Impact of Low Resolution on Image Recognition with Deep Neural Networks: An Experimental Study

Abstract

Due to the advances made in recent years, methods based on deep neural networks have been able to achieve a state-of-the-art performance in various computer vision problems. In some tasks, such as image recognition, neural-based approaches have even been able to surpass human performance. However, the benchmarks on which neural networks achieve these impressive results usually consist of fairly high quality data. On the other hand, in practical applications we are often faced with images of low quality, affected by factors such as low resolution, presence of noise or a small dynamic range. It is unclear how resilient deep neural networks are to the presence of such factors. In this paper we experimentally evaluate the impact of low resolution on the classification accuracy of several notable neural architectures of recent years. Furthermore, we examine the possibility of improving neural networks’ performance in the task of low resolution image recognition by applying super-resolution prior to classification. The results of our experiments indicate that contemporary neural architectures remain significantly affected by low image resolution. By applying super-resolution prior to classification we were able to alleviate this issue to a large extent as long as the resolution of the images did not decrease too severely. However, in the case of very low resolution images the classification accuracy remained considerably affected.

A Deep Learning Semantic Segmentation-Based Approach for Field-Level Sorghum Panicle Counting

“…In Koziarski and Cyganek [72], the robustness of deep learning models was evaluated against distortions such as blurring and concluded that the models were susceptible to such distortions. A similar study [73] evaluated the impact of low resolution on classification accuracy of several state of art models including VGGNet [23], ResNet [24] and AlexNet [74], and also confirmed the negative impact of lower resolution on deep learning models. From the foregoing, we would recommend high resolution images, which should not only contribute to better accuracies but also allow adequate visualization of target features during ground truth labeling.…”

“…However, previous studies on UAS imaging performance have shown that lower resolution data tend to reduce the accuracy of derived metrics e.g., plant height and biomass [69][70][71]. Other studies have also shown that lower resolution images tend to lower the performance of deep learning models [72,73]. In Koziarski and Cyganek [72], the robustness of deep learning models was evaluated against distortions such as blurring and concluded that the models were susceptible to such distortions.…”

Deep learning approaches for challenging species and gender identification of mosquito vectors

“…Although this process may introduce an additional step in the image annotation process before the training of the deep neural network, the quality of detection can be enhanced 34 . Furthermore, as reported, the proposed models cannot identify the objects when using low-quality images with a low resolution as the input 18 , 43 , and a low image resolution may degrade the result of the identification 44 . Suitable ranges for the image resolution should be determined to increase the detection rate of the model.…”

Do Radiographic Assessments of Periodontal Bone Loss Improve with Deep Learning Methods for Enhanced Image Resolution?

“…Moreover, it demonstrated that there are significant differences between the interpolation methods in this application, which suggests that using appropriate resolution improvements in the input images can positively impact the performances of deep-learning methods. Previous works also assessed the impacts of low-quality input images on the performances of deep-learning methods [ 15 , 16 ]. Dodge and Karam [ 15 ] evaluated how different aspects, such as blur, noise, contrast, and compression, can impact the performances of VGG-CNN-S [ 17 ], GoogleNet [ 18 ], and VGGNet [ 19 ] classifier models.…”

“…The results obtained in the aforementioned studies demonstrated that the tested networks are sensitive to blur effects, which, according to the authors, can be related to the interference in the images’ textures used by CNNs to identify patterns, and consequently, objects. Additionally, Koziarski and Cyganek [ 16 ] evaluated how using low-resolution input images affect the AlexNet [ 20 ], VGGNet, and ResNet networks in an image recognition task (Large Scale Visual Recognition Challenge 2012). The results presented in that study show that using resolution improvements leads to an increase in CNNs’ classification accuracy compared with using low-resolution images.…”