Data Augmentation Using Random Image Cropping and Patching for Deep CNNs
Open accessThe DOI matches its Crossref record; 54 of the 54 references it cites were found in a registry or an index.
Abstract
Deep convolutional neural networks (CNNs) have achieved remarkable results in image processing tasks. However, their high expression ability risks overfitting. Consequently, data augmentation techniques have been proposed to prevent overfitting while enriching datasets. Recent CNN architectures with more parameters are rendering traditional data augmentation techniques insufficient. In this study, we propose a new data augmentation technique called random image cropping and patching (RICAP) which randomly crops four images and patches them to create a new training image. Moreover, RICAP mixes the class labels of the four images, resulting in an advantage of the soft labels. We evaluated RICAP with current state-of-the-art CNNs (e.g., the shake-shake regularization model) by comparison with competitive data augmentation techniques such as cutout and mixup. RICAP achieves a new state-of-the-art test error of 2.19% on CIFAR-10. We also confirmed that deep CNNs with RICAP achieve better results on classification tasks using CIFAR-100 and ImageNet, an image-caption retrieval task using Microsoft COCO, and other computer vision tasks.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“However, inserting a dropout between the convolutional layers (as opposed to within the residual block) made the learning more effective in WideResNet [121, 122].”
From Review of deep learning: concepts, CNN architectures, challenges, applications, future directions (2021) · cites this paper as [122]
Experimental — accuracy not yet measured
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
Deep Residual Learning for Image Recognition
He, K., Zhang, X., Ren, S., & Sun, J. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.90 · Cited by 235,474 · cited 5×
“The backbone model was the 50 layer ResNet [11] pre-trained on ImageNet.”
BackgroundMethodInfluentialVery Deep Convolutional Networks for Large-Scale Image Recognition
Simonyan, K., & Zisserman, A. (2014) · arXiv · doi:10.48550/arxiv.1409.1556 · Cited by 114,148
ImageNet classification with deep convolutional neural networks
Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2017) · Communications of the ACM · doi:10.1145/3065386 · Cited by 112,155 · cited 4×
“AlexNet also performed principal component analysis (PCA) on a set of RGB values to alter the intensities of the RGB channels for evaluation on the ImageNet dataset [24].”
MethodInfluentialAdam: A Method for Stochastic Optimization
Kingma, D. P., & Ba, J. (2014) · arXiv · doi:10.48550/arxiv.1412.6980 · Cited by 98,946 · cited 2×
“The weight parameters were updated using the Adam optimizer [42] over 100 epochs with a mini-batch size of 16.”
MethodYou Only Look Once: Unified, Real-Time Object Detection
Redmon, J., Divvala, S., Girshick, R., & Farhadi, A. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.91 · Cited by 53,803 · cited 1×
“YOLO [46] is an end-to-end model achieving faster inference than previous methods.”
MethodDensely Connected Convolutional Networks
Huang, G., Liu, Z., Van Der Maaten, L., & Weinberger, K. Q. (2017) · 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2017.243 · Cited by 48,566
Going deeper with convolutions
Szegedy, C., Liu, W., Jia, Y., Sermanet, P., Reed, S., Anguelov, D., Erhan, D., Vanhoucke, V., & Rabinovich, A. (2015) · 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2015.7298594 · Cited by 48,562
Microsoft COCO: Common Objects in Context
Lin, T.-Y., Maire, M., Belongie, S., Hays, J., Perona, P., Ramanan, D., Dollár, P., & Zitnick, C. L. (2014) · Lecture Notes in Computer Science · doi:10.1007/978-3-319-10602-1_48 · Cited by 45,460
ImageNet Large Scale Visual Recognition Challenge
Russakovsky, O., Deng, J., Su, H., Krause, J., Satheesh, S., Ma, S., Huang, Z., Karpathy, A., Khosla, A., Bernstein, M., et al. (2015) · International Journal of Computer Vision · doi:10.1007/s11263-015-0816-y · Cited by 39,380 · cited 16×
“We apply RICAP to existing deep CNNs and evaluate them on the classification tasks using the CIFAR-10, CIFAR-100 [8], and ImageNet [24] datasets in Sections IV-A, IV-B and IV-C.”
MethodInfluentialRethinking the Inception Architecture for Computer Vision
Szegedy, C., Vanhoucke, V., Ioffe, S., Shlens, J., & Wojna, Z. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.308 · Cited by 32,138
Cited by
Items 101–200 of 476
Deep Learning Innovations in Video Classification: A Survey on Techniques and Dataset Evaluations
Mao, M., Lee, A., & Hong, M. (2024) · Electronics · doi:10.3390/electronics13142732 · Cited by 27
Age Range Estimation Using MTCNN and VGG-Face Model
Gyawali, D., Pokharel, P., Chauhan, A., & Shakya, S. C. (2020) · International Conference on Computing, Communication and Networking Technologies (ICCCNT) · doi:10.1109/icccnt49239.2020.9225443 · Cited by 23
“images are extracted of size 224x224 pixels using random crop[18].”
