论文信息 - Characteristic Similarity Using Classical CNN Model

Characteristic Similarity Using Classical CNN Model

Deep neural network such as classical CNN become very popular used in image classification. CIFAR-10 dataset provides the ten image categories dataset with 32x32 pixels of size. Since the similarity became issue when using dataset which is not providing the enough information for classification [2] and labeled subset of tiny image in CIFAR-10 [3], we modify the classical CNN model with several parameters to reach the good similarity result in learning. We find the model can give the high similarity with 90% for representative image in dataset.

Motoki Amagasaki | Morihiro Kuga | Mery Diana | Juntaro Chikama | Masahiro Iida

[1] Mohamed Moustafa,et al. CIFAR-10: KNN-Based Ensemble of Classifiers , 2016, 2016 International Conference on Computational Science and Computational Intelligence (CSCI).

[2] Srikar Appalaraju,et al. Image similarity using Deep CNN and Curriculum Learning , 2017, ArXiv.

[3] Shan Sung Liew,et al. Gender classification: a convolutional neural network approach , 2016 .

[4] Benjamin Recht,et al. Do CIFAR-10 Classifiers Generalize to CIFAR-10? , 2018, ArXiv.