IAES International Journal of Artificial Intelligence (IJ-AI)

Vol. 13, No. 3, September 2024, pp. 3211~3217

ISSN: 2252-8938, DOI: 10.11591/ijai.v13.i3.pp3211-3217  3211

Barcode-less fruits classification using deep learning

Alaa Sheta1, Abdel Karim Baareh2, Amal Abdel-Raouf1, Peter Rausch3

1
Department of Computer Science, Southern Connecticut State University, New Haven, USA
2
Department Computer Science, Information Technology College, Isra University, Amman, Jordan
3
Department Computer Science, Nuremberg Institute of Technology Georg Simon Ohm, Nuremberg, Germany

Article Info ABSTRACT

Article history: Barcode-less fruit recognition technology has revolutionized checkout by
eliminating manual barcode scanning. This technology automatically
Received Sep 21, 2023 identifies and adds fruit items to the purchase list, significantly reducing
Revised Nov 13, 2023 waiting times at the cash register. Faster checkouts enhance customer
Accepted Jan 6, 2024 convenience and optimize operational efficiency for retailers. Adding
barcodes to fruits requires adhesives on the fruit surface that may cause
health hazards. Leveraging deep learning techniques for barcode-less fruit
Keywords:
recognition brings valuable advantages to industries, including advanced
Automatic fruit recognition automation, enhanced accuracy, and increased efficiency. These benefits
Classification translate into improved productivity, cost reduction, and superior quality
Convolution neural network control. This research provides our initial idea of developing a convolutional
Deep learning neural network (CNN) designed specifically for automatic fruit recognition,
Fruit detection even in challenging real-world scenarios. The proposed method assists fruit
sellers in accurately identifying and distinguishing between different types
of fruit that may exhibit similarities. A dataset with 44,406 images of
different fruit types is used to train and test our technique. Employing a
CNN, the developed model achieves an impressive classification accuracy of
97.4% during the training phase and 88.6% during the testing phase,
respectively, showcasing its effectiveness in precise fruit recognition.
This is an open access article under the CC BY-SA license.

Corresponding Author:
Amal Abdel-Raouf
Department of Computer Science, Southern Connecticut State University
New Haven, CT, USA
Email: [email protected]

1. INTRODUCTION
Barcode scanning streamlines the checkout process for most products in retail stores. Nowadays,
barcodes are utilized everywhere. During checkout in the supermarket, customers scan barcodes to purchase
various products. A barcode is a readable code with encoded numbers or patterns of lines and spaces. This
sequence together is unique for a product. Barcodes are attached to products to specify essential information
about an item. This information is commonly accessed from an inventory database. Scanning barcodes can
speedily get data about a product, such as the value, make, or description. There are several types of
barcodes, either one or two dimensions. A 1-dimensional (1D) barcode, sometimes called a linear barcode,
can store text information of up to 20 characters. A 2-dimensional (2D) barcode can be presented using
various shapes such as rectangles, hexagons, and other geometric patterns. They are named matrix codes. 2D
barcodes can store larger datasets of up to about 2,000 characters. 2D barcodes are relatively more secure
than 1D barcodes because they allow data encryption. Scanning fruits and vegetables presents a unique
challenge, requiring manual identification and searching within the system.