MethodAdversarial Learning-Based Sentiment Analysis for Socially Implemented IoMT Systems
Xu, Y., Gao, H., Li, R., Jiang, Y., Mumtaz, S., Jiang, Z., Fang, J., & Dong, L. (2023) · IEEE Transactions on Computational Social Systems · doi:10.1109/tcss.2023.3252028 · Cited by 19
“Some researchers add random noise to the input or hidden layers of the model to prevent overfitting of the model [40], [41].”
BackgroundHigh Precision Cervical Precancerous Lesion Classification Method Based on ConvNeXt
Tang, J., Zhang, T., Gong, Z., & Huang, X. (2023) · Bioengineering · doi:10.3390/bioengineering10121424 · Cited by 22
“However, for a model, such as a simple and effective data augmentation algorithm, there are a series of works of related research [61,62].”
BackgroundOrtho-Shot: Low Displacement Rank Regularization with Data Augmentation for Few-Shot Learning
Osahor, U., & Nasrabadi, N. M. (2022) · IEEE Winter Conference on Applications of Computer Vision · doi:10.1109/wacv51458.2022.00210 · Cited by 29
Rapid detection of soybean nutrient deficiencies with YOLOv8s for precision agriculture advancement
Jeong, M., Park, S., Kwon, S.-M., Lim, K., Jung, D.-R., Lee, H., Kim, H. J., & Shin, J.-H. (2025) · Scientific Reports · doi:10.1038/s41598-024-83295-6 · Cited by 23
“Mosaic augmentation was fully used with a factor of 1.0, whereas MixUp and Copy-Paste augmentations were not used in this configuration 30 .”
InfluentialHameed, K., Chai, D., & Rassau, A. (2021) · Information Processing in Agriculture · doi:10.1016/j.inpa.2021.09.003 · Cited by 24
Detection of moving objects in multi-complex environments using selective attention networks (SANet)
Cho, J., & Kim, K. (2023) · Automation in Construction · doi:10.1016/j.autcon.2023.105066 · Cited by 21
Kishore, P., Mondal, A., Trivedi, A., Singh, P. K., & Alankar, A. (2024) · International Journal of Fatigue · doi:10.1016/j.ijfatigue.2024.108724 · Cited by 20
Self-Supervised Scene-Debiasing for Video Representation Learning via Background Patching
Assefa, M., Jiang, W., Gedamu, K., Yilma, G., Kumeda, B., & Ayalew, M. (2022) · IEEE Transactions on Multimedia · doi:10.1109/tmm.2022.3193559 · Cited by 17
Ahmad, F., Khan, M. U. G., Tahir, A., & Masud, F. (2023) · BMC Bioinformatics · doi:10.1186/s12859-023-05398-7 · Cited by 18
“Researches have elaborated that image patching help increase dataset size [32] and preserve essential local image details [30].”
BackgroundLee, S.-J., Oh, H. J., Son, Y., Kim, J., Kwon, I.-J., Kim, B., Lee, J.-H., & Kim, H.-K. (2024) · BMC Oral Health · doi:10.1186/s12903-024-03898-3 · Cited by 27
“This technique was motivated by two existing methods: RICAP [28] and mosaic data augmentation [29] .”
MethodDai, Q., Xiao, Y., Lv, S., Song, S., Xue, X., Liang, S., Huang, Y., & Li, Z. (2024) · Agriculture · doi:10.3390/agriculture14111964 · Cited by 22
Perlin Random Erasing for Data Augmentation
Saran, M., Nar, F., & Saran, A. N. (2021) · Signal Processing and Communications Applications Conference (SIU) · doi:10.1109/siu53274.2021.9477804 · Cited by 20
Adversarial and Isotropic Gradient Augmentation for Image Retrieval With Text Feedback
Huang, F., Zhang, L., Zhou, Y., & Gao, X. (2022) · IEEE Transactions on Multimedia · doi:10.1109/tmm.2022.3222624 · Cited by 17
“Typical data augmentation operations [37], [38], [39] include flipping, rotation, scaling, cropping, adding noise, and so on.”
BackgroundPrompt-Based Modality Alignment for Effective Multi-Modal Object Re-Identification
Zhang, S., Luo, W., Cheng, D., Xing, Y., Liang, G., Wang, P., & Zhang, Y. (2025) · IEEE Transactions on Image Processing · doi:10.1109/tip.2025.3556531 · Cited by 32
“Each image is padded with 10 pixels and augmented with random cropping [64], horizontal flipping, and erasing [65].”
