Must Read Papers

Selected Papers and blogs for Perception Deep Learning

Must read papers on CNN

A brief note on these selected papers:

Popular Backbone for CNN

• Alexnet: review on alexnet

• VGGNet: review on VGGNet

• ResNet 2015

  • An Overview of ResNet and its Variants

  • Understanding and visualizing ResNets

• Inception: simple guide to Inception

  • v1 2014: GoogLeNet

  • v3: 2015, review on v3

  • v4: paper, review on v4

• Feature Pyramid Network: review on FPN

Object Detection

1. Two-Stage Detector

2. R-CNN: reading on R-CNN and variants

   • R-CNN 2013 , review on RCNN

   • Fast R-CNN: 2015

   • Faster R-CNN: 2015

2.One-Stage Detector

• YOLO

  • v1 2016: v2

    • YOLO explained, YOLO in python, YOLO simpler explanation

    • YOLOv1 CVPR2016 presentation

    • Official slides, resource for YOLO

    • YOLOv2 simple explain

  • v3 2018:

    • beginner guide, theory explained , whats new inYOLOv3

    • YOLOv3 in Keras, YOLOv3 in PyTorch

  • v4: reading

  • v5: reading

• SSD

• RetinaNet 2017: review on retinanet

• EfficientDet: 2019

• EfficientNet: 2019

• MobileNets

• Squeezenet

• ShuffleNet

Review on Object Detection

• Object Detection in 20 Years: A Survey

Image from zou2019

Advances in 2D Computer Vision

Recent Advances in Modern Computer VisionMedium

Review: 3D object detection using LiDAR

• ****The state of 3D object detection: A review of the state of the art based upon the KITTI leaderboard****

Detector (LIDAR only) latency vs vehicle AP

Image General approaches for LIDAR+RGB fusion. Images are adapted from MV3D (Chen et. at. 2016), F-Pointnet (Qi et. al. 2017), ContFuse (Liang et. al. 2018), and LaserNet (Meyer et. al. 2018).for post

Detector (LIDAR+RGB fusion labeled) latency vs vehicle APost

Trade-offs between RV and BEV projectionspost

Figure from PointPillars

3D Object Orientation Detection

Multimodal Regression — Beyond L1 and L2 LossMedium

https://towardsdatascience.com/orientation-estimation-in-monocular-3d-object-detection-f850ace91411towardsdatascience.com

Pseudo LiDAR: from 2D to 3D

Making a Pseudo LiDAR With Cameras and Deep LearningMedium

Time Series Classification

• Deep Learning for Time series classification: a review

• InceptionTime: https://towardsdatascience.com/deep-learning-for-time-series-classification-inceptiontime-245703f422db