Journal homepage: http://ijai.iaescore.com

3212  ISSN: 2252-8938

There are two primary classification problems: i) distinguishing between various fruit types, such as
apples and oranges, and ii) classifying fruits of the same type (e.g., various apples). Exact classification is
difficult due to variations in the fruit's form, color, and various factors. Additionally, fruits can be enclosed in
plastic bags, adding another layer of complexity.
In today's age, barcode technology remains extensively used in fruit stores and superstores to
acquire fruit prices and essential information, such as tracing the source of a product. One significant
advantage of barcode scanning technology is its reasonably low cost compared to other technologies.
Barcode scanning is implemented using scanners, which are relatively low-cost and easy to sustain. This
technology also allows stores to easily access their inventory database to check the status and availability of
any item, so it helps minimize cost and effort in managing and planning reserves.
Machine learning has gained extensive attention, mainly algorithms focused on object detection and
recognition [1]. Typically, fruit stores and supermarkets package fruits and vegetables in small boxes and
utilize barcodes to determine their prices [2], [3]. Nevertheless, many customers still prefer the hands-on
experience of selecting fruits rather than prepackaged alternatives. Huo et al. [1] provided a method to
improve the accuracy of QR image code recognition by proposing an enhanced adaptive median filter
algorithm and QR code distortion correction technique based on artificial neural networks. The proposed
method compares the distorted QR images and their pattern. This method shows significance in QR code
recognition. Barcode-less fruit recognition technology helps achieve threefold: i) it advances a customer's
shopping experience by allowing no-bar code scanning. Therefore, the purchase process becomes much more
accessible by only recognizing the fruit or vegetables by camera, ii) the purchase process becomes faster,
thus helping minimize the customer’s wait time and enhance customer experience, and iii) minimize the loss
of items, intentional or unintentional.
Deep learning models can extract the best features that describe an object from many images and
recognize or classify many objects in these images [4], [5]. Deep learning was also utilized in disease
diagnosis, such as sleep apnea, with promising results [6], [7]. CNN was successfully used in image
classification for fruit and vegetable recognition [8]. Aranda et al. [9] developed a CNN with a few layers to
achieve this task so that the checkout process at the supermarket is quick and straightforward. Innovative
technology in fruit recognition offers numerous advantages in various applications [10], [11]. Firstly, enables
efficient fruit detection and classification, facilitating automation and streamlining processes in the
agricultural industry. With smart devices equipped with advanced sensors and imaging capabilities, fruits can
be accurately recognized and categorized based on features such as shape, color, size, and texture. This
technology brings benefits to both farmers and consumers. Farmers can leverage fruit recognition to optimize
harvesting operations, ensuring that only ripe and high-quality fruits are picked, thereby reducing waste and
maximizing productivity. In earlier research, scientists proposed various methods merging computer vision
and machine learning to manually extract features from fruits and classify them based on these computer
vision features [2], [12]. These methods employed computer vision algorithms to analyze fruits' color, shape,
size, and texture characteristics, which can be used as input for classification algorithms. Many of these
approaches involved preprocessing or feature extraction using computer vision techniques and applying
different classifiers. However, these classifiers often needed more robustness across all types of fruits,
leading to higher misclassification rates.
Although barcode scanning technology does have many advantages, it was found that it still has
several drawbacks. For example, barcode scanners are not free of charge, particularly in big stores where we
need many machines that might encounter malfunctions. This technology can not avoid human error. Human
workers can still make mistakes while scanning the product. Errors in providing product information might
cause both inventory errors and customer disappointment. Moreover, this dependence on barcodes presents a
significant challenge for shopkeepers who must remember and manage the barcodes associated with each
fruit category [13], [14]. One important issue related to barcodes is that, in many cases, the adhesives used on
the barcode stickers might contain harmful chemicals that could affect human health [15]. The skin of fruits
and vegetables might absorb the substances in adhesives. Therefore, it is mandatory to investigate more about
the adhesives utilized for barcode stickers. A wide variety of chemicals are utilized in the manufacture of
adhesives. Adhesives might have harmful chemicals that affect customer’s health. An important question
arises: can we invest in barcode-less item recognition technology?
Our proposed work includes the development of a barcode-less fruit recognition model based on
CNN. The proposed model was trained on a large data set of images containing 15 different fruits and
vegetables. We designed our CNN model with limited complexity and a limited number of layers. The
images in our database were resized to 64×64 pixels. Each class had a maximum of 100 images. The study
reports the challenge of correctly recognizing the characteristics of fruits and vegetables with a barcode-
based method. The basic idea of recognizing fruits or goods without barcodes can be presented in Figure 1.

Int J Artif Intell, Vol. 13, No. 3, September 2024: 3211-3217

Int J Artif Intell ISSN: 2252-8938  3213

Figure 1. Barcode less recognition system [16]

