25 Years of CNNs: Can We Compare to Human Abstraction Capabilities?
By Sebastian Stabinger, Antonio Jose Rodríguez-Sánchez, Justus H. Piater. 2016
58 citations recorded by the source corpus.
We try to determine the progress made by convolutional neural networks over the past 25 years in classifying images into abstract classes. For this purpose we compare the performance of LeNet to that of GoogLeNet at classifying randomly generated images which are differentiated by an abstract property (e.g., one class contains two objects of the same size, the other class two objects of different sizes). Our results show that there is still work to do in order to solve vision problems humans are able to solve without much difficulty.
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
- Luc Van Gool
- Andrew Zisserman
- Cordelia Schmid
- Jitendra Malik
- Alan L. Yuille
- Ming-Hsuan Yang
- Xiaogang Wang
- Andrea Vedaldi
Citations in the corpus
- A Study and Comparison of Human and Deep Learning Recognition Performance under Visual Distortions (2017, background): “Stabinger et al. [8] test modern deep networks on these same classification tasks, but find that the networks can still not achieve human accuracy.”
- Human and DNN Classification Performance on Images With Quality Distortions (2019, background): “Some existing studies that compare HVS and DNN performance do not consider distorted images (e.g., Russakovsky et al. 2015; Kheradpisheh et al. 2016; Stabinger et al. 2016).”
- The Notorious Difficulty of Comparing Human and Machine Perception (2020, unknown): “Previous studies (J. Kim et al., 2018; Stabinger et al., 2016) explored how well deep neural networks can learn visual relations by testing them on the Synthetic Visual Reasoning Test (Fleuret et al., 2011).”
- Object Level Visual Reasoning in Videos (2018, background): “Humans have this extraordinary ability of performing visual reasoning on very complicated tasks while it is currently very hard for contemporary computer vision algorithms [9, 10].”
- Not-So-CLEVR: learning same–different relations strains feedforward neural networks (2018, methodology): “recent work has shown how CNNs, including variants of the popular LeNet [6] and AlexNet [7] architectures, could only solve a handful of the 23 SVRT problems [8,9].”