Pathological Myopia Image Recognition Strategy Based on Data Augmentation and Model Fusion
Cui, J., Zhang, X., Xiong, F., & Chen, C. (2021) · Journal of Healthcare Engineering · doi:10.1155/2021/5549779 · Cited by 17
“presented a detailed analysis of image enhancement techniques based on multifusion, thereby giving an insight into the algorithm used in each method, along with its implementation framework [36, 42].”
MethodRoslan, N. A. M., Diah, N. M., Ibrahim, Z., Munarko, Y., & Minarno, A. E. (2023) · International Journal of Advances in Intelligent Informatics · doi:10.26555/ijain.v9i1.1076 · Cited by 28
“Random image cropping and patching (RICAP) [23] is a data augmentation technique that randomly crops four photos and patches them together to create a new image, as well as mixing the class labels of the four images to take use of the soft labels.”
BackgroundNighttime Thermal Infrared Image Colorization With Feedback-Based Object Appearance Learning
Luo, F., Liu, S., Cao, Y., Yang, K.-F., Xie, C.-Y., Liu, Y., & Li, Y. (2023) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2023.3331499 · Cited by 16
“RICAP [41] stitched four images into one image after cropping out small pieces at a certain ratio.”
BackgroundArtificial and beneficial – Exploiting artificial images for aerial vehicle detection
Weber, I., Bongartz, J., & Roscher, R. (2021) · ISPRS Journal of Photogrammetry and Remote Sensing · doi:10.1016/j.isprsjprs.2021.02.015 · Cited by 17
“Several more complex augmentation methods [16, 27, 28, 29], of whom especially mixup [30] and Random Image Cropping and Patching (RICAP, [16]) are suitable for object detection, highlighted that well-designed augmentations have the potential to improve performance significantly, particularly in the…”
BackgroundPorosity prediction from pre-stack seismic data via a data-driven approach
Yang, N., Li, G., Zhao, P., Zhang, J., & Zhao, D. (2023) · Journal of Applied Geophysics · doi:10.1016/j.jappgeo.2023.104947 · Cited by 16
SAR target augmentation and recognition via cross-domain reconstruction
Dong, G., & Song, Y. (2024) · Pattern Recognition · doi:10.1016/j.patcog.2024.111117 · Cited by 15
Tackling the Background Bias in Sparse Object Detection via Cropped Windows
Varga, L. A., & Zell, A. (2021) · IEEE International Conference on Computer Vision Workshops · doi:10.1109/iccvw54120.2021.00311 · Cited by 19
“Data augmentation techniques such as Random Cropping, Random Image Cropping And Patching (RICAP) [19], or Cut Mix [29] commonly use a similar approach.”
Benchmarking PathCLIP for Pathology Image Analysis
Zheng, S., Cui, X., Sun, Y., Li, J., Li, H., Zhang, Y., Chen, P., Jing, X., Ye, Z., & Yang, L. (2024) · Journal of Imaging Informatics in Medicine · doi:10.1007/s10278-024-01128-4 · Cited by 26
GA3N: Generative adversarial AutoAugment network
Chinbat, V., & Bae, S. (2022) · Pattern Recognition · doi:10.1016/j.patcog.2022.108637 · Cited by 15
Mixed Stage Partial Network and Background Data Augmentation for Surveillance Object Detection
Chen, P.-Y., Hsieh, J.-W., Gochoo, M., & Chen, Y. (2022) · IEEE Transactions on Intelligent Transportation Systems · doi:10.1109/tits.2022.3206709 · Cited by 19
“However, to prevent the model from overfitting, different data augmentation methods, such as random crop [62], AutoAug-ment [63], and RandAugment [64] have played a major role in achieving SoTA results in image classification [65].”
BackgroundMethodYang, S. W., Park, H. J., Kim, J. S., Choi, W., Park, J., & Han, S. W. (2023) · Waste Management · doi:10.1016/j.wasman.2023.04.044 · Cited by 16
Background Instance-Based Copy-Paste Data Augmentation for Object Detection
Zhang, L., Xing, Z., & Wang, X. (2023) · Electronics · doi:10.3390/electronics12183781 · Cited by 14
Deep learning for Ethiopian indigenous medicinal plant species identification and classification
Kiflie, M. A., Sharma, D. P., & Haile, M. A. (2024) · Journal of Ayurveda and Integrative Medicine · doi:10.1016/j.jaim.2024.100987 · Cited by 19
“This technique is commonly used to achieve different goals, including enhancing the image ’ s composition, bringing focus to a specific subject, or eliminating any distracting elements in the background [28].”
Video Recognition of American Sign Language Using Two-Stream Convolution Neural Networks
Nugraha, F., & Djamal, E. C. (2019) · doi:10.1109/iceei47359.2019.8988872 · Cited by 15
“The small sets of data to prevent overfitting and make a more robust model use random cropping [11] and jittering [12].”