2. METHOD
Deep learning is a subdivision of artificial intelligence (AI). Deep learning is a standard extension of
the well-known artificial neural network model. It can learn from various data types, including images, texts,
and numerical data. It has the credit of being the reason for the success of various achievements in the AI
domain, such as natural language processing, self-driving cars, and many others. Deep learning is one of the
potent techniques that has been successfully used for detection, recognition, and classification in many
real-life applications [17], [18]. Each layer comprises multiple neural nodes with activation functions
connected to nodes in preceding and subsequent layers.
During training, the parameters of the neural network are adjusted to best fit the available data. With
sufficient training, deep neural network models can achieve remarkable pattern recognition capabilities
comparable to or surpassing human-level performance. Deep learning has been successfully used in
significant applications such as troop camouflage detection [19], automated COVID-19 detection per x-ray
chest images [20], biomedical applications [21], computer vision [22], improving computational chemistry
and drug design [23], addressing safety issues in the interface between pedestrians and autonomous vehicles
[24], and classifying alzheimer's disease [25]. The typical structure of CNNs comprises the following layer
types:
− Convolutional layers: these layers apply a set of learnable filters (kernels) to the input image, extracting
features by convolutions. Each filter detects specific patterns or features in the image, such as edges or
textures. Convolutional layers help to capture spatial hierarchies and local patterns in the data.
− Pooling layers: these layers are utilized in neural networks to decrease the input's spatial dimensions
(width and height) through downsampling. There are several types of pooling in CNN. They include max,
min, and average pooling. The pooling technique helps reduce the features' complexity from one layer to
another. It also helps reduce overfitting and the model sensitivity to changes in the input. Pooling also
causes downsampling, which allows the CNN model to pay more attention to significant features of the
input images. Mainly, pooling helps maintain the most relevant features while reducing the input's spatial
size.
− Batch normalization layers: this layer is responsible for reducing the possible overfitting and making the
model better generalized. It makes the model less sensitive to variations in the initial random weight and
hybrid parameters adopted for training it.
− Activation layers: several possible activation functions have been utilized in the literature. The rectified
linear unit (ReLU) is the most adopted function in CNNs. The ReLU function produces zeros for negative
values while keeping positive values unchanged.
− Fully connected layers: these layers, also recognized as dense layers, show a central role in deep neural
networks. These layers establish connections between every neuron in one layer and the subsequent layer.
They contribute to the classification process by leveraging the extracted features to make predictions. In
classification tasks, the neurons in the final fully connected layer align with the given output classes,
enabling the network to assign class labels to the input data.
These layers are stacked together to form the architecture of a CNN. The input flows through the
convolutional, pooling, and activation layers to extract features and progressively reduce the spatial
dimensions. After flattening the developed feature maps, they are advanced through fully connected layers
for some applications (i.e., classification). Designing the appropriate architecture for a CNN within a deep
learning context necessitates a blend of domain expertise, experimental exploration, and adherence to
established methodologies.
Our proposed approach introduces a simple and efficient CNN framework comprising 16 layers
(see Figure 2). The adopted CNN architecture design for fruit detection is provided in Table 1. In our specific

Barcode-less fruits classification using deep learning (Alaa Sheta)

3214  ISSN: 2252-8938

scenario, we focus on constructing a streamlined architecture that optimizes computational efficiency while
upholding a commendable level of classification accuracy. Notably, indiscriminate augmentation of layers
can often lead to overfitting the data. In our pursuit, however, we managed to strike a harmonious
equilibrium, attaining a satisfactory classification performance across the training and testing stages.
Furthermore, we bound the inherent characteristics of original color images with dimensions of 64×64×3.
This choice of resolution affords us the advantage of constraining the complexity of the CNN architecture,
consequently enhancing its efficiency. This framework permits the network to learn the best features from a
large dataset of input images without preprocessing. To evaluate the performance of our proposed network,
we implement experiments using an image dataset incorporating diverse real-world scenarios, described in
the following section.

Figure 2. Proposed CNN architecture

Table 1. Proposed CNN network architecture developed using MATLAB

No. Name Description
1 imageinput Image Input: 64×64×3 images with' zerocenter' normalization
2 conv 1 Convolution: 8 3×3×3 convolutions, stride [1, 1], padding' same'
3 maxpool 1 Max Pooling: 2×2, stride [2, 2], padding [0, 0, 0, 0]
4 batchnorm 1 Batch Normalization: 8 channels
5 relu 1 ReLU: Rectified Linear Unit activation
6 conv 2 Convolution: 16 3×3×8 convolutions, stride [1, 1], padding' same'
7 maxpool 2 Max Pooling: 2×2, stride [2, 2], padding [0, 0, 0, 0]
8 batchnorm 2 Batch Normalization: 16 channels
9 relu 2 ReLU: Rectified Linear Unit activation
10 conv 3 Convolution: 32 3×3×16 convolutions, stride [1, 1], padding' same'
11 maxpool 3 Max Pooling: 2×2, stride [2, 2], padding [0, 0, 0, 0]
12 batchnorm 3 Batch Normalization: 32 channels
13 relu 3 ReLU: Rectified Linear Unit activation
14 fc Fully Connected: 15 fully connected layer
15 softmax Softmax: Softmax activation
16 classoutput Classification Output: Crossentropy with' Apple' and 14 other classes

