Professional Documents
Culture Documents
Deep-Learning
1st Prof. Sandeep Shinde 2nd Swanand Kulkarni 3rd Akhilesh Kulkarni
Head of Department Department of Computer Engineering Department of Computer Engineering
Computer Engineering Vishwakarma Institute of Technology Vishwakarma Institute of Technology
Vishwakarma Institute of Technology Pune, India Pune, India
Pune, India swanand.kulkarni21@vit.edu kshitij.meshram21@vit.edu
sandeep.shinde@vit.edu
Abstract — In response to the global demand due to its potential health hazards. Upon contact with
for efficiently handling mango quality control moisture, calcium carbide releases acetylene gas,
and inspection processes, this project harnesses acting as a ripening agent but posing risks to
the capabilities of deep learning and object consumers. Our project delves into these risks,
detection. By focusing on automating the emphasizing the need for a vigilant examination of
mango classification process, the primary the effects on physical appearance, including
objective is to create a streamlined system for uniform ripening, texture, softness, and external
detecting artificial fruit ripening using a blemishes.
combination of sensors, image processing, and
machine learning models. The overarching aim Implemented in Python, with PyTorch as the chosen
is to ensure food safety and quality through the deep-learning framework, particularly leveraging
development of a cost-effective and reliable YOLOv5, the project utilizes Google Colab as the
solution. integrated development environment (IDE).
Additionally, Roboflow serves as a comprehensive
tool for data collection, preprocessing, and model
Keywords - Python, PyTorch, YOLOv5, Google
training techniques, providing a holistic approach to
Colab, Roboflow, Mango Quality Control,
addressing the challenges associated with mango
Artificial Ripening, Deep Learning, Object
quality control and the detection of artificial
Detection.
ripening. Through this project, we envision not just
an automation of processes but a significant stride
I. INTRODUCTION toward ensuring the safety and quality of mangoes
Mangoes, being a widely consumed fruit globally, cost-effectively and reliably.
undergo various quality control measures that often II. LITERATURE REVIEW
involve labor-intensive and time-consuming
Individuals in contemporary times, driving a
processes. In a bid to revolutionize this paradigm,
heightened focus on maintaining well-being
our project employs cutting-edge technologies,
through the consumption of nutritious Mangoes.
specifically deep learning and object detection. The
The regular intake of Mangoes is associated with
central goal is to automate and streamline the
numerous health benefits, contributing to a
mango classification process, enhancing efficiency
diminished risk of chronic diseases due to the
and accuracy.
essential nutrients they provide for the proper
The project addresses a critical aspect of fruit maintenance of the body. However, a growing
quality – the detection of artificial ripening. issue within the fruit market revolves around the
Traditional methods of ripeness inspection are artificial ripening of Mangoes using substances
prone to inefficiencies. Leveraging the power of like calcium carbide, raising serious health
machine learning, our system aims to provide a concerns, including the potential to induce cancer.
swift and accurate determination of the ripeness News reports during mango seasons frequently
status, ensuring adherence to quality standards. underscore the prevalence of artificially ripened
Mangoes in markets, despite persistent warnings
The use of ethylene gas and chemical agents for and enforcement efforts. The use of calcium
artificial ripening is a prevalent practice. However, carbide in mango ripening, in particular, has
the introduction of calcium carbide raises concerns
emerged as a cause for alarm. This chemical growing importance of health-conscious choices in
method has been linked to adverse health effects, the modern food landscape.
ranging from vomiting, diarrhea, and ulcers of
the throat and abdomen to general weakness, The maturation process and the shift in color in
permanent eye damage, and respiratory issues. Mangoes and vegetables exhibit a close correlation.
