iT邦幫忙

第 12 屆 iThome 鐵人賽

DAY 24
1
AI & Data

30天只學U-net系列 第 24

[day-24] U-net Data Augmentation

前言

Data Augmentation

Data augmentation is essential to teach the network the desired invariance and robustness properties, when only few training samples are available.

影像增量的3個優點

  1. invariance 變異性
  2. robustness 健碩性/魯棒性[1]
  3. few training samples 少樣本訓練

其中robustness是(應用的模型應用的關鍵環境/光線變化等)

In case of microscopical images we primarily need shift and rotation invariance as well as
robustness to deformations and gray value variations.

這一段要我們去思考,這種case需要注意哪一種的 robustness,以顯微鏡細胞圖的例子:

  1. 旋轉
  2. gray value的不確定

Especially random elastic deformations of the training samples seem to be the key concept to train a segmentation network with very few annotated images. We generate smooth deformations using random displacement vectors on a coarse 3 by 3 grid.

這邊提出彈性變形是在少樣本分類的關鍵。所以用了randome displacement在3X3的網格上。

The displacements are sampled from a Gaussian distribution with 10 pixels standard
deviation. Per-pixel displacements are then computed using bicubic interpolation.

說明如何彈性變形。

Drop-out layers at the end of the contracting path perform further implicit data augmentation.

這邊說明的加入 drop out layer 目的是做data augmentation

Conclusion

今天又看到新的東西了,drop out 跟 data augmentation 竟然有關連性,明天我們看看到底是怎麼回事吧~~

Reference

[1]Model-Based Robust Deep Learning
[2]Dropout as data augmentation


上一篇
[day-23] U-net Training 細節 (5) - initial weight
下一篇
[day-25] U-net Experiments (1) - rule
系列文
30天只學U-net30

尚未有邦友留言

立即登入留言