BackgroundPolarization imaging feature characterization of different endometrium phases by machine learning
Yao, Y., Zuo, M., Dong, Y., Shi, L., Zhu, Y., Si, L., Ye, X., & Ma, H. (2021) · OSA Continuum · doi:10.1364/osac.414109 · Cited by 17
A Thresholded Gabor-CNN Based Writer Identification System for Indic Scripts
Mridha, M. F., Ohi, A. Q., Shin, J., Kabir, M. M., Monowar, M. M., & Hamid, M. A. (2021) · IEEE Access · doi:10.1109/access.2021.3114799 · Cited by 15
“CNN architecture [59].”
Diffeomorphic transforms for data augmentation of highly variable shape and texture objects
Vállez, N., Bueno, G., Déniz, Ó., & Blanco, S. (2022) · Computer Methods and Programs in Biomedicine · doi:10.1016/j.cmpb.2022.106775 · Cited by 19
YOLOv7-GCA: A Lightweight and High-Performance Model for Pepper Disease Detection
Yue, X., Li, H., Song, Q., Zeng, F., Zheng, J., Ding, Z., Kang, G., Cai, Y., Lin, Y., Xu, X., et al. (2024) · Agronomy · doi:10.3390/agronomy14030618 · Cited by 15
Gu, Z., Krishna, K. V. M., Parsazadeh, M., Sharma, S., Manjunath, A., Tran, H., Fu, S., & Dahotre, N. B. (2024) · Progress in Additive Manufacturing · doi:10.1007/s40964-024-00603-2 · Cited by 18
“The annotated masks are prepared for the training input by assigning unique class numbers (background:0, MP:1, porosity: 2, mounting material: 3 and solidified metal: 4) [59].”
Combined Oriented Data Augmentation Method for Brain MRI Images
Farhan, A. S., Khalid, M., & Manzoor, U. (2025) · IEEE Access · doi:10.1109/access.2025.3526684 · Cited by 17
“The goal is to improve the generalization and robustness of deep CNN models by diversifying the training data [46].”
Li, W., Xiao, L., Yao, S., Hou, C., Wen, Z., & Ren, D. (2025) · Multimedia Systems · doi:10.1007/s00530-025-01783-9 · Cited by 15
“Techniques such as random cropping [11], scale transformations [12], and feature resampling [13] enhance a model’s detection accuracy and generalization for small objects.”
Köhler, A., Myklebust, E. B., & Maeland, S. (2022) · Geophysical Journal International · doi:10.1093/gji/ggac117 · Cited by 13
Lightweight Object Detection Model with Data Augmentation for Tiny Pest Detection
Yuan, Z., Li, S., Yang, P., & Li, Y. (2022) · IEEE International Conference on Industrial Informatics (INDIN) · doi:10.1109/indin51773.2022.9976137 · Cited by 15
“In addition to those intuitive methods, mixing images by random cropping and patching [23] is an effective method to reduce the error rate, especially in the situation of a low number of training samples.”
MethodLiu, B., Li, L., Xiao, Q., Ni, W., & Yang, Z. Z. (2022) · IEEE Transactions on Geoscience and Remote Sensing · doi:10.1109/tgrs.2022.3211517 · Cited by 17
“To improve the integrity of the comparative experiment, we selected seven basic data manipulations: cutmix [54], cutout [23], dual cutout, random crop, random erasing [55], random horizontal flip, and anywhere-RICAP [56].”
MethodPham, L. D., Ngo, D. T., Salovic, D., Jalali, A., Schindler, A., Nguyen, P. X., Tran, K., & Vu, H. C. (2023) · Applied Acoustics · doi:10.1016/j.apacoust.2023.109489 · Cited by 21
Hierarchical Attention-Based Astronaut Gesture Recognition: A Dataset and CNN Model
Lingyun, G., Lin, Z., & Zhaokui, W. (2020) · IEEE Access · doi:10.1109/access.2020.2986473 · Cited by 13
“A data augmentation approach named Random image cropping and patching (RICAP) [43] randomly crops four images and patches them to synthesize a new image, and mixes the class labels of the four images to get an advantage of the soft labels.”
MethodA review of image processing and analysis of computed tomography images using deep learning methods
Anderson, D., Ramachandran, P., Trapp, J., & Fielding, A. L. (2025) · Physical and Engineering Sciences in Medicine · doi:10.1007/s13246-025-01635-w · Cited by 16
“Several methods for data augmentation also utilise smaller regions of the input image as “patches”, which can be combined together to create composite images using techniques like random image cropping and patching (RICAP) [ 209].”
Efficient Dataset Collection for Concrete Crack Detection With Spatial-Adaptive Data Augmentation
Kim, J., & Lee, J. (2023) · IEEE Access · doi:10.1109/access.2023.3328243 · Cited by 12
“Takahashi et al. proposed the RICAP (Random image cropping patching) technique, which combines randomly cropped patches from four images to create a single image [50].”
