FIGURE 3 - uploaded by Dang N. H. Thanh
Content may be subject to copyright.
EfficientNet-B0 architecture, where K × K -a filter size, S -stride, B -feature maps, h × w -image size.
Source publication
Dish understanding from digital media is an interesting problem, but it also contains a big challenge. The challenge comes from the complexity of ingredients in the dish. With the development of deep learning, several effective tools can solve the problem partially. In this work, the task of dish recognition is considered. A novel dish recognition...
Context in source publication
Context 1
... family of efficient baseline networks (EfficientNet) [10] was developed based on convolutional neural networks by scaling all dimensions including width, height, and resolution, uniformly. EfficientNet-B0 is one of the architectures of the EfficientNet family and its architecture is presented in Figure 3. The EfficientNet-B0 architecture contains several layers: ...
Similar publications
The use of quantum technology in NLP tasks, especially sentiment classification, has the potential to be developed. In this research, we investigate the best technique to represent sentiment sentences so that sentiment can be analyzed using the Quantum-Enhanced Support Vector Machine (QE-SVM) algorithm. Investigations were carried out using circuit...
Citations
... G is considered an undirected PFG if e ij= e ij , ∀i,j=1,2,….,n.. On the contrary, G is considered a directed PFG. For a directed PFG, some degree values on G [11,[21][22][23] such as the picture fuzzy in degree (d I ), picture fuzzy out-degree (d o ), and picture fuzzy degree centrality (d) can be calculated in Eq. (6), Eq. (7), and Eq. (8) respectively. ...
In conventional monitoring forest protection, detection methods use optical sensors or RGB cameras combine features including smokes, fires and human-destroyed forests at national forests. This paper has presented a new approach using Deep learning integrated with Picture Fuzzy Set for the surveillance monitoring system to be activated to confirm human behaviour in real-time in forest protection. Picture Fuzzy Graph (PFG) are applied to solve many complex problems in the real-world problems. The paper has presented a novel approach using deep learning with knowledge graphs to find a human profile including the detection of humans in large data. In the proposed model, digital human profiles are collected from conventional databases combination with social networks in real-time, and a knowledge graph is created to represent complex-relational user attributes of human profile in large datasets. PFG is applied to quantify the degree centrality of nodes. To confirm the proposed model, the proposed model has been tested with data sets through case studies of a forest. Experimental results show that the proposed model has been validated on real world datasets to demonstrate this method’s effectiveness.
... A dozen of works have concerned the deployment of food recognition systems on smartphone or on cloud for realworld dish image analysis [32][33][34][35][36][37][38][39][40][41][42][43][44][45][46][47][48][49]. As to multistage approaches, Kawano and Yanai implemented a food recognition system on smartphone with the purpose of recording calories, nutrition, and eating habits [32]. is multistage system adopts linear SVM, bounding box adjustment, and food region estimation, and the classification accuracy of top 5 candidates reaches 79.2% on a 100-category food dataset. ...
... ey consider foods in restaurants and foods in wild, and fine-tuning GoogLeNet is found superior over k-nearest neighbor and ensemble SVM on the Food-500 dataset [41]. On foods from 6 restaurant chains dataset, its top-3 accuracy reaches 92.1%, while on foods across wild datasets, its accuracy should be further improved [43]. Aguilar proposed a deep network as a semantic food detector for smart restaurants [46]. ...
... (2) 31,127 images (3) > 1000 images per class [41] End-to-end approach Food201-MultiLabel Accuracy 50% < 1s (1) 201 classes (2) 50,374 images [42] End-to-end approach Office-31 Accuracy 97.71% Server: GPU (12 GB) Meal-300 Accuracy 72.45% [43] End-to-end approach UEH-VDR Accuracy 92.33% (1) 9 classes Precision 94.13% (2) 7,848 images Recall 90.82% [44] End-to-end approach FOOD-500 Accuracy 97.2% Server: CPU (1.6 GHz) ...
Deep learning is a new research direction in the field of machine learning, which was introduced into machine learning to bring it closer to its original goal. Accurate dish recognition becomes increasingly important in the multimedia community since it can help cuisine recommendation, calorie management, service improvement, and other food computing tasks. Many novel approaches have been developed on web recipes and menu pictures, while few are concerned real-life dish image analysis. In this study, a deep learning-based prototype system is deployed in a Chinese canteen, and 28 dish types, 16,904 images, and 45,061 instances have been collected. Specifically, in the prototype system, three practical issues are explored, including the backbone network selection, the training strategy determination, and the minimum number of samples for model upgrading. Experimental results suggest that fine-tuned Faster-RCNN can serve as the backbone network of the prototype system since it outperforms the other four fine-tuned networks on dish recognition (accuracy, 98.10%; recall, 97.20%; MAP (mean average precession), 98.30%) and satisfies real-time requirement (0.15 second per image). Meanwhile, the transferred backbone network achieves superior results (MAP, 96.48%) over the same architecture trained from image scratches (MAP, 87.84%). On model upgrading, a good (MAP, 91.34%) to better (MAP, 96.48%) outcome is obtained when the training size is increased from 50 to 200 samples per dish type, and 150 and more instances should be annotated if a new dish type is added to the system’s recognition list. Conclusively, the real-life deployment and evaluation of the prototype system indicate that deep learning is full of potential to enhance customer experience through accurate daily dish recognition.
... Although our method predicts pig breeds very accurately, but some misclassifications occurred where two pigs belonging to two different breeds have very similar visual characteristics. In the future, we are planning to use deep learning-based approaches (Hoang et al., 2021;Thanh et al., 2022) to improve the accuracy of pig breed classification problem. We are planning to test different algorithms such as monarch butterfly optimisation (MBO) (Wang et al., 2019), earthworm optimisation algorithm (EWA) , elephant herding optimisation (EHO) (Wang et al., 2015), moth search (MS) algorithm (Wang, 2018), Slime mould algorithm (SMA) (Li et al., 2020a), hunger games search (HGS) (Yang et al., 2021), Runge Kutta optimiser (RUN) (Ahmadianfar et al., 2021), colony predation algorithm (CPA) (Tu et al., 2021), and Harris hawks optimisation (HHO) (Heidari et al., 2019) for pig breed prediction. ...
... Although our method predicts pig breeds very accurately, but some misclassifications occurred where two pigs belonging to two different breeds have very similar visual characteristics. In the future, we are planning to use deep learning-based approaches (Hoang et al., 2021;Thanh et al., 2022) to improve the accuracy of pig breed classification problem. We are planning to test different algorithms such as monarch butterfly optimisation (MBO) (Wang et al., 2019), earthworm optimisation algorithm (EWA) , elephant herding optimisation (EHO) (Wang et al., 2015), moth search (MS) algorithm (Wang, 2018), Slime mould algorithm (SMA) (Li et al., 2020a), hunger games search (HGS) (Yang et al., 2021), Runge Kutta optimiser (RUN) (Ahmadianfar et al., 2021), colony predation algorithm (CPA) (Tu et al., 2021), and Harris hawks optimisation (HHO) (Heidari et al., 2019) for pig breed prediction. ...
