Performance Analysis of Deep Learning CNN Models for Variety Classification in Hazelnut

Taner, Alper; Öztekin, Yeşim Benal; Duran, Hüseyin

doi:10.3390/su13126527

Cited by 49 publications

(28 citation statements)

References 54 publications

(56 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Table 1 shows the evaluation metrics of the train set, validation set and test set of the CNN model. To consider a model as the best model, it must have performed well in the training data set and the validation data set [39]. In this sense, results shown in Table 1 were found to be from the best model.…”

Section: Resultsmentioning

confidence: 98%

A Deep CNN Model for Inferring 3D Human Body Shapes Using Front and Side Images

Crespo¹,

Spanlang²

2021

Proceedings of 3DBODY.TECH 2021 - 12th International Conference and Exhibition on 3D Body Scanning and Processing Technologies,

View full text Add to dashboard Cite

Immersive Virtual Reality (IVR) studies indicate that there is some level of the brain that does not distinguish between reality and virtual reality. In this context, a self avatar embodied from first person perspective brings a significant and lasting change to the user. IVR is therefore widely used in research and for psychological and physiological health rehabilitation. We use IVR in a wide range of areas in pain and mobility, emotional health, diversity equity and inclusion, to rehabilitate domestic violence offenders and to promote healthier lifestyles among obese people (Fig. 1). A remaining challenge is to accurately and efficiently create avatars with body shapes and appearance that closely match those of the real user's bodies. This is owing to the huge differences in human body forms, the reduction of the complex human shape by body scanning technology and the complexity of acquiring accurate body measurements. The primary objective of this work was to construct a cost-effective and accurate model to infer the 3D shape from a front and side image of a person taken with a smartphone. To achieve this, we used a fully morphable human body model to change the body shape using a set of body shape modifying parameters. We create a dataset of thousands of computer generated front and side images varying the shape modifiers of the morphable model. We then train a convolutional neural network (CNN) using that dataset (Fig. 3). Our approach efficiently infers 3D human body shapes from a person's front and side image generating an accurate representation of a person. We made preliminary tests using a set of 10 body scans with known measurements, creating computer generated front and side images of the scans and using these images as input to the CNN and to compare the resulting body shape with the original 3D body scan. Our results demonstrate the effectiveness of the designed approach (Fig. 5, Fig. 6). Our proposed model enables us to create a fully movable avatar that can be embodied in IVR from a front and side smartphone photo in a fully automated way. The same inferred shape modifiers can also be used on the clothing of the avatar to enable us to dress the avatar. Although a larger comparative study needs to be performed before the use of this approach can be routinely recommended, we believe that the convenience and ease-of-use of this model will contribute to increase the reach of VR tools with look-alike avatars also in clinical settings.

show abstract

Section: Resultsmentioning

confidence: 98%

A Deep CNN Model for Inferring 3D Human Body Shapes Using Front and Side Images

Crespo¹,

Spanlang²

2021

Proceedings of 3DBODY.TECH 2021 - 12th International Conference and Exhibition on 3D Body Scanning and Processing Technologies,

View full text Add to dashboard Cite

show abstract

“…These results can be categorized as the good ones (accuracy between 26% -36% for YOLOv5s lower than other YOLO versions [30]) which potential to be implemented for real-world applications such as object detection of obstacle avoidance mechanisms in mobile robot for volcano monitoring application. However, the model performance should be improved for the application by: getting more data, rescaling the data, transforming the data, using feature selection for the data, rearranging the dataset distribution, using better resolution images, and using a specific group of objects being trained [19,31].…”

Section: Resultsmentioning

confidence: 99%

Design of Real-Time Object Detection in Mobile Robot for Volcano Monitoring Application

Evita

Mustikawati

Djamal

2022

J. Phys.: Conf. Ser.

View full text Add to dashboard Cite

Indonesia is one of the countries located at the ring of fire which should be monitored to predict the eruption earlier and set the risk zones around with no human involvement especially while eruption taking place. Therefore, in this research, it is used a 4 wheeled mobile robot called PRAWIRA for this purpose. The robot should have the ability to avoid the obstacles in front of it in this area. It has been designed a real-time object detection system for volcano monitoring application using deep learning from the YOLOv5s model for 4 objects (trees, persons, stones, and stairs). It was used 484 images for the dataset after the pre-train process was conducted with several steps: object identification; dataset downloading (Google Chrome Extension and Open Images v6); image labeling (LabeImg); augmentation process (flip, blur, and rotation); and data training for varies epochs and batches by Jupyter Notebook GPU. The preliminary result for this research was presented in the mean average precision (mAP) of YOLOv5s (the smallest version). The first variation (batch = 16, epochs = 100) resulted in mAP_0.5 = 17.9% and mAP_0.5:0.95 = 7.27% with 0.262 hours of training time. The second (batch = 16, epochs = 500) resulted in mAP_0.5 = 25.7% and mAP_0.5:0.95 = 12.3% with 1.296 hours of training time, while the third (batch = 80, epochs = 100) resulted in mAP_0.5 = 17.7% and mAP_0.5:0.95 = 5.63% with 0.232 hours of training time. Furthermore, the last variation (batch = 80, epochs = 500) resulted in mAP_0.5 = 19.5% and mAP_0.5:0.95 = 8.92% with 1.085 hours of training time. Therefore, the second variation is the best result for the model with 14.8 MB of size. Moreover, interfaces for the best model were displayed to show the result of the training.

show abstract

“…In recent years, many efficient pre-trained image classifier algorithms have been developed and made available for research, owing to the efforts of the machine learning developer communities [ 33 , 34 , 35 , 36 , 37 ]. A pre-trained model has been trained to solve a problem similar to the one in hand using a large benchmark dataset.…”

Section: Methodsmentioning

confidence: 99%

Identification of Autism in Children Using Static Facial Features and Deep Neural Networks

Rahman

Subashini

2022

Brain Sciences

View full text Add to dashboard Cite

Autism spectrum disorder (ASD) is a complicated neurological developmental disorder that manifests itself in a variety of ways. The child diagnosed with ASD and their parents’ daily lives can be dramatically improved with early diagnosis and appropriate medical intervention. The applicability of static features extracted from autistic children’s face photographs as a biomarker to distinguish them from typically developing children is investigated in this study paper. We used five pre-trained CNN models: MobileNet, Xception, EfficientNetB0, EfficientNetB1, and EfficientNetB2 as feature extractors and a DNN model as a binary classifier to identify autism in children accurately. We used a publicly available dataset to train the suggested models, which consisted of face pictures of children diagnosed with autism and controls classed as autistic and non-autistic. The Xception model outperformed the others, with an AUC of 96.63%, a sensitivity of 88.46%, and an NPV of 88%. EfficientNetB0 produced a consistent prediction score of 59% for autistic and non-autistic groups with a 95% confidence level.

show abstract

Performance Analysis of Deep Learning CNN Models for Variety Classification in Hazelnut

Cited by 49 publications

References 54 publications

A Deep CNN Model for Inferring 3D Human Body Shapes Using Front and Side Images

A Deep CNN Model for Inferring 3D Human Body Shapes Using Front and Side Images

Design of Real-Time Object Detection in Mobile Robot for Volcano Monitoring Application

Identification of Autism in Children Using Static Facial Features and Deep Neural Networks

Contact Info

Product

Resources

About