3. DATASET
In this research, we employed a database compiled by Hussain et al. [26]. The database comprises a
total of 44,406 fruit images that were gathered over a span of 6 months. These images were captured in a
controlled laboratory environment. They encompassed various scenarios and diverse lighting conditions,
including fluorescent lighting, natural light shadows, sunshine, pose variations, illumination variations,
camera-capturing artifacts, specular reflection shading, and shades. The images were captured on a
transparent background using an HD Logitech web camera with a 320×258 pixels resolution. The dataset, as
presented in [26] consists of multiple fruit classes, such as apple, banana, carambola, guava, kiwi, mango,
muskmelon, orange, peach, pear, persimmon, pitaya, plum, pomegranate, and tomato. A sample of the
adopted dataset is shown in Figure 3. To determine calibration, all images were resized to a resolution of
64×64 pixels. Each class had a maximum of 100 images.

Int J Artif Intell, Vol. 13, No. 3, September 2024: 3211-3217

Int J Artif Intell ISSN: 2252-8938  3215

Apple Banana Carambola Guava Kiwi

Mango Muskmelon Orange Peach Pear

Persimmon Pitaya Plum Pomegranate Tomatoes

Figure 3. Sample of the fruits [25]

4. RESULTS AND DISCUSSION

To assess the performance of our model, we employ the accuracy metric, which gauges the ratio of
correct predictions to the total predictions made during each phase. In our experiments, we achieved a
training accuracy of 97.4% and a testing accuracy of 88.6%. These accuracy metrics provide insights into the
model's performance on the training and testing datasets. The high training accuracy of 97.4% indicates that
the model has effectively learned to make accurate predictions on the training data. This suggests that the
model has successfully captured the underlying patterns and features within the training dataset. On the
contrary, the testing accuracy of 88.6% demonstrates the model's ability to generalize its predictions to
previously unseen data. Although the testing accuracy is slightly lower than the training accuracy, it still
showcases the model's competence in handling new data effectively. For a visual representation of the
model's learning process, please refer to Figure 4 which illustrates the convergence curves for both loss and
accuracy. Figures 5(a) and (b) provides a confusion chart for training and testing cases. These visualizations
affirm that the model has learned to generalize effectively and can accurately predict new, unseen examples.

Figure 4. Accuracy and loss curves after training the proposed CNN

Barcode-less fruits classification using deep learning (Alaa Sheta)

3216  ISSN: 2252-8938

(a) (b)

Figure 5. The confusion chart of (a) training and (b) testing cases

5. CONCLUSION
This research designed a CNN model with 16 hidden layers to recognize 15 fruits and vegetables.
Scanning fruits and vegetables presents a challenge, demanding manual work by employees to pick up the
correct item in a database. The proposed model can distinguish between similar fruits in sizes and shapes.
Our data set had 44,406 images of various fruit types utilized for model development. The performance of
the developed CNN model was 97.4% and 88.6% at both the training and testing phases, respectively. This
application of deep Learning can help enhance market productivity, reduce the need for workers with unique
skills, and allow better quality control. This model could be deployed on mobile phones for possible use at
supermarkets.

