolutional Encoder-Decoder Architecture for Image Segmentation》
論文閱讀筆記：《CRNet: Cross-Reference Networks for Few-Shot Segmentation》
論文閱讀筆記：《CRNet: Cross-Reference Networks for Few-Shot Segmentation》 論文下載地址：https://arxiv.org/abs/2003.10658
論文閱讀筆記《Automatic Fabric Defect Detection with a Multi-Scale Convolutional Denoising Autoencoder Net》
核心思想 本文提出一種多尺度卷積去噪自動編碼器網路（MSCDAE）實現了對布匹紡織物缺陷的無監督檢測，演算法的思路其實並不複雜，作者使用正樣本對卷積自動編碼器（CAE）進行訓練，使得其具備提取織物特
論文閱讀筆記：Social Collaborative Filtering for Cold-start Recommendations
論文：Social Collaborative Filtering for Cold-start Recommendations / 使用者冷啟動推薦的社會化協同過濾 作者：Suvash Sedhain,Scott Sanner,Darius Braziunas,Lexing Xie,Jordan Christensen 發表刊物