Papayafreshnet: a hybrid deep learning framework for non-destructive freshness classification of papayas using convolutional and transformer networks

Abstract The freshness of fruits, especially papayas, is essential for consumer satisfaction and marketability. This article introduces PapayaFreshNet, an innovative deep learning network designed for the automatic categorization of papaya freshness through picture analysis. The architecture integra...

Full description

Saved in:
Bibliographic Details
Main Authors: Ayan Sar, Tanupriya Choudhury, Tanmay Sarkar, Ketan Kotecha
Format: Article
Language:English
Published: Springer 2025-04-01
Series:Discover Food
Subjects:
Online Access:https://doi.org/10.1007/s44187-025-00368-9
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849699823584804864
author Ayan Sar
Tanupriya Choudhury
Tanmay Sarkar
Ketan Kotecha
author_facet Ayan Sar
Tanupriya Choudhury
Tanmay Sarkar
Ketan Kotecha
author_sort Ayan Sar
collection DOAJ
description Abstract The freshness of fruits, especially papayas, is essential for consumer satisfaction and marketability. This article introduces PapayaFreshNet, an innovative deep learning network designed for the automatic categorization of papaya freshness through picture analysis. The architecture integrates the advantages of Convolutional Neural Networks (CNNs) with sophisticated Transformer-based feature extraction and attention methods. Utilizing ResNet50 for hierarchical feature extraction, succeeded by a multi-head self-attention Transformer encoder, our model proficiently captures both local and global picture patterns that signify fruit freshness. The model integrates Squeeze-and-Excitation (SE) blocks for channel-wise attention, improving the extraction of the most prominent features. Comprehensive data augmentation methods are employed to reduce overfitting and enhance model generalization. The suggested model attains an impressive training accuracy of 99.65% and a testing accuracy of 97.68%, illustrating its capability in high-precision, real-time freshness classification. The findings indicate that PapayaFreshNet functions as an efficient instrument for non-destructive, automated quality control in agricultural and retail, delivering dependable freshness evaluations to improve supply chain efficacy and minimize waste. Graphical Abstract
format Article
id doaj-art-f09121b77dbc41c48b86aa88b2c092d3
institution DOAJ
issn 2731-4286
language English
publishDate 2025-04-01
publisher Springer
record_format Article
series Discover Food
spelling doaj-art-f09121b77dbc41c48b86aa88b2c092d32025-08-20T03:18:28ZengSpringerDiscover Food2731-42862025-04-015111810.1007/s44187-025-00368-9Papayafreshnet: a hybrid deep learning framework for non-destructive freshness classification of papayas using convolutional and transformer networksAyan Sar0Tanupriya Choudhury1Tanmay Sarkar2Ketan Kotecha3School of Computer Science, University of Petroleum and Energy Studies (UPES)School of Computer Science, University of Petroleum and Energy Studies (UPES)Department of Food Processing Technology, Malda Polytechnic, West Bengal State Council of Technical Education, Government of West BengalSymbiosis Institute of Technology, Symbiosis International (Deemed University)Abstract The freshness of fruits, especially papayas, is essential for consumer satisfaction and marketability. This article introduces PapayaFreshNet, an innovative deep learning network designed for the automatic categorization of papaya freshness through picture analysis. The architecture integrates the advantages of Convolutional Neural Networks (CNNs) with sophisticated Transformer-based feature extraction and attention methods. Utilizing ResNet50 for hierarchical feature extraction, succeeded by a multi-head self-attention Transformer encoder, our model proficiently captures both local and global picture patterns that signify fruit freshness. The model integrates Squeeze-and-Excitation (SE) blocks for channel-wise attention, improving the extraction of the most prominent features. Comprehensive data augmentation methods are employed to reduce overfitting and enhance model generalization. The suggested model attains an impressive training accuracy of 99.65% and a testing accuracy of 97.68%, illustrating its capability in high-precision, real-time freshness classification. The findings indicate that PapayaFreshNet functions as an efficient instrument for non-destructive, automated quality control in agricultural and retail, delivering dependable freshness evaluations to improve supply chain efficacy and minimize waste. Graphical Abstracthttps://doi.org/10.1007/s44187-025-00368-9AttentionFreshness classificationPapayaTransformer NetworksDeep Learning
spellingShingle Ayan Sar
Tanupriya Choudhury
Tanmay Sarkar
Ketan Kotecha
Papayafreshnet: a hybrid deep learning framework for non-destructive freshness classification of papayas using convolutional and transformer networks
Discover Food
Attention
Freshness classification
Papaya
Transformer Networks
Deep Learning
title Papayafreshnet: a hybrid deep learning framework for non-destructive freshness classification of papayas using convolutional and transformer networks
title_full Papayafreshnet: a hybrid deep learning framework for non-destructive freshness classification of papayas using convolutional and transformer networks
title_fullStr Papayafreshnet: a hybrid deep learning framework for non-destructive freshness classification of papayas using convolutional and transformer networks
title_full_unstemmed Papayafreshnet: a hybrid deep learning framework for non-destructive freshness classification of papayas using convolutional and transformer networks
title_short Papayafreshnet: a hybrid deep learning framework for non-destructive freshness classification of papayas using convolutional and transformer networks
title_sort papayafreshnet a hybrid deep learning framework for non destructive freshness classification of papayas using convolutional and transformer networks
topic Attention
Freshness classification
Papaya
Transformer Networks
Deep Learning
url https://doi.org/10.1007/s44187-025-00368-9
work_keys_str_mv AT ayansar papayafreshnetahybriddeeplearningframeworkfornondestructivefreshnessclassificationofpapayasusingconvolutionalandtransformernetworks
AT tanupriyachoudhury papayafreshnetahybriddeeplearningframeworkfornondestructivefreshnessclassificationofpapayasusingconvolutionalandtransformernetworks
AT tanmaysarkar papayafreshnetahybriddeeplearningframeworkfornondestructivefreshnessclassificationofpapayasusingconvolutionalandtransformernetworks
AT ketankotecha papayafreshnetahybriddeeplearningframeworkfornondestructivefreshnessclassificationofpapayasusingconvolutionalandtransformernetworks