MethodQuery-Aware Cross-Mixup and Cross-Reconstruction for Few-Shot Fine-Grained Image Classification
Zhang, Z., Chang, D., Zhu, R., Li, X., Ma, Z., & Xue, J. (2024) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2024.3484530 · Cited by 19
“Data augmentation aims to increase sample information to assist feature learning and improve the generalization ability of the model [20], [27]–[29].”
CrossViT with ECAP: Enhanced deep learning for jaw lesion classification
Panyarak, W., Suttapak, W., Mahasantipiya, P. M., Charuakkra, A., Boonsong, N., Wantanajittikul, K., & Iamaroon, A. (2024) · International Journal of Medical Informatics · doi:10.1016/j.ijmedinf.2024.105666 · Cited by 13
Emotion Recognition of Autistic Children from Facial Images Using Hybrid Model
Afrin, M., Hoque, K. E., & Chaiti, R. D. (2024) · International Conference on Computing, Communication and Networking Technologies (ICCCNT) · doi:10.1109/icccnt61001.2024.10724499 · Cited by 16
“Random Crop: Random crop layer cut the input images during training phrase [15].”
BackgroundDo Affective Cues Validate Behavioural Metrics for Search?
McDuff, D., Thomas, P., Craswell, N., Rowan, K., & Czerwinski, M. P. (2021) · International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR) · doi:10.1145/3404835.3462894 · Cited by 13
Abdullahi, H. S., & Sheriff, R. E. (2022) · Elsevier eBooks · doi:10.1016/b978-0-323-85214-2.00013-6 · Cited by 14
Data Augmentation Using Mixup and Random Erasing
Dai, X.-P., Zhao, X., Cen, F., & Zhu, F.-L. (2022) · 2022 IEEE International Conference on Networking, Sensing and Control (ICNSC) · doi:10.1109/icnsc55942.2022.10004083 · Cited by 11
“Image Cropping and Patching [10] is inspired by the above methods and has achieved good results on the general CNN structure.”
MethodMCMC Guided CNN Training and Segmentation for Pancreas Extraction
Tian, M., He, J., Yu, X., Cai, C., & Gao, Y. (2021) · IEEE Access · doi:10.1109/access.2021.3070391 · Cited by 14
Harnessing deep learning for plant species classification: A comprehensive review
Zulfiqar, A., Ebroul Izquierdo, M., & Chandramouli, K. (2025) · Computers and Electronics in Agriculture · doi:10.1016/j.compag.2025.110663 · Cited by 13
Silveira, A., Greving, I., Longo, E., Scheel, M., Weitkamp, T., Fleck, C., Shahar, R., & Zaslansky, P. (2024) · Journal of Synchrotron Radiation · doi:10.1107/s1600577523009852 · Cited by 13
NBBOX: Noisy Bounding Box Improves Remote Sensing Object Detection
Kim, Y., Kim, S., & Jeon, M. (2025) · IEEE Geoscience and Remote Sensing Letters · doi:10.1109/lgrs.2025.3527712 · Cited by 17
“The image erasing like in [5] and [6] randomly selects a subregion in an image and deletes its contents, while the image mix like in [7] and [8] combines two or more images (or patches) as a single image during training.”
A Good Data Augmentation Policy is not All You Need: A Multi-Task Learning Perspective
Zhang, L., & Ma, K. (2023) · IEEE transactions on circuits and systems for video technology (Print) · doi:10.1109/tcsvt.2022.3219339 · Cited by 13
“Moreover, Mixup-based methods have been proposed to synthesize new images as the linear interpolation of existing images [17], [53], [63] [51].”
MethodPark, P., Jeong, H., & Kim, J. (2020) · IEEE/ACM International Symposium on Microarchitecture (MICRO) · doi:10.1109/micro50266.2020.00072 · Cited by 15
“[43] proposed an efficient cropping algorithm that randomly crops four images and merges them to create a new training image.”
BackgroundLutnick, B. R., Murali, L. K., Ginley, B. G., Rosenberg, A. Z., & Sarder, P. (2022) · Journal of Pathology Informatics · doi:10.4103/jpi.jpi_59_20 · Cited by 13
“This improves upon the popular data augmentation strategy of randomly cropping training images, which has been shown to improve training.[22] As a result of the elimination of a preselected dataset, the notion of the training epoch (one training loop through all the data) is no longer valid, the number of training ste…”
BackgroundVirtual computed-tomography system for deep-learning-based material decomposition
Fujiwara, D., Shimomura, T., Zhao, W., Li, K., Haga, A., & Geng, L. (2022) · Physics in Medicine and Biology · doi:10.1088/1361-6560/ac7bcd · Cited by 11
“In this study, the training data of BT phantomswere augmented by horizontal and vertical flipping, rotationwith amaximumangle of20°, and random image cropping and patching (RICAP) (Takahashi et al 2019).”
