sunjac/cv_paperdaily

CV 每日论文阅读笔记（日更）

Notice！

Busy with recent things , I'll update without day.

Notes are attached to raw PDF files from now on .

CV--PaperDaily

创建人	知乎论文阅读专栏	个人博客	其他
ming71	论文笔记入口	chaser	CSDN

Update CV papers here everday .
The content includes but is not limited to Object detection , Semantic segmentation , and other papers about deep learning .
Comments are welcome , and you can e-mail me by [email protected] .

Paper reading

Divided by Conference & Journal .

AAAI

• M2Det: A Single-Shot Object Detector based on Multi-Level Feature Pyramid
• Weakly Supervised Semantic Segmentation Using Superpixel Pooling Network
• Gradient Harmonized Single-stage Detector

---

CVPR

• Assisted Excitation of Activations: A Learning Technique to Improve Object
• Borrow from Anywhere Pseudo Multi-modal Object Detection in Thermal Imagery
• Cascade R-CNN: Delving into High Quality Object Detection
• Feature Pyramid Networks for Object Detection
• Not All Pixels Are Equal: Difficulty-Aware Semantic Segmentation via Deep Layer Cascade
• Path Aggregation Network for Instance Segmentation
• Region Proposal by Guided Anchoring
• Scale-Transferable Object Detection
• DOTA: A Large-scale Dataset for Object Detection in Aerial Images
• R2CNN: Rotational Region CNN for Orientation Robust Scene Text Detection
• Pseudo Mask Augmented Object Detection
• Single-Shot Object Detection with Enriched Semantics
• Weakly Supervised Instance Segmentation using Class Peak Response
• Learning Deep Features for Discriminative Localization
• Simple Does It: Weakly Supervised Instance and Semantic Segmentation
• Weakly Supervised Learning of Instance Segmentation with Inter-pixel Relations
• Panoptic Segmentation
• Learning Instance Activation Maps for Weakly Supervised Instance Segmentation
• Generalized Intersection over Union: A Metric and A Loss for Bounding Box Regression
• Single-Shot Refinement Neural Network for Object Detection
• ScratchDet : Training Single-Shot Object Detectors
• Squeeze-and-Excitation Networks
• Dual Attention Network for Scene Segmentation
• An Analysis of Scale Invariance in Object Detection
• Rotation Sensitive Regression for Oriented Scene Text Detection
• Oriented Response Networks

---

ECCV

• DetNet: A Backbone network for Object Detection
• Receptive Field Block Net for Accurate and Fast Object Detection
• Modeling Visual Context is Key to Augmenting Object Detection Datasets
• Contextual Priming and Feedback for Faster R-CNN
• Learning to Segment via Cut-and-Paste
• Acquisition of Localization Confidence for Accurate Object Detection

---

ICCV

• Focal Loss for Dense Object Detection
• InstaBoost: Boosting Instance Segmentation via Probability Map Guided
• Scale-Aware Trident Networks for Object Detection
• EGNet: Edge Guidance Network for Salient Object Detection
• ThunderNet: Towards Real-time Generic Object Detection
• Cut, Paste and Learn: Surprisingly Easy Synthesis for Instance Detection
• FCOS: Fully Convolutional One-Stage Object Detection
• Gaussian YOLOv3: An Accurate and Fast Object Detector Using Localization Uncertainty for Autonomous Driving
• Single Shot Text Detector with Regional Attention

---

ICML

• Making Convolutional Networks Shift-Invariant Again
• How much real data do we actually need: Analyzing object detection performance using synthetic and real data (workshop)

---

ICLR

• Why do deep convolutional networks generalize so poorly to small image transformations?
• Dataset Augmentationin In Feature Space
• ImageNet-trained CNNs are biased towards texture: increasing shape bias improves accuracy and robustness
• Approximating CNNs with Bag-of-local-Features models works surprisingly well on ImageNet

---

ArXiv

• FSSD: Feature Fusion Single Shot Multibox Detector
• MDSSD: Multi-scale Deconvolutional Single Shot Detector for Small Objects
• MMDetection: Open MMLab Detection Toolbox and Benchmark
• Double-Head RCNN: Rethinking Classification and Localization for Object Detection
• Learning Data Augmentation Strategies for Object Detection
• A Preliminary Study on Data Augmentation of Deep Learning for Image Classification
• Improved Regularization of Convolutional Neural Networks with Cutout
• Data Augmentation by Pairing Samples for Images Classification
• Bag of Freebies for Training Object Detection Neural Networks
• The Effectiveness of Data Augmentation in Image Classification using Deep Learning
• Natural Adversarial Examples
• Recent Advances in Deep Learning for Object Detection
• Matrix Nets: A New Deep Architecture for Object Detection
• Needles in Haystacks: On Classifying Tiny Objects in Large Images
• CBNet: A Novel Composite Backbone Network Architecture for Object Detection
• Light-Head R-CNN: In Defense of Two-Stage Object Detector
• R3Det Refined Single-Stage Detector with Feature Refinement for Rotating Object
• Beyond Skip Connections: Top-Down Modulation for Object Detection
• Deep Learning for 2D and 3D Rotatable Data An Overview of Methods
• Is Sampling Heuristics Necessary in Training Deep Object Detectors
• A Real-Time Scene Text Detector with Learned Anchor
• RAM: Residual Attention Module for Single Image Super-Resolution

---

Others

• (Acess) Smart Augmentation: Learning an Optimal Data Augmentation Strategy
• (ICANN) Further advantages of data augmentation on convolutional neural networks
• (WACV) Understanding Convolution for Semantic Segmentation
• (BMVC) Enhancement of SSD by concatenating feature maps for object detection
• (Big Data) A survey on Image Data Augmentation for Deep Learning
• (DICTA) Understanding data augmentation for classification: when to warp?
• (IJCV) What Makes Good Synthetic Training Data for Learning Disparity and Optical Flow Estimation?
• (ACCV) Reverse Densely Connected Feature Pyramid Network for Object Detection
• (IJAC) An Overview of Contour Detection Approaches
• (ICIP) SSSDET: Simple Short and Shallow Network for Resource Efficient Vehicle Detection in Aerial Scenes
• (Remote Sensing) Automatic Ship Detection of Remote Sensing Images from Google Earth in Complex Scenes Based on Multi-Scale Rotation Dense Feature Pyramid Networks
• (Multimedia) Arbitrary-oriented scene text detection via rotation proposals
• (NIPS) R-FCN： Object Detection via Region-based Fully Convolutional Networks
• (JMLR) Neural Architecture Search: A Survey
• (OCEANS) Ship Detection: An Improved YOLOv3 Method
• (VISIGRAPP) Learning Transformation Invariant Representations with Weak Supervision
• (ICDAR) ICDAR 2015 competition on Robust Reading
• (BMVC) Rethinking Classification and Localization for Cascade R-CNN

---