This paper gives a comparative study on the performances of several deep learning methods for the food images recognition challenge. The experiments were conducted on the UEC Food-100 dataset using ResNet-18, Inception-V3, Resnet-50, Densenet-121, Wide Resnet-50 and ResNext-50 with images of size 320x320 and 299x299. The limited size of the database required the transfer learning approach; that is, all models were trained with pretrained ImageNet weights. The best classification result was obtained using ResNext-50 with 87.7 % accuracy.
Paper
Full text
A Comparative Study of Deep Learning Methods on Food Classification Problem
Semantic Scholar · Computer Science · 2020
Abstract
This paper gives a comparative study on the performances of several deep learning methods for the food images recognition challenge. The experiments were conducted on the UEC Food-100 dataset using ResNet-18, Inception-V3, Resnet-50, Densenet-121, Wide Resnet-50 and ResNext-50 with images of size 320x320 and 299x299. The limited size of the database required the transfer learning approach; that is, all models were trained with pretrained ImageNet weights. The best classification result was obtained using ResNext-50 with 87.7 % accuracy.