Recognizing the challenge of visually identifying The alteration in pigments such as chlorophyll,
artificially ripened Mangoes, a novel solution has carotenoids, flavonoids, and other vital compounds
been developed—a device leveraging image plays a decisive role in determining the color of
processing technology. This innovative system these crops. As Mangoes and vegetables ripen, the
involves capturing an image of the mango under levels of these specific pigments change,
scrutiny and comparing its features, such as consequently influencing the color of the produce.
histogram values, with those of naturally ripened Leveraging this intrinsic relationship between
Mangoes. Integrated into a smartphone through ripeness and color, and employing suitable
an Android application, this technology provides technology, one can ascertain the full development
an efficient and accurate means of detecting or ripeness of a particular fruit or vegetable. The
artificially ripened Mangoes. The proposed transformation in color is a key indicator of the
system demonstrates a commendable 91% ripening stage, and various color spaces can be
efficiency in identifying Mangoes subjected to employed for spectral analysis to gain insights into
artificial ripening. this process. Notably, research has demonstrated
that different color spaces yield distinct outcomes in
spectral analysis. Among these, the L*A*B* color
In response to the escalating concerns surrounding space has been identified as particularly effective in
the widespread adoption of artificial ripening the analysis of components. This color space proves
practices in the contemporary fruit market, advantageous in capturing and interpreting the
particularly the use of the carcinogenic substance spectral characteristics associated with the
Calcium Carbide (CaC2), this literature survey maturation of Mangoes and vegetables. In this
explores a novel approach using multi-spectral study, an innovative approach is introduced that
imaging to discern between natural and artificial incorporates artificial neural networks to develop a
ripened Mangoes. The proposed methodology system for predicting the maturity stage of a fruit or
captures spatial and spectral information across vegetable. The RGB color space is utilized as the
eight narrow bands within the visible and near- basis for this prediction model. By harnessing the
infrared wavelength range. The study introduces a capabilities of an artificial neural network, this
meticulously curated dataset comprising multi- research aims to enhance the accuracy and
spectral images of naturally and artificially ripened reliability of forecasting the ripeness of Mangoes
Mangoes, forming the foundation for extensive and vegetables based on their color characteristics.
experiments conducted on a large-scale database. This technological advancement holds promise for
The remarkable outcome reveals an average applications in agriculture and food industries,
classification accuracy of 94.66%, underscoring providing a non-invasive and efficient method for
the effectiveness of multi-spectral imaging as a gauging the maturity levels of produce.
tool for identifying artificially ripened Mangoes.
Additionally, the research delves into the The natural ripening process of Mangoes, driven by
advantages of employing different color spaces for ethylene production, enhances their sweetness and
spectral analysis, with the LAB* color space edibility. However, concerns have arisen about the
proving particularly effective. Furthermore, an artificial ripening of Mangoes, particularly in regions
innovative approach integrating artificial neural like Bangladesh, a major producer of tropical
networks within the RGB color space is Mangoes. While the practical need for effective
introduced, aiming to enhance the accuracy and transportation and distribution has led to the use of
reliability of predicting the maturity stage of chemicals like calcium carbide for accelerated
Mangoes and vegetables. This technological ripening, it poses significant health risks. This
advancement holds promise for non-invasive and prompts a closer examination of artificial ripening
efficient applications in agriculture and the food methods, associated health hazards, and the quest for
industry, offering a robust method for assessing the safer alternatives. The focus is on shedding light on
maturity levels of produce. offering a promising critical aspects and exploring potential remedies to
solution to the prevalent issue of artificial ripening contribute to a broader understanding and address
in the fruit market. By leveraging advancements in the implications of artificial ripening practices. The
imaging technology, this research contributes to way for more informed and health-conscious
ongoing efforts to ensure the quality and safety of approaches in the handling and ripening of Mangoes.
Mangoes reaching consumers, aligning with the
Overall, these findings highlight the importance of
considering the potential health risks associated
with the consumption of artificially ripened
Mangoes. Reliable manner.
III. METHODOLOGY
Detecting artificial fruit ripening involves
leveraging computer vision with tools like
Roboflow and the YOLOv5 object detection model.
The process begins by defining the problem and
collecting a diverse dataset that includes images of
artificially ripened Mangoes. This dataset is then
labeled using Roboflow, with bounding boxes
indicating the regions of interest, such as ripe and
unripe Mangoes. The dataset is split into training,
validation, and test sets, facilitating model
development and evaluation.