yuanpengwu/paper_reading

Paper notes

This repository contains my paper reading notes on deep learning and machine learning. New year resolution for 2019: read at least one paper a week!

Summary of Topics

Topics	Description
DRL	Deep Reinforcement Learning
CLS	Classification
OD	Object Detection
InsSeg, SemSeg, PanSeg	Instance/Semantic/Panoptic Segmentation
Video	Video understanding
MI	Medical Imaging
GAN	Generative Adversarial Network
NIPS, CVPR, ICCV, ECCV, etc	Conference papers

2019-05

• Notes on Focal Loss and RetinaNet [Notes]
• Anchor Free [知乎]
• YOLO [知乎]
• RCNN, Fast RCNN, Faster RCNN, MaskRCNN[CSDN]

2019-04

• TensorMask: A Foundation for Dense Object Segmentation [Notes]
• InstanceFCN Review [Note]

2019-03

• Bag of Freebies for Training Object Detection Neural Networks [Notes]
• mixup: Beyond Empirical Risk Minimization [Notes] ICLR 2018
• Multi-view Convolutional Neural Networks for 3D Shape Recognition (MVCNN) [Notes] ICCV 2015 Does doing CNN on RGBD work?
• 3D ShapeNets: A Deep Representation for Volumetric Shapes [Notes] CVPR 2015
• Volumetric and Multi-View CNNs for Object Classification on 3D Data [Notes] CVPR 2016
• Group Normalization [Notes] ECCV 2018
• Spatial Transformer Networks [Notes]NIPS 2015

2019-02

• Quo Vadis, Action Recognition? A New Model and the Kinetics Dataset (I3D) [Notes]Video CVPR 2017
• Initialization Strategies of Spatio-Temporal Convolutional Neural Networks [Notes] Video
• Detect-and-Track: Efficient Pose Estimation in Videos [Notes] ICCV 2017 Video
• Deep Learning Based Rib Centerline Extraction and Labeling [Notes] MI MICCAI 2018
• SlowFast Networks for Video Recognition [Notes] Video
• Aggregated Residual Transformations for Deep Neural Networks (ResNeXt) [Notes] CVPR 2017.
• Beyond the pixel plane: sensing and learning in 3D (blog, 中文版本)
• VoxNet: A 3D Convolutional Neural Network for Real-Time Object Recognition (VoxNet) [Notes]
• PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation CVPR 2017 [Notes]
• PointNet++: Deep Hierarchical Feature Learning on Point Sets in a Metric Space NIPS 2017 [Notes]
• Review of Geometric deep learning 几何深度学习前沿 (from 知乎) (Up to CVPR 2018)

2019-01

• Human-level control through deep reinforcement learning (Nature DQN paper) [Notes] DRL
• Retina U-Net: Embarrassingly Simple Exploitation of Segmentation Supervision for Medical Object Detection [Notes] MI
• Panoptic Segmentation [Notes] PanSeg
• Panoptic Feature Pyramid Networks [Notes] PanSeg
• Attention-guided Unified Network for Panoptic Segmentation [Notes] PanSeg
• Bag of Tricks for Image Classification with Convolutional Neural Networks [Notes] CLS
• Deep Reinforcement Learning for Vessel Centerline Tracing in Multi-modality 3D Volumes [Notes] DRL MI
• Deep Reinforcement Learning for Flappy Bird [Notes] DRL
• Long-Term Feature Banks for Detailed Video Understanding [Notes] Video
• Non-local Neural Networks [Notes] Video CVPR 2018

2018

• Mask R-CNN
• Cascade R-CNN: Delving into High Quality Object Detection
• Focal Loss for Dense Object Detection (RetinaNet)
• Squeeze-and-Excitation Networks
• Progressive Growing of GANs for Improved Quality, Stability, and Variation
• Deformable Convolutional Networks
• Learning Region Features for Object Detection

2017 and before

• Learning notes on Deep Learning
• List of Papers on Machine Learning
• Notes of Literature Review on CNN in CV This is the notes for all the papers in the recommended list here
• Notes of Literature Review (Others)
• Notes on how to set up DL/ML environment
• Useful setup notes

Papers to Read

Here is the list of papers waiting to be read.