REFERENCES
[1] L. Huo, J. Zhu, P. K. Singh, and P. A. Pavlovich, “Research on QR image code recognition system based on artificial intelligence
algorithm,” Journal of Intelligent Systems, vol. 30, no. 1, pp. 855–867, 2021, doi: 10.1515/jisys-2020-0143.
[2] G. R. Narayan and V. James, “Barcode recognition from video by combining image processing and xilinx,” Procedia
Engineering, vol. 38, pp. 2140–2146, 2012, doi: 10.1016/j.proeng.2012.06.257.
[3] N. Pradhan, D. R. K. Tyagi, and M. P. Nagpal, “Barcode recognition techniques: review & application,” International Journal of
Innovative Research in Computer Science & Technology, vol. 9, no. 3, 2021, doi: 10.21276/ijircst.2021.9.3.6.
[4] D. T. P. Chung and D. V. Tai, “A fruits recognition system based on a modern deep learning technique,” Journal of Physics:
Conference Series, vol. 1327, no. 1, 2019, doi: 10.1088/1742-6596/1327/1/012050.
[5] C. C. Ukwuoma, Q. Zhiguang, M. B. B. Heyat, L. Ali, Z. Almaspoor, and H. N. Monday, “Recent advancements in fruit detection and
classification using deep learning techniques,” Mathematical Problems in Engineering, vol. 2022, pp. 1–29, 2022, doi: 10.1155/2022/9210947.
[6] A. Sheta et al., “Diagnosis of obstructive sleep apnea using feature selection, classification methods, and data grouping based age,
sex, and race,” Diagnostics, vol. 13, no. 14, 2023, doi: 10.3390/diagnostics13142417.
[7] A. Sheta et al., “Diagnosis of obstructive sleep apnea from ECG signals using machine learning and deep learning classifiers,”
Applied Sciences, vol. 11, no. 14, 2021, doi: 10.3390/app11146622.
[8] L. Ale, A. Sheta, L. Li, Y. Wang, and N. Zhang, “Deep learning based plant disease detection for smart agriculture,” in 2019
IEEE Globecom Workshops, GC Wkshps 2019 - Proceedings, IEEE, 2019. doi: 10.1109/GCWkshps45667.2019.9024439.
[9] J. L. R. -Aranda, J. I. N. -Varela, J. C. C. -Tello, and G. R. -Ramirez, “Fruit classification for retail stores using deep learning,” in
Pattern Recognition, vol. 12088, Springer International Publishing, 2020, pp. 3–13. doi: 10.1007/978-3-030-49076-8_1.
[10] N. E. A. Mimma, S. Ahmed, T. Rahman, and R. Khan, “Fruits classification and detection application using deep learning,”
Scientific Programming, vol. 2022, pp. 1–16, Nov. 2022, doi: 10.1155/2022/4194874.
[11] D. Hussain, I. Hussain, M. Ismail, A. Alabrah, S. S. Ullah, and H. M. Alaghbari, “A simple and efficient deep learning-based framework
for automatic fruit recognition,” Computational Intelligence and Neuroscience, vol. 2022, pp. 1–8, 2022, doi: 10.1155/2022/6538117.
[12] B. Reed, “Increase speed, accuracy with mobile barcode scanning,” Smartsheet, [Online]. Available:
https://www.smartsheet.com/blog/increase-speed-accuracy-mobile-barcode-scanning
[13] A. Rocha, D. C. Hauagge, J. Wainer, and S. Goldenstein, “Automatic fruit and vegetable classification from images,” Computers
and Electronics in Agriculture, vol. 70, no. 1, pp. 96–104, 2010, doi: 10.1016/j.compag.2009.09.002.
[14] Y. Cha, T. Kim, D. Kim, and B. Cha, “Draft design of fruit object recognition using transfer learning in smart farm,” in ACM
International Conference Proceeding Series, ACM, Sep. 2020, pp. 117–120. doi: 10.1145/3426020.3426048.
[15] “Stickers on fruits and vegetables: are they safe?,” Food Safety Mantra. [Online]. Available:
https://www.foodsafetymantra.com/regulatory-insight/consumer-products/stickers-on-fruits-and-vegetables-are-they-safe/
[16] “New grocery checkout scanner identifies produce and packaged goods without barcodes,” Barcode Inc, 2012. [Online]. Available:
https://www.barcodesinc.com/news/new-grocery-checkout-scanner-identifies-produce-and-packaged-goods-without-barcodes/
[17] W. C. Seng and S. H. Mirisaee, “A new method for fruits recognition system,” in Proceedings of the 2009 International
Conference on Electrical Engineering and Informatics, ICEEI 2009, pp. 130–134. doi: 10.1109/ICEEI.2009.5254804.
[18] X. Chang, “Application of computer vision technology in post-harvest processing of fruits and vegetables: starting from shape
recognition algorithm,” in International Conference on Applied Artificial Intelligence and Computing, ICAAIC 2022, IEEE, 2022,

Int J Artif Intell, Vol. 13, No. 3, September 2024: 3211-3217

Int J Artif Intell ISSN: 2252-8938  3217

pp. 934–937. doi: 10.1109/ICAAIC53929.2022.9793255.

