跳转至

论文汇总

论文分类

Backbone

论文 年份 简介
VGG43 2014.09 VGG Net
ResNet13 2015.12 ResNet
DenseNet44 2016.08 DenseNet

目标检测

论文 年份 简介
YOLOv138 2015.6 YOLOv1
YOLOv248 2016.12 YOLOv2
YOLOv349 2018.04 YOLOv3
YOLOv450 2020.04 YOLOv4
DETR55 2020.05 DETR，Transformer直接查询BBOX
YOLOv5 2020.06 YOLOv5
70 2020.08 OBB对象检测
YOLOv651 2022.09 YOLOv6

目标分割

论文 年份 简介
FCN41 2014.11 全卷积网络，语义分割开山之作

关键点

论文 年份 简介
61 2014.06 早期关键点检测工作
60 2019.08 两个步骤，一个是回归热图，一个是精修热图
59 2019.11 受FCOS启发，以做检测的思路来做关键点检测，直接回归关键点的位置

UNet系列

论文 年份 简介
1 2015.05 UNet开山之作，提出了包含编码路径和解码路径的U形网络
3D UNet2 2016.06 做了数据增强，尤其是弹性形变
UNet++17 2018.07 增加网络之间的连接节点
nnU-Net5 2018.09 主要做流程标准化及自动化
nnU-Net6 2019.04 nnU-Net的第二篇论文，做了大量实验放了一些比赛结果来表明nnU-Net的有效性
37 2019.09 深度可分离卷积，做3D医学图像分割
\(U^2-Net\)16 2020.05 UNet里套UNet
15 2021.02 CNN提底层特征，Transformer提上层远程特征
12 2021.10 通过大量消融实验来试，哪个效果好用哪个
UNeXt32 2022.03 加入MLP机制，减少参数提高效率
优化的nnU-Net28 2022.07 优化nnU-Net，通过大模型训练小模型，用上了未标注的数据集

Transformer系列

论文 年份 简介
Attention is all you need18 2017.06 Transformer的开山之作，发扬光大
VIT39 2020.10 第一次将纯Transformer用于图像分类
SwinTransformer42 2021.03 Swin-Transformer
Swin-UNet40 2021.05 U型Transformer
MaskFormer79 2021.07 把语义分割从像素级分类任务改成了mask级分割+分类任务
Mask DINO23 2022.06 基于Transformer的图片检测分割统一框架
81 2022.11 TODO
80 2023.09 TODO
82 2023.11 TODO

自监督、半监督、无监督

牙齿、颌骨检测

论文 年份 简介
33 2021.02 类似CenterNet，通过损失函数惩罚解决牙齿重叠的问题
56 2022.05 基于YOLO V3的牙齿检测方法，单层检测牙齿然后算出来BBOX

牙齿、颌骨分割

论文 年份 简介
14 2019.07 牙齿分割，先检测牙齿边缘然后用于指导牙齿分割。此外，提出了相似性矩阵用于替代传统NMS来解决牙齿检测问题中不容易NMS的问题
8 2020.04 牙齿实例分割，主要工作是1、回归牙齿中心点用于指导牙齿定位 2、边界敏感的Dice Loss，优化边界的分割
11 2021.06 牙齿分割，类似8的思路，先回归牙齿的质心和牙齿的骨架，然后用于指导牙齿分割
10 2022.01 颌骨分割
3 2022.04 做了牙齿和颌骨分割
7 2021.07 下颌骨分割的论文综述

牙齿数据融合

论文 年份 简介
66 2021.12 CBCT和口扫模型（IOS模型）融合
4 2022.03 CBCT和口扫模型（IOS模型）融合

损失函数

论文 年份 简介
9 2020.03 主要用于血管等管状细小物体的分割的损失函数，可以提升分割稳定性和连续性，防止断裂

MONAI

论文 年份 简介
UNETR31 2021.03 U型的Transformer，Transformer提特征然后U型构建
SwinUnetr30 2021.11 Transformer医学图像预训练
35 2022.09 HECKTOR 2022第一名技术报告

数据集

论文 年份 简介
68 2015 SUN RGB-D

3D点云

论文 年份 简介
78 2008.09 点云对齐中使用的PFH特征
77 2009.08 点云对齐中使用的FPFH特征，78的延续工作
76 2015 Open3d使用的RANSAC点云配置方法
72 2017.11 71从这里参考的回归角度任务设计成分类+回归的任务，这个又是参考的73
71 2019.04 3DETR参考的论文，主要是3D检测
57 2021.09 3DETR，DETR的3D点云版本