Deep Learning in general

• MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications (MobileNet)
• MobileNetV2: Inverted Residuals and Linear Bottlenecks (MobileNet v2)
• Xception: Deep Learning with Depthwise Separable Convolutions (Xception)
• SqueezeDet: Unified, Small, Low Power Fully Convolutional Neural Networks for Real-Time Object Detection for Autonomous Driving (SqueezeDet)
• Accurate, Large Minibatch SGD: Training ImageNet in 1 Hour
• ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness ICML 2019
• Approximating CNNs with Bag-of-local-Features models works surprisingly well on ImageNet (BagNet) blog ICML 2019
• A disciplined approach to neural network hyper-parameters: Part 1 -- learning rate, batch size, momentum, and weight decay
• Understanding deep learning requires rethinking generalization

Object detection

• Training Region-based Object Detectors with Online Hard Example Mining

Instance and Panoptic Segmentation

• Path Aggregation Network for Instance Segmentation

Video Understanding

• Learning Spatiotemporal Features with 3D Convolutional Networks (C3D) Video  ICCV 2015
• AVA: A Video Dataset of Spatio-temporally Localized Atomic Visual Actions
• Spatiotemporal Residual Networks for Video Action Recognition (decouple spatiotemporal) NIPS 2016
• Learning Spatio-Temporal Representation with Pseudo-3D Residual Networks (P3D, decouple spatiotemporal) ICCV 2017 
• A Closer Look at Spatiotemporal Convolutions for Action Recognition (decouple spatiotemporal) CVPR 2018
• Rethinking Spatiotemporal Feature Learning: Speed-Accuracy Trade-offs in Video Classification (decouple spatiotemporal) ECCV 2018
• Can Spatiotemporal 3D CNNs Retrace the History of 2D CNNs and ImageNet? CVPR 2018

Medical Imaging

• nnU-Net: Self-adapting Framework for U-Net-Based Medical Image Segmentation
• Foveal fully convolutional nets for multi-organ segmentation
• Evaluation of deep learning methods for parotid gland segmentation from CT images
• DeepMedic for Brain Tumor Segmentation
• Attention U-Net: Learning Where to Look for the Pancreas
• Attention-Gated Networks for Improving Ultrasound Scan Plane Detection
• 3D Context Enhanced Region-based Convolutional Neural Network for End-to-End Lesion Detection

Orthoganal architecture improvements

• Concurrent Spatial and Channel Squeeze & Excitation in Fully Convolutional Networks
• CBAM: Convolutional Block Attention Module

GAN

Reinforcement Learning

• Playing Atari with Deep Reinforcement Learning NIPS 2013
• Multi-Scale Deep Reinforcement Learning for Real-Time 3D-Landmark Detection in CT Scan
• An Artificial Agent for Robust Image Registration

3D Perception

• ** 3D-CNN：3D Convolutional Neural Networks for Landing Zone Detection from LiDAR**
• Generative and Discriminative Voxel Modeling with Convolutional Neural Networks
• Orientation-boosted Voxel Nets for 3D Object Recognition (ORION) <BMVC 2017>
• GIFT: A Real-time and Scalable 3D Shape Search Engine CVPR 2016
• 3D Shape Segmentation with Projective Convolutional Networks (ShapePFCN)CVPR 2017
• Learning Local Shape Descriptors from Part Correspondences With Multi-view Convolutional Networks
• Open3D: A Modern Library for 3D Data Processing
• Multimodal Deep Learning for Robust RGB-D Object Recognition IROS 2015
• ** Frustum PointNets for 3D Object Detection from RGB-D Data**
• ** FlowNet3D: Learning Scene Flow in 3D Point Clouds CVPR 2019
• Mining Point Cloud Local Structures by Kernel Correlation and Graph Pooling CVPR 2018
• PU-Net: Point Cloud Upsampling Network CVPR 2018
• Recurrent Slice Networks for 3D Segmentation of Point Clouds CVPR 2018
• SPLATNet: Sparse Lattice Networks for Point Cloud Processing CVPR 2018
• Dynamic Graph CNN for Learning on Point Clouds
• Convolutional Neural Networks on Graphs with Fast Localized Spectral Filtering NIPS 2016
• Semi-Supervised Classification with Graph Convolutional Networks ICLR 2017
• Geometric Matrix Completion with Recurrent Multi-Graph Neural Networks NIPS 2017
• Graph Attention Networks ICLR 2018
• 3D-SSD: Learning Hierarchical Features from RGB-D Images for Amodal 3D Object Detection (3D SSD)
• Escape from Cells: Deep Kd-Networks for the Recognition of 3D Point Cloud Models ICCV 2017
• PointCNN: Convolution On X-Transformed Points NIPS 2018
• PointSIFT: A SIFT-like Network Module for 3D Point Cloud Semantic Segmentation
• ** Shape Completion using 3D-Encoder-Predictor CNNs and Shape Synthesis CVPR 2017