MethodKiflie, M. A., Sharma, D. P., Haile, M. A., & Srinivasagan, R. (2024) · Computers · doi:10.3390/computers13020038 · Cited by 13
“Image augmentation involves diversifying a dataset by applying various adjustments to the original images, including changes in brightness, contrast, rotation, flipping, or zooming [41,42].”
BackgroundAn Overview on Data Augmentation for Machine Learning
Volkova, S. (2024) · Lecture notes in networks and systems · doi:10.1007/978-3-031-55349-3_12 · Cited by 19
Evidence-Based Real-Time Road Segmentation With RGB-D Data Augmentation
Xue, F., Chang, Y., Xu, W., Liang, W., Sheng, F., & Ming, A. (2025) · IEEE Transactions on Intelligent Transportation Systems · doi:10.1109/tits.2024.3509140 · Cited by 12
“Takahashi et al. [27] designed a combined strategy, randomly cropping images and concatenating the cropped parts to form a new image.”
Leung, E. T. Y., Mei, X., Lee, B. K. M., Lam, K. K. W., Lee, C., Li, R., Ng, E. H. Y., Yeung, W. S. B., Yu, L., & Chiu, P. C. N. (2025) · Human Reproduction Open · doi:10.1093/hropen/hoaf024 · Cited by 11
“This can be achieved by exposing the model to a wide range of variations across clinical datasets, incorporating data augmentation (Takahashi et al. , 2020), and employing regularization techniques such as dropout, weight decay, and early stopping based on validation performance (Lee and Seok,…”
Zheng, X., Huang, Y., Xie, Z., & Zheng, A. (2025) · npj Heritage Science · doi:10.1038/s40494-025-01907-4 · Cited by 14
“Random horizontal fl ipping and a maximum 10-degree rotation are applied for data augmentation 25,26 .”
Quaternary synapses network for memristor-based spiking convolutional neural networks
Sun, S., Li, J., Li, Z., Liu, H., Liu, H., & Li, Q. (2019) · IEICE Electronics Express · doi:10.1587/elex.16.20190004 · Cited by 10
“To re fl ect the practical application of algorithm performance better, we do not use any data augmentation technologies [25, 26, 27].”
MethodHepatic Vein and Arterial Vessel Segmentation in Liver Tumor Patients
Kuang, H., Yang, Z., Zhang, X., Tan, J.-P., Wang, X., & Zhang, L. (2022) · Computational Intelligence and Neuroscience · doi:10.1155/2022/2303733 · Cited by 10
“ere are many ways of data augmentation, such as rotation, translation, scaling, and cropping [41, 42].”
BackgroundFeature combination mixup: novel mixup method using feature combination for neural networks
Takase, T. (2023) · Neural computing & applications (Print) · doi:10.1007/s00521-023-08421-3 · Cited by 12
“The random image cropping and patching [18] method randomly crops four images and then patches them to generate a new image.”
MethodDifferential augmentation data for vehicle classification using convolutional neural network
Prasetiyo, B., Alamsyah, A., Hakim, M. F. A., Jumanto, J., & Adi, M. H. (2023) · AIP conference proceedings · doi:10.1063/5.0126720 · Cited by 13
Measuring Visual Generalization in Continuous Control from Pixels
Grigsby, J., & Qi, Y. (2020) · arXiv · doi:10.48550/arxiv.2010.06740 · Cited by 10
MSE Loss with Outlying Label for Imbalanced ClassificationPreprint
Kato, S., & Hotta, K. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2107.02393 · Cited by 31
“The method for generating new data from existing data is called oversampling[23, 24, 28, 32].”
BackgroundMethodDRFnet: Dynamic receptive field network for object detection and image recognition
Tan, M.-J., Yuan, X., Liang, B., & Han, S. (2023) · Frontiers in Neurorobotics · doi:10.3389/fnbot.2022.1100697 · Cited by 10
“0005 to train our network, and the learning rate decay strategy uses a cosine decay scheduler (Takahashi et al., 2019), with a warmup of two epochs in the beginning (Goyal et al.”
MethodData augmentation and data mining towards microstructure and property relationship for composites
Guo, Z., Liu, X., Pan, Z., Zhou, Y., Zhong, Z., & Yan, Z. (2023) · Engineering Computations · doi:10.1108/ec-10-2022-0639 · Cited by 10
Zamani, S. A., & Baleghi, Y. (2025) · Computers and Electronics in Agriculture · doi:10.1016/j.compag.2025.110151 · Cited by 11
An Improved Semantic Segmentation Method for Remote Sensing Images Based on Neural Network
Jiang, N., & Li, J. (2020) · Traitement du signal · doi:10.18280/ts.370213 · Cited by 9
“Many typical CNNs have been successfully designed to semantically segment multiple objects, namely, fully-connected network (FCN) [1], SegNet [2], pyramid scene parsing network (PSPNet) [3], DeepLab [4], and deconvolution network.”
