Filters








37 Hits in 1.5 sec

Differentiable Data Augmentation with Kornia [article]

Jian Shi, Edgar Riba, Dmytro Mishkin, Francesc Moreno, Anguelos Nicolaou
2020 arXiv   pre-print
This module leverages differentiable computer vision solutions from Kornia, with an aim of integrating data augmentation (DA) pipelines and strategies to existing PyTorch components (e.g. autograd for  ...  In this paper we present a review of the Kornia differentiable data augmentation (DDA) module for both for spatial (2D) and volumetric (3D) tensors.  ...  Discussion We presented Kornia DDA that aligned with PyTorch API design principles with a focus on usability, to perform efficient differentiable augmentation pipelines for both production and research  ... 
arXiv:2011.09832v1 fatcat:p6wgj4eumrdmhja6dyvnwfmkfu

A survey on Kornia: an Open Source Differentiable Computer Vision Library for PyTorch [article]

E. Riba, D. Mishkin, J. Shi, D. Ponsa, F. Moreno-Noguer, G. Bradski
2020 arXiv   pre-print
This work presents Kornia, an open source computer vision library built upon a set of differentiable routines and modules that aims to solve generic computer vision problems.  ...  The package uses PyTorch as its main backend, not only for efficiency but also to take advantage of the reverse auto-differentiation engine to define and compute the gradient of complex functions.  ...  Example 1: Data augmentation pipeline Example showing how flexible is Kornia to define a data augmentation pipeline using other PyTorch components.  ... 
arXiv:2009.10521v1 fatcat:wjyp5nzbfzfmjpvuxfaxydh5ra

Kornia: an Open Source Differentiable Computer Vision Library for PyTorch [article]

Edgar Riba, Dmytro Mishkin, Daniel Ponsa, Ethan Rublee, Gary Bradski
2019 arXiv   pre-print
This work presents Kornia -- an open source computer vision library which consists of a set of differentiable routines and modules to solve generic computer vision problems.  ...  The package uses PyTorch as its main backend both for efficiency and to take advantage of the reverse-mode auto-differentiation to define and compute the gradient of complex functions.  ...  With the rise of deep learning, most of the standard computer vision frameworks have moved to being used more for certain geometric vision functions, data pre-processing, data augmentation on the CPU in  ... 
arXiv:1910.02190v2 fatcat:shnrjgnvyfbgzaskdjznlb6psu

Kornia: an Open Source Differentiable Computer Vision Library for PyTorch

Edgar Riba, Dmytro Mishkin, Daniel Ponsa, Ethan Rublee, Gary Bradski
2020 2020 IEEE Winter Conference on Applications of Computer Vision (WACV)  
This work presents Kornia -an open source computer vision library which consists of a set of differentiable routines and modules to solve generic computer vision problems.  ...  The package uses PyTorch as its main backend both for efficiency and to take advantage of the reverse-mode auto-differentiation to define and compute the gradient of complex functions.  ...  , data augmentation on the CPU in order to be transferred later to the GPU as well as post processing to refine results.  ... 
doi:10.1109/wacv45572.2020.9093363 dblp:conf/wacv/RibaMPRB20 fatcat:dtyl4bijcfee3dv7se3hhgfmpa

Kaolin: A PyTorch Library for Accelerating 3D Deep Learning Research [article]

Krishna Murthy Jatavallabhula, Edward Smith, Jean-Francois Lafleche, Clement Fuji Tsang, Artem Rozantsev, Wenzheng Chen, Tommy Xiang, Rev Lebaredian, Sanja Fidler
2019 arXiv   pre-print
Kaolin provides efficient implementations of differentiable 3D modules for use in deep learning systems.  ...  Kaolin packages together several differentiable graphics modules including rendering, lighting, shading, and view warping.  ...  This allows for operator overloading over common functions for data augmentation and modifications supported by the package.  ... 
arXiv:1911.05063v2 fatcat:2mbh3eqljrfv5ervwrwtcu54ry

Local Patch AutoAugment with Multi-Agent Collaboration [article]

Shiqi Lin, Tao Yu, Ruoyu Feng, Xin Li, Xin Jin, Zhibo Chen
2021 arXiv   pre-print
Data augmentation (DA) plays a critical role in improving the generalization of deep learning models. Recent works on automatically searching for DA policies from data have achieved great success.  ...  The agents cooperate with each other to achieve the optimal augmentation effect of the entire image by sharing a team reward.  ...  Therefore, we compare the feedback of target network φ(•) on the augmented data processed by our proposed PAA x with the original data x and take their difference on the training losses as the reward for  ... 
arXiv:2103.11099v2 fatcat:zxbiwct4urepxcmen7uba76iki

Faster AutoAugment: Learning Augmentation Strategies using Backpropagation [article]

Ryuichiro Hataya, Jan Zdenek, Kazuki Yoshizoe, Hideki Nakayama
2019 arXiv   pre-print
As the objective of training, we minimize the distance between the distributions of augmented data and the original data, which can be differentiated.  ...  In this paper, we propose a differentiable policy search pipeline for data augmentation, which is much faster than previous methods.  ...  We transplanted the operations described in section 3.1 to PyTorch [22] , a tensor computation library with automatic differentiation. For geometric operations, we extend functions in kornia [25] .  ... 
arXiv:1911.06987v1 fatcat:euoxrg2coredvlsbfkqod3wr2i

