workshop
-
Robust vision challenge 2018
robust vision.JPG
- 语义分割11个submissions，
第一名 Peter Samuel Rota Bulo (Mapillary): In-Place Activated BatchNorm for Memory-Optimized Training of DNNs
key observation of 当前分割方法
我们关注的是减少memory consume in 训练，因为这可以让我们用larger crops, larger batch size or deeper models.
image.png
image.png
image.png
image.png
结论
大数据集上预训练非常有用！！！
第二名 Marin Oršić (Uni Zagreb): Ladder-DenseNet Architecture for Robust Semantic Segmentation
image.png
image.png
image.png
image.png
Instance segment 3 submissions.
Shou-Yao Roy Tseng (NTHU): Non-local RoI for Instance Segmentation
Hsien-Tzu Cheng (NTHU): Improved MaskRCNN
-
Invited talk
- Stefan Roth (TU Darmstadt): Robust Scene Analysis: Energy-based models, deep learning, and something in between
- Uwe Franke (Daimler AG): 30 Years Fighting for Robustness
- Judy Hoffman (UC Berkeley): Making our Models Robust to Changing Visual Environments
- CV for microscopy image analysis
Cell Image Segmentation by Integrating Multiple CNNs
Comparison of deep transfer learning strategies for digital pathology
Understanding Pixel-to-Label Prediction Model
Invited Talk: Life and death decisions- classification, characterization and predictions of death in neuronal models of neurodegenerative disease (Jeremy Linsley (UCSF))
- Deep Learning for Biomedical Imaging: Can We Get Better, Higher or Faster? Tammy Riklin Raviv, PhD
- Deep Lesion Database and Deep Lesion Graphs on Relationship Learning and Organization of Significant Radiology Image Findings Ke YAN's blog
- Mine Deeper & Learn Wider: a Perspective on Distilling Radiological Reports for Chest X-ray Analysis Wang Xiaosong
ChestX-ray8 数据贡献者
- Population imaging analytics: progress, challenges and opportunities
感兴趣文章
- MIT- Learning Network Architectures of Deep CNNs under Resource Constraints
- Recurrent Segmentation for Variable Computational Budgets
- Probabilistic model for 3D interactive segmentation
MIT 周博磊等人组织