其他

论文 年份 简介
73 2016.12 2D图片算3D BBOX，提出了把角度回归改成多个bin的分类+残差角度回归，后面的VoteNet、3DETR都是受他影响
69 2018.01 人脸识别的，ArcLoss
19 2019.08 高分辨率的特征图，用于关键点、分割等下游任务
Wave2lip36 2020.08 Wave2lip，语音驱动的任意人脸视频口型生成
53 2020.09 深度学习算法医学图像分割的一篇综述，基本涉及到了医学图像分割的方方面面
64 2020.12 抠图
63 2021.01 人脸关键点检测综述
65 2021.01 三维点云数据预训练及数据增强
22 2022.03 公开了一个高分辨率的分割数据集，包含5000张的数据
47 2022.03 医学图像自监督的工作，感觉比较偏理论，没有特别仔细看
58 2022.08 使用nnUNet分割了大量的器官，并提供python包方便使用
52 2022.09 医学图像炼丹trick避坑指南，做了大量对比实验验证医学图像分割中各种trick的效果

相关问题

牙齿检测中的NMS问题

牙齿由于比较密集，且空间形状和朝向较为复杂，使用简单的基于IOU的NMS难以达到很好的效果，NMS的阈值难以确定，针对这个问题目前看到两篇论文进行了尝试解决

  • 14， 通过在训练过程中学习一个相似性矩阵来来学习牙齿之间的相似性，然后通过相似性矩阵来进行NMS
  • 33，类似CenterNet的检测模型，通过损失函数惩罚高斯热图，引导网络使得检测到的牙齿之间不会重叠

TODO

20.21.24.26.27.29.34.45.46.

  1. U-Net: Convolutional Networks for Biomedical Image Segmentation ↩

  2. 3D U-Net: Learning Dense Volumetric Segmentation from Sparse Annotation ↩

  3. A fully automatic AI system for tooth and alveolar bone segmentation from cone-beam CT images ↩

  4. AI-enabled Automatic Multimodal Fusion of Cone-Beam CT and Intraoral Scans for Intelligent 3D Tooth-Bone Reconstruction and Clinical Applications ↩

  5. nnU-Net: Self-adapting Framework for U-Net-Based Medical Image Segmentation ↩

  6. Automated Design of Deep Learning Methods for Biomedical Image Segmentation ↩

  7. Automatic Segmentation of Mandible from Conventional Methods to Deep Learning—A Review ↩

  8. Center-Sensitive and Boundary-Aware Tooth Instance Segmentation and Classification from Cone-Beam CT ↩↩

  9. clDice：A Novel Topology-Preserving Loss Function for Tubular Structure Segmentation ↩