[19] Muslikhin, A. Nasuha, Suprapto, A. Winursito, and F. Arifin, “Troop camouflage detection based on deep action learning,” IAES
International Journal of Artificial Intelligence, vol. 11, no. 3, pp. 859–871, Sep. 2022, doi: 10.11591/ijai.v11.i3.pp859-871.
[20] L. Wang, Z. Q. Lin, and A. Wong, “COVID-Net: a tailored deep convolutional neural network design for detection of COVID-19
cases from chest X-ray images,” Scientific Reports, vol. 10, no. 1, Nov. 2020, doi: 10.1038/s41598-020-76550-z.
[21] M. Wainberg, D. Merico, A. Delong, and B. J. Frey, “Deep learning in biomedicine,” Nature Biotechnology, vol. 36, no. 9, pp.
829–838, 2018, doi: 10.1038/nbt.4233.
[22] A. Voulodimos, N. Doulamis, A. Doulamis, and E. Protopapadakis, “Deep learning for computer vision: a brief review,”
Computational Intelligence and Neuroscience, vol. 2018, pp. 1–13, 2018, doi: 10.1155/2018/7068349.
[23] A. C. Mater and M. L. Coote, “Deep learning in chemistry,” Journal of Chemical Information and Modeling, vol. 59, no. 6, pp.
2545–2559, 2019, doi: 10.1021/acs.jcim.9b00266.
[24] Z. Zhu, Z. Hu, W. Dai, H. Chen, and Z. Lv, “Deep learning for autonomous vehicle and pedestrian interaction safety,” Safety
Science, vol. 145, 2022, doi: 10.1016/j.ssci.2021.105479.
[25] S. Bringas, S. Salomón, R. Duque, C. Lage, and J. L. Montaña, “Alzheimer’s disease stage identification using deep learning
models,” Journal of Biomedical Informatics, vol. 109, Sep. 2020, doi: 10.1016/j.jbi.2020.103514.
[26] I. Hussain, Q. He, Z. Chen, and W. Xie, “Fruit recognition dataset (version V 1.0),” Zenodo, pp. 1–7, 2018.

BIOGRAPHIES OF AUTHORS

Alaa Sheta obtained his Ph.D. of Information Technology from George Mason
University, Virginia, U.S.A. in 1997. Prior to that, he earned his B.E. and M.Sc. degrees in
Electronics and Communication Engineering from Cairo University, Egypt in 1988 and 1994,
respectively. Currently, he holds a tenured professorship at the Department of Computer Science
in Southern Connecticut State University in New Haven, Connecticut, USA. His research
interests span various areas including deep Learning, meta-heuristics, machine learning, data
mining, image processing, and robotics. He has made significant contributions to these fields,
publishing more than 170 papers in esteemed international journals and conferences. In addition
to his scholarly achievements, he has actively participated in various capacities such as serving
as a chair, guest editor, and program committee member for numerous international events. He
can be contacted at email: [email protected].

Abdel Karim Baareh holds a Doctor of Informatics/Artificial Intelligence degree

from Damascus University, Syria, which he earned in 2009. He also completed his B.Sc. and
M.C.A (Computer Application) degrees from Mysore and Bangalore University, India, in 1992
and 1999, respectively. Currently, he is on sabbatical leave and affiliated with the Department of
Data Science and Artificial Intelligence at the Information Technology College, Isra University in
Jordan. He holds a permanent professorial position within the Department of Computer Science
at Al-Balqa Applied University, Jordan. His extensive research portfolio spans various domains,
including metaheuristics, global optimization, machine learning, data mining, bioinformatics,
graph theory, and parallel programming. Over the years, he has contributed significantly to his
field, with a publication record that boasts over 35 papers in international journals and
conferences. He can be contacted at email: [email protected] or [email protected].

Amal Abdel-Raouf received her M.Sc. in Computer Engineering from Cairo

University in 1994 and her Ph.D. in Computer Science and Engineering in June 2005 from the
University of Connecticut, United States. She is currently a professor at the Department of
Computer Science, Southern Connecticut State University, CT, United States. She is the author
of many papers in journals and international conferences in the field of software engineering.
She has worked as a technical commit- tee member and a reviewer of many journals,
international conferences, and workshops. Her research interests include big data, software
engineering, software quality, real time systems, parallel and distributed computing, object-
oriented systems, and artificial intelligence. She can be contacted at email:
[email protected].

Peter Rausch is a professor of Information Systems at the Nuremberg Institute of

Technology-George Simon Ohm, with an extensive publication record comprising numerous
papers, articles, and book chapters. Notably, he recently authored a book on performance
management and business intelligence. His ongoing research endeavors are centered on fuzzy
technologies, business planning, and process automation. His academic journey includes earning a
Ph.D. in business administration from Frankfurt University and several years of professional
experience in software development, business process optimization, and consulting. In a previous
role as a program manager, he led process and tool management within the Department of Sales
Support at Sun Microsystems, Germany. Subsequently, he took charge of an integration
environment as part of an industry solution at a software company. His academic career continued
as a professor of Information Systems at Coburg University of Applied Sciences before he accepted
an offer in 2008 to become a professor of Information Systems at the Technische Hochschule
Nürnberg-Georg Simon Ohm. He can be contacted at email: [email protected].
Barcode-less fruits classification using deep learning (Alaa Sheta)