Unsupervised Deep Feature Transfer for Low Resolution Image Classification

Wu, Yuanwei; Zhang, Ziming; Wang, Guanghui

doi:10.1109/iccvw.2019.00136

Cited by 24 publications

(11 citation statements)

References 42 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Considering the practical challenges with real-world data collection, the main problem is the low quality of images (Zhu et al, 2020). Therefore, research attention should focus on investigating various transfer learning models for identifying the best solution for leverage on the discriminative feature representation from high-resolution images to low-resolution images (Wu et al, 2019). Based on some of the highlighted challenges with plant disease classification, this study aims to improve the classification of plant diseases on low-quality images using data augmentation methods.…”

Section: Related Workmentioning

confidence: 99%

Cassava disease recognition from low‐quality images using enhanced data augmentation model and deep learning

et al. 2021

View full text Add to dashboard Cite

Improvement of deep learning algorithms in smart agriculture is important to support the early detection of plant diseases, thereby improving crop yields. Data acquisition for machine learning applications is an expensive task due to the requirements of expert knowledge and professional equipment. The usability of any application in a real-world setting is often limited by unskilled users and the limitations of devices used for acquiring images for classification. We aim to improve the accuracy of deep learning models on low-quality test images using data augmentation techniques for neural network training. We generate synthetic images with a modified colour value distribution to expand the trainable image colour space and to train the neural network to recognize important colour-based features, which are less sensitive to the deficiencies of low-quality images such as those affected by blurring or motion. This paper introduces a novel image colour histogram transformation technique for generating synthetic images for data augmentation in image classification tasks. The approach is based on the convolution of the Chebyshev orthogonal functions with the probability distribution functions of image colour histograms. To validate our proposed model, we used four methods (resolution down-sampling, Gaussian blurring, motion blur, and overexposure) for reducing image quality from the Cassava leaf disease dataset. The results based on the modified MobileNetV2 neural network showed a statistically significant improvement of cassava leaf disease recognition accuracy on lower-quality testing images when compared with the baseline network.The model can be easily deployed for recognizing and detecting cassava leaf diseases in lower quality images, which is a major factor in practical data acquisition.

show abstract

Section: Related Workmentioning

confidence: 99%

Cassava disease recognition from low‐quality images using enhanced data augmentation model and deep learning

et al. 2021

View full text Add to dashboard Cite

show abstract

“…Deep learning models: Inspired by the success of AlexNet [16] in the ImageNet Large Scale Visual Recognition Challenge (ILSVRC) 2012, convolutional neural networks (CNN) have attracted a lot of attention and been successfully applied to image classification [20][21][22], object detection [4,23,24], depth estimation [25,26], image transformation [27,28], and crowd counting [29]citesajid2020plug. VGGNets [14], and GoogleNet [17], the ILSVRC winners of 2014 and 2015, proved that deeper models could significantly increase the ability of representations.…”

Section: Related Workmentioning

confidence: 99%

A comparative study on polyp classification using convolutional neural networks

Patel

Tao

et al. 2020

PLoS ONE

Self Cite

View full text Add to dashboard Cite

Colorectal cancer is the third most common cancer diagnosed in both men and women in the United States. Most colorectal cancers start as a growth on the inner lining of the colon or rectum, called 'polyp'. Not all polyps are cancerous, but some can develop into cancer. Early detection and recognition of the type of polyps is critical to prevent cancer and change outcomes. However, visual classification of polyps is challenging due to varying illumination conditions of endoscopy, variant texture, appearance, and overlapping morphology between polyps. More importantly, evaluation of polyp patterns by gastroenterologists is subjective leading to a poor agreement among observers. Deep convolutional neural networks have proven very successful in object classification across various object categories. In this work, we compare the performance of the state-of-the-art general object classification models for polyp classification. We trained a total of six CNN models end-to-end using a dataset of 157 video sequences composed of two types of polyps: hyperplastic and adenomatous. Our results demonstrate that the state-of-the-art CNN models can successfully classify polyps with an accuracy comparable or better than reported among gastroenterologists. The results of this study can guide future research in polyp classification.

show abstract

“…Computer vision has progressed rapidly with deep learning techniques and more advanced and accurate models for object detection, image classification, image segmentation, pose estimation, and tracking emerging almost every day [31], [42], [49]. Even though computer vision enters a new era with deep learning, there are still plenty of problems unsolved and domain shift is one of them.…”

Section: Introductionmentioning

confidence: 99%

Six-channel Image Representation for Cross-domain Object Detection

Zhang,

Ma,

Wang

2021

Preprint

Self Cite

View full text Add to dashboard Cite

Most deep learning models are data-driven and the excellent performance is highly dependent on the abundant and diverse datasets. However, it is very hard to obtain and label the datasets of some specific scenes or applications. If we train the detector using the data from one domain, it cannot perform well on the data from another domain due to domain shift, which is one of the big challenges of most object detection models. To address this issue, some image-to-image translation techniques are employed to generate some fake data of some specific scenes to train the models. With the advent of Generative Adversarial Networks (GANs), we could realize unsupervised image-to-image translation in both directions from a source to a target domain and from the target to the source domain. In this study, we report a new approach to making use of the generated images. We propose to concatenate the original 3-channel images and their corresponding GAN-generated fake images to form 6-channel representations of the dataset, hoping to address the domain shift problem while exploiting the success of available detection models. The idea of augmented data representation may inspire further study on object detection and other applications.

show abstract

Unsupervised Deep Feature Transfer for Low Resolution Image Classification

Cited by 24 publications

References 42 publications

Cassava disease recognition from low‐quality images using enhanced data augmentation model and deep learning

Cassava disease recognition from low‐quality images using enhanced data augmentation model and deep learning

A comparative study on polyp classification using convolutional neural networks

Six-channel Image Representation for Cross-domain Object Detection

Contact Info

Product

Resources

About