  10. Deep Learning-Based Automatic Segmentation of Mandible and Maxilla in Multi-Center CT Images ↩

  11. Hierarchical Morphology-Guided Tooth Instance Segmentation from CBCT Images ↩

  12. Optimized U-Net for Brain Tumor Segmentation ↩

  13. Deep Residual Learning for Image Recognition ↩

  14. ToothNet：Automatic Tooth Instance Segmentation and Identification from Cone Beam CT Images ↩↩

  15. TransUNet：Transformers Make Strong Encoders for Medical Image Segmentation ↩

  16. U2-Net：Going Deeper with Nested U-Structure for Salient Object Detection.md ↩

  17. UNet++： A Nested U-Net Architecture for Medical Image Segmentation.md ↩

  18. Attention Is All You Need ↩

  19. Deep High-Resolution Representation Learning for Visual Recognition ↩

  20. Focused Decoding Enables 3D Anatomical Detection by Transformers ↩

  21. Generalized Decoding for Pixel, Image, and Language ↩

  22. Highly Accurate Dichotomous Image Segmentation ↩

  23. Mask DINO：Towards A Unified Transformer-based Framework for Object Detection and Segmentation ↩

  24. Masked-attention Mask Transformer for Universal Image Segmentation ↩

  25. MedSegDiff：Medical Image Segmentation with Diffusion Probabilistic Model ↩

  26. OneFormer：One Transformer to Rule Universal Image Segmentation ↩

  27. Polarized Self-Attention：Towards High-quality Pixel-wise Regression ↩

  28. Revisiting nnU-Net for Iterative Pseudo Labeling and Efficient Sliding Window Inference ↩

  29. Rich feature hierarchies for accurate object detection and semantic segmentation ↩

  30. Swin UNETR：Swin Transformers for Semantic Segmentation of Brain Tumors in MRI Images ↩

  31. UNETR: Transformers for 3D Medical Image Segmentation ↩

  32. UNeXt：MLP-based Rapid Medical Image Segmentation Network ↩

  33. Tooth Instance Segmentation from Cone-Beam CT Images through Point-based Detection and Gaussian Disentanglement ↩↩

  34. CenterNet：Keypoint Triplets for Object Detection ↩

  35. Automated head and neck tumor segmentation from 3D PET/CT ↩

  36. A Lip Sync Expert Is All You Need for Speech to Lip Generation In The Wild ↩

  37. 3D U2-Net：A 3D Universal U-Net for Multi-Domain Medical Image Segmentation ↩

  38. You Only Look Once：Unified, Real-Time Object Detection.md ↩

  39. An Image is Worth 16x16 Words：Transformers for Image Recognition at Scale ↩

  40. Swin-Unet：Unet-like Pure Transformer for Medical Image Segmentation ↩

  41. Fully Convolutional Networks for Semantic Segmentation ↩

  42. Swin Transformer：Hierarchical Vision Transformer using Shifted Windows ↩

  43. Very Deep Convolutional Networks for Large-Scale Image Recognition ↩

  44. Densely Connected Convolutional Networks ↩

  45. DiRA：Discriminative, Restorative, and Adversarial Learning for Self-supervised Medical Image Analysis ↩

  46. ACPL：Anti-curriculum Pseudo-labelling for Semi-supervised Medical Image Classification ↩

  47. BoostMIS：Boosting Medical Image Semi-supervised Learning with Adaptive Pseudo Labeling and Informative Active Annotation ↩

  48. YOLO9000：Better, Faster, Stronger ↩

  49. YOLOv3：An Incremental Improvement ↩

  50. YOLOv4：Optimal Speed and Accuracy of Object Detection ↩

  51. YOLOv6：A Single-Stage Object Detection Framework for Industrial Applications ↩

  52. Deep Learning for Medical Image Segmentation：Tricks, Challenges and Future Directions ↩

  53. Medical Image Segmentation Using Deep Learning：A Survey ↩

  54. Boundary loss for highly unbalanced segmentation ↩

  55. End-to-End Object Detection with Transformers ↩

  56. A Combined Approach for Accurate and Accelerated Teeth Detection on Cone Beam CT Images ↩

  57. An End-to-End Transformer Model for 3D Object Detection ↩

  58. TotalSegmentator：robust segmentation of 104 anatomical structures in CT images ↩

  59. DirectPose：Direct End-to-End Multi-Person Pose Estimation ↩

  60. Integrating Spatial Configuration into Heatmap Regression Based CNNs for Landmark Localization ↩

  61. Joint Training of a Convolutional Network and a Graphical Model for Human Pose Estimation ↩

  62. Explicit Box Detection Unifies End-to-End Multi-Person Pose Estimation ↩

  63. Fast Facial Landmark Detection and Application ↩

  64. Real-Time High-Resolution Background Matting ↩

  65. Self-Supervised Pretraining of 3D Features on any Point-Cloud ↩

  66. Fully automatic integration of dental CBCT images and full-arch intraoral impressions with stitching error correction via individual tooth segmentation and identification ↩

  67. Deep learning method for reducing metal artifacts in dental cone-beam CT using supplementary information from intra-oral scan ↩

  68. SUN RGB-D：A RGB-D Scene Understanding Benchmark Suite ↩

  69. ArcFace：Additive Angular Margin Loss for Deep Face Recognition ↩

  70. Align Deep Features for Oriented Object Detection ↩

  71. Deep Hough Voting for 3D Object Detection in Point Clouds ↩↩

  72. Frustum PointNets for 3D Object Detection from RGB-D Data ↩

  73. 3D Bounding Box Estimation Using Deep Learning and Geometry ↩↩

  74. Image-to-Image Translation with Conditional Adversarial Networks ↩

  75. Segment Anything ↩

  76. Robust Reconstruction of Indoor Scenes ↩

  77. Fast Point Feature Histograms (FPFH) for 3D Registration ↩

  78. Aligning Point Cloud Views using Persistent Feature Histograms ↩↩

  79. Per-Pixel Classification is Not All You Need for Semantic Segmentation ↩

  80. Mask-Attention-Free Transformer for 3D Instance Segmentation ↩

  81. Superpoint Transformer for 3D Scene Instance Segmentation ↩

  82. OneFormer3D: One Transformer for Unified Point Cloud Segmentation ↩

  83. Query Refinement Transformer for 3D Instance Segmentation ↩

评论