Voting combinations-based ensemble of fine-tuned convolutional neural networks for food image recognition
- PDF / 2,237,316 Bytes
- 22 Pages / 439.642 x 666.49 pts Page_size
- 67 Downloads / 212 Views
Voting combinations-based ensemble of fine-tuned convolutional neural networks for food image recognition Erdal Tasci1 Received: 10 January 2020 / Revised: 8 July 2020 / Accepted: 28 July 2020 / © Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract Obesity is one of today’s most visible, uncared, and common public health problems worldwide. To manage weight loss, obtain calorie intake and record eating lists, the development of the diverse automatic dietary assessment applications has great importance. Recently, deep learning becomes a popular approach that provides outstanding image recognition results. In this paper, we use ResNet, GoogleNet, VGGNet, and InceptionV3 with finetuning based on deep learning for image-based and computer-aided food recognition task. We also apply six voting combination rules (namely, minimum probability, average of probabilities, median, maximum probability, product of probabilities, and weighted probabilities) for ensemble methods. The experimental results demonstrate that our proposed ensemble voting scheme with transfer learning gives promising results compared to the state-of-the-art methods on Food-101, UEC-FOOD100, and UEC-FOOD256 image datasets. Keywords Food recognition · Deep learning · CNN · Image processing · Ensemble learning · Classification · Voting · Optimization
1 Introduction Obesity, known as adiposity, is a disease and global concern that can cause major health risks including diabetes, cardiovascular diseases, high blood cholesterol, sleep disorders, and some cancers. In 2016, more than 1.9 billion adults, over 340 million children, and adolescents were overweight worldwide according to the World Health Organization (WHO) [34]. Nowadays, people are more conscious about their health conditions due to increased health risks. To prevent and treat overweight and obesity problem, lifestyle changes including heart-healthy eating, controlled and increased physical exercise programs, and accurate dietary assessment is required according to the National Heart, Lung, and Blood Institute [25]. Erdal Tasci
[email protected] 1
Ege University, Computer Engineering Department, Izmir, Turkey
Multimedia Tools and Applications
Documenting food, calories and physical exercise intake is crucial for weight management and dietary assessment. With the increase in technological devices such as mobile phones, tablet PCs, cameras, and notebooks, food logging is facilitated in daily life and people can use cameras for capturing and recording food images. Food image recognition is concerned with identifying food types from images. In other words, food recognition is considered as a type of relatively harder visual recognition problem due to its fine-grained structure for providing automatic dietary monitoring and assessment [35]. CNN is a type of deep neural network which is inspired by the biological process called the animal visual cortex [10]. This network is commonly used for solving complex image recognition problems for beating records and produ
Data Loading...
