RICAP: Random Image Cropping and Patching Data Augmentation for Deep CNNs

By Ryo Takahashi, Takashi Matsubara, Kuniaki Uehara. 2018

131 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

Citations in the corpus

CutMix: Regularization Strategy to Train Strong Classifiers With Localizable Features (2019, background): “Recently, Mixup variants [41, 34, 10, 39] have been proposed; they perform feature-level interpolation and other types of transformations.”
Shakedrop Regularization for Deep Residual Learning (2018, methodology): “Widely used regularization methods include data augmentation [18], stochastic gradient descent (SGD) [37], weight decay [20], batch normalization (BN) [15], label smoothing [28], adversarial training [6], mixup [29]–[31], [38], and dropout [26], [32].”
Fixing Data Augmentation to Improve Adversarial Robustness (2021, methodology): “The second cluster, containing RICAP, Cutout and CutMix, includes the three methods that occlude local information with patching and provide a significant boost upon the baseline with +3.06% in robust accuracy for CutMix and an average improvement of +1.79% in clean accuracy.”
A Survey of Deep Learning Techniques for Weed Detection from Images (2021, background): “, 2019) Cropping Remove a certain portion of an image (Takahashi et al., 2018).”
Data Augmentation Can Improve Robustness (2021, methodology): “For RICAP [49], the size of the four patches is determined by the image center point whose coordinates are sampled with a beta distribution Beta( α, β ) with parameters α = 0 .”
Data Augmentation Using Random Image Cropping and Patching for Deep CNNs (2018, background): “Limited preliminary results can be found in a recent conference proceeding [27].”
Remix: Rebalanced Mixup (2020, background): “erpolations of samples. Despite its simplicity, it works surprisingly well for improving generalization of deep neural networks. Mixup inspires several follow-up works like Manifold Mixup [32], RICAP [29] and CutMix [36] that have shown signicant improvement over Mixup. Mixup also shed lights upon o”
ResizeMix: Mixing Data with Preserved Object Information and True Labels (2020, background): “Rather than mixing the element-wise convex, RICAP [49] randomly gets four patches from different images and combines them to a new sample.”