EatSafe-Fake Product Detection and Nutrition Advisor

Abstract

Food fraud and poor nutritional awareness are two problems that directly affect what people eat and how safe those choices are. Packaged goods frequently slip past quality checks, and for most shoppers, decoding a nutrition label is harder than it should be. This paper looks at where current research stands on both fronts: detecting counterfeit products and