Self-Distilled Hashing for Deep Image Retrieval [article]

Young Kyun Jang, Geonmo Gu, Byungsoo Ko, Nam Ik Cho
2021 arXiv   pre-print
To mitigate this issue, data augmentation can be applied during training.  ...  In this work, we propose a novel self-distilled hashing scheme to minimize the discrepancy while exploiting the potential of augmented data.  ...  Instead of including non-differentiable quantization process in model training, we learn H in the real space to estimate optimal b with continuously relaxed h while fully exploiting the power of data augmentation  ... 
arXiv:2112.08816v1 fatcat:32wo44a3yfdmjp7hmsflvtchui

Learning Data Augmentation with Online Bilevel Optimization for Image Classification [article]

Saypraseuth Mounsaveng, Issam Laradji, Ismail Ben Ayed, David Vazquez, Marco Pedersoli
2020 arXiv   pre-print
This framework can be used as a general solution to learn the optimal data augmentation jointly with an end task model like a classifier.  ...  Data augmentation is a key practice in machine learning for improving generalization performance.  ...  In order to have differentiable affine and color transformations, we use the Kornia [43] library and the affine grid and grid samples functions of the torchvision package of pytorch framework.  ... 
arXiv:2006.14699v2 fatcat:bsolo6fsjneh3kknfbs3kbmccq

Meta Approach to Data Augmentation Optimization [article]

Ryuichiro Hataya, Jan Zdenek, Kazuki Yoshizoe, Hideki Nakayama
2020 arXiv   pre-print
Data augmentation policies drastically improve the performance of image recognition tasks, especially when the policies are optimized for the target data and tasks.  ...  In this paper, we propose to optimize image recognition models and data augmentation policies simultaneously to improve the performance using gradient descent.  ...  Differentiable Data Augmentation To differentiate through φ, MADAO adopts the differentiable data augmentation pipeline following [17] .  ... 
arXiv:2006.07965v1 fatcat:gje32fsvqnaelipg4n5wc5eq3e

Deep invariant networks with differentiable augmentation layers [article]

Cédric Rommel, Thomas Moreau, Alexandre Gramfort
2022 arXiv   pre-print
It can incorporate any type of differentiable augmentation and be applied to a broad class of learning problems beyond computer vision.  ...  State-of-the-art methods for learning data augmentation policies require held-out data and are based on bilevel optimization problems, which are complex to solve and often computationally demanding.  ...  In this experiments, differentiable augmentations were implemented using the Kornia package Riba et al. [2020] , as well as the official code of Faster AutoAugment Hataya et al. [2020] .  ... 
arXiv:2202.02142v3 fatcat:mnkrxreyejgpfaho4fnweilidu

AutoDO: Robust AutoAugment for Biased Data with Label Noise via Scalable Probabilistic Implicit Differentiation [article]

Denis Gudovskiy, Luca Rigazio, Shun Ishizaka, Kazuki Kozuka, Sotaro Tsukizawa
2021 arXiv   pre-print
In particular, we include hyperparameters for augmentation, loss weights, and soft-labels that are jointly estimated using implicit differentiation.  ...  AutoAugment has sparked an interest in automated augmentation methods for deep learning models.  ...  The code is in PyTorch [27] with the Kornia library [30] for differentiable image operations.  ... 
arXiv:2103.05863v2 fatcat:dblc7apwxrd6tnjdjn6txclwfa

An overview of mixing augmentation methods and augmentation strategies [article]

Dominik Lewy, Jacek Mańdziuk
2022 arXiv   pre-print
An alternative approach is application of data augmentation (DA) techniques that aim at model regularization by creating additional observations from the available ones.  ...  First, the presented methods are briefly described, and then qualitatively compared with respect to their key characteristics.  ...  Fourthly, differentiable augmentation operations from Kornia library [59] are used.  ... 
arXiv:2107.09887v2 fatcat:isue7dmwxzdihgwiq2efj3k3qm

Soft-Median Choice: An Automatic Feature Smoothing Method for Sound Event Detection [article]

Fengnian Zhao, Ruwei Li, Xin Liu, Liwen Xu
2021 arXiv   pre-print
To resolve these problems, we provide different channels of features smoothed to different extents along with the original feature, so the model can optimize the weights while cognizing all the errors.  ...  Although the differentiable module of median blur with the backend of pytorch exists in kornia package[18], our study shows that a system with a differentiable median filter layer cannot converge well,  ...  With the help of data-augmentation, enough quantity of data is obtained for training [3] [8] [9] . This technique improves generalization performance.  ... 
arXiv:2011.12564v3 fatcat:gljqxe6q3ravnnmorxzlm22ldu

A Benchmark and Baseline for Language-Driven Image Editing [article]

Jing Shi, Ning Xu, Trung Bui, Franck Dernoncourt, Zheng Wen, Chenliang Xu
2020 arXiv   pre-print
Not only performing well on challenging user data, but such an approach is also highly interpretable.  ...  However, most similar work can only deal with a specific image domain or can only do global retouching.  ...  .: Kornia: an open source differentiable computer vision library for pytorch (2020) 21 29. Gonzales, R.C., Woods, R.E.: Digital image processing (2002) 21 30.  ... 
arXiv:2010.02330v1 fatcat:7rz36nz2mjfknbxuwivd2a2pii
« Previous Showing results 1 — 15 out of 37 results