BackgroundInfluentialS 4 : Self-supervised learning with sparse-dense sampling
Tian, Y., Zhang, W., Su, P., Xu, Y., Zhuang, P., Xie, X., & Zhao, W. (2024) · Knowledge-Based Systems · doi:10.1016/j.knosys.2024.112040 · Cited by 11
Assessing the reliability of point mutation as data augmentation for deep learning with genomic data
Lee, H., Özbulak, U., Park, H.-M., Depuydt, S., Neve, W. D., & Vankerschaver, J. (2024) · BMC Bioinformatics · doi:10.1186/s12859-024-05787-6 · Cited by 11
Su, T., Wu, J. C., Chiu, W.-C., Chen, T., Lo, W., & Lu, H. H. (2024) · Journal of Dental Sciences · doi:10.1016/j.jds.2024.06.001 · Cited by 8
Synthetic data for visual machine learning : A data-centric approach
Tsirikoglou, A. (2022) · Linköping studies in science and technology. Dissertations · doi:10.3384/9789179291754 · Cited by 10
Saraswathi, C., & Pushpa, B. (2023) · Engineering Technology & Applied Science Research · doi:10.48084/etasr.5852 · Cited by 8
“In [17], a novel data augmentation method was proposed, called Random Image Cropping and Patching (RICAP), which randomly cropped and patched an input image to generate a new training image for CNN.”
MethodInfluentialTong, Y., Luo, X., Ma, L., Xie, S., Yang, W., & Guo, Y. (2024) · Pattern Analysis and Applications · doi:10.1007/s10044-024-01258-z · Cited by 8
“Therefore, data augmentation methods [13, 29, 49, 57] are proposed to improve the accuracy of deep learning models when the quantity and diversity of training images are insufficient.”
Automated Grading System to Evaluate Ripeness of Tomatoes Using Deep Learning Methods
Malhotra, S., & Chhikara, R. R. (2021) · Lecture notes in networks and systems · doi:10.1007/978-981-15-9689-6_15 · Cited by 8
Prior-guided GAN-based interactive airplane engine damage image augmentation method
Huang, R., Duan, B., Zhang, Y., & Fan, W. (2021) · Chinese Journal of Aeronautics · doi:10.1016/j.cja.2021.11.021 · Cited by 8
Self-Supervised Multi-Label Transformation Prediction for Video Representation Learning
Assefa, M., Jiang, W., Yilma, G., Kumeda, B., Ayalew, M., & Seid, A. M. (2022) · Journal of Circuits Systems and Computers · doi:10.1142/s0218126622501596 · Cited by 8
A2-Aug: Adaptive Automated Data Augmentation
Li, L., & Li, A.-G. (2023) · 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw59228.2023.00221 · Cited by 14
“Data augmentation [36, 2, 3], such as cropping , flipping, and color jittering, is an effective training technique, which is widely used for many computer vision tasks ( e.”
MethodMolin, A. E., Blekanov, I. S., Mitrofanov, E. P., & Mitrofanova, O. (2024) · Vestnik of Saint Petersburg University Applied Mathematics Computer Science Control Processes · doi:10.21638/11701/spbu10.2024.103 · Cited by 11
Real-Time Barge Detection Using Traffic Cameras and Deep Learning on Inland Waterways
Agorku, G. E., Hernandez, S. V., Falquez, M., Poddar, S., & Amankwah-Nkyi, K. (2024) · Transportation Research Record Journal of the Transportation Research Board · doi:10.1177/03611981241263574 · Cited by 9
“Gaussian blur simulated adverse weather conditions, enabling the model to better handle degraded image quality.”
Progress and Thinking on Self-Supervised Learning Methods in Computer Vision: A Review
Chen, Z., Hu, B., Chen, Z., & Zhang, J. (2024) · IEEE Sensors Journal · doi:10.1109/jsen.2024.3443885 · Cited by 17
Latent Code Augmentation Based on Stable Diffusion for Data-Free Substitute Attacks
Shao, M., Meng, L., Qiao, Y., Zhang, L., & Zuo, W. (2025) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2025.3526338 · Cited by 14
“Then, we define the set of multicode augmentation operations 1 , ψ 2 , . . . , ψ l } that include MixUp [47], CutMix [48], RICAP [49], and so on.”
Enhancing CNNs via structural intervention with XGBoost
Yavas, C. E., Chen, L., Kadlec, C., & Kim, J. (2025) · Engineering Research Express · doi:10.1088/2631-8695/add084 · Cited by 7
RDFCNet: RGB-guided depth feature calibration network for RGB-D salient object detection
Wei, W., Shi, Y., Wang, Y., & Liu, S. (2025) · Neurocomputing · doi:10.1016/j.neucom.2025.131127 · Cited by 7
ROFI: a deep learning-based ophthalmic sign-preserving and reversible patient face anonymizer
Tian, Y., Zhou, M., Chen, Y., Li, F., Qi, L., Wang, S., Xu, X., Yu, Y., Xu, S., Lei, C., et al. (2025) · npj Digital Medicine · doi:10.1038/s41746-025-02062-7 · Cited by 13
Intelligent Lithological Mapping: Challenges and Future Prospective
Wang, Z., & Zuo, R. (2025) · Natural Resources Research · doi:10.1007/s11053-025-10563-1 · Cited by 16
Hydrogen-assisted cracking: A deep learning approach for fractographic analysis
Campari, A., Konert, F., Razavi, N., Sobol, O., & Alvaro, A. (2026) · Computational materials science · doi:10.1016/j.commatsci.2025.114366 · Cited by 7
Zhao, H., Wang, J., Chen, Z., Lin, S., & Bao, P. (2022) · Neural Processing Letters · doi:10.1007/s11063-022-11022-1 · Cited by 7
“The former [20, 40–43] mixed a pair image and their label with a certain proportion, while the latter [19–22 , 44 In this work, our proposed data augmentation is built on the region-level removal strategies.”
BackgroundDetection of infill development and contributing factors using deep learning and multilevel modeling
Jun, H.-J., Kim, D., Kim, J., & Heo, J. (2024) · Cities · doi:10.1016/j.cities.2024.105019 · Cited by 7
Chromosomal Mutation-Inspired Radio Augmentation for Enhanced Automatic Modulation Classification
Pu, X., Luo, C., Yin, Y., Liu, Z., & Luo, Y. (2024) · IEEE Internet of Things Journal · doi:10.1109/jiot.2024.3457908 · Cited by 8
Tran, K., Pham, L. D., Nguyen, V.-R., & Nguyen, H.-S.-H. (2025) · Signal Image and Video Processing · doi:10.1007/s11760-025-03848-8 · Cited by 11
Valencia-Rosado, L. O., Guzman-Zavaleta, Z. J., & Starostenko, O. (2020) · IEEE Access · doi:10.1109/access.2020.3048083 · Cited by 8
“Additionally, data augmentation was performed to avoid over fitting, specifically random cropping and flipping, as stated in [45].”
MethodDistribution-preserving data augmentation
Saran, N., Saran, M., & Nar, F. (2021) · PeerJ Computer Science · doi:10.7717/peerj-cs.571 · Cited by 9
Sun, Z., Shi, Z., Xin, Y., Zhao, S., Jiang, H., Wang, D., Zhang, L., Wang, Z., Dai, Y., & Jiang, H. (2021) · Frontiers in Bioengineering and Biotechnology · doi:10.3389/fbioe.2021.761548 · Cited by 7
“Random image cropping and patching (RICAP) were employed for data augmentation for deep convolutional neural network training (Takahashi et al., 2018).”
MethodImage Classification Based On improved VGG Network
Tang, L. (2021) · IEEE International Conference on Signal and Image Processing (ICSIP) · doi:10.1109/icsip52628.2021.9688778 · Cited by 7
“In the process of image classification, some objects in the pictures are not fully displayed, and there may be occlusions, which means that the pictures may only show a part of the target object [6].”
Background
Metadata
- Authors
- Takahashi, R., Matsubara, T., & Uehara, K.
- Affiliations
- Kobe University (JP)
- Venue
- IEEE Transactions on Circuits and Systems for Video Technology
- Published
- 2020-09
- Volume
- 30
- Issue
- 9
- Pages
- 2917-2931
- ISSN
- 1051-8215 (ISSN-L)
- 1558-2205
- OpenAlex
- W2900595477
- Cited by
- 477
- References
- 54
- Citations by year
- 2026: 30 · 2025: 68 · 2024: 73 · 2023: 89 · 2022: 100 · 2021: 72 · 2020: 34
- FWCI
- 24.17
- Type
- journal article
- Language
- en
- Open access
- green
- arxiv.org
- Licence
- Locations
- IEEE Transactions on Circuits and Systems for Video Technology (publishedVersion)
- arXiv (Cornell University) (acceptedVersion)
- Institutional Repositories DataBase (IRDB) (acceptedVersion)
- Kobe University Repository Kernel (Kobe University) (submittedVersion, other-oa)
- arXiv (Cornell University)
- Funders
- Strategic Information and Communications R&D Promotion Programme of Ministry of Internal Affairs and Communications, Japan (172107101)
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
- Physical Sciences › Computer Science › Artificial Intelligence › Domain Adaptation and Few-Shot Learning
- Physical Sciences › Computer Science › Artificial Intelligence › Machine Learning and Data Classification
- Keywords
- data augmentation, overfitting, convolutional neural networks, soft labels, cutouts, mixup, image classification
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.