2019 年下半年 CV 类论文阅读情况

Classification

• InceptionV1-V4, Inception-ResNet(10.8), Xception(12.5)
• ResNet(9.2), DenseNet(12.4)
• SqueezeNet(12.5)
• MoblieNetv1(12.5), MobileNetv2(12.11), MobileNetv3
• EfficientNet(12.13)
• ReNeXt(12.6)
• ShuffleNetv1(12.5), ShuffleNetv2(12.11)
• NasNet, MnasNet
• DPN
• MixNet：谷歌新出最强轻量级模型
• SEnet(12.12)

Object Detection

• R-CNN(10.10), Fast R-CNN(10.24), Faster R-CNN(10.25), Mask R-CNN
• SPPnet(10.23)
• YOLOv1(11.28), YOLO9000(12.3), YOLOv3
• FreeAnchor
• CornetNet
• CenterNet
• SSD(12.4)
• EdgeNet：速度最快
• EfficientDet：性能最强
• ASFF
• FPN
• MnasFPN：最强轻量级模型
• RetinaNet
• TridentNet

Semantic Segmentation

• PolarMask
• FCN(11.15), R-FCN
• U-Net(11.18), U-Net++, mU-Net
• ICNet
• SegNet(12.4)
• PSPNet
• LinkNet
• DeepLabv1, DeepLabv2, DeepLabv3, DeepLabv3+(12.15)

其它

Face Recognition: From Traditional to Deep Learning Methods：人脸识别综述

A Survey of Deep Learning Techniques for Autonomous Driving：自动驾驶综述

USE OF A CAPSULE NETWORK TO DETECT FAKE IMAGES AND VIDEOS：伪造图片和视频识别

Style Mixer: Semantic-aware Multi-Style Transfer Network：风格迁移新网络

Momentum Contrast for Unsupervised Visual Representation Learning：恺明大佬的无监督语义分割

On the Relationship between Self-Attention and Convolutional Layers：自注意力机制与卷积层之间的关系

Multi-Label Learning with Deep Forest：深度森林多标签学习

Faster AutoAugment: Learning Augmentation Strategies using Backpropagation：faster自动增强（前面还有谷歌的fastaa、aa、randaa）

Research Guide: Model Distillation Techniques for Deep Learning：知识蒸馏技术研究指南

Object Detection in 20 Years: A Survey：目标检测综述

语义分割综述盘点

Deep Learning for Visual Tracking: A Comprehensive Survey：目标跟踪综述

Ultrafast Photorealistic Style Transfer via Neural Architecture Search：NAS风格迁移

Advances and Open Problems in Federated Learning：联邦学习综述

Analyzing and Improving the Image Quality of StyleGAN：最强大的风格迁移StyleGAN2

在计算机视觉深度学习方向，这几个月对基本分类网络、轻量级网络、目标检测、语义分割几个方向进行了了解，快要期末考了，剩下的交给2020年了(╯#-_-)╯。预计下学期重点会放在人脸识别、实例分割、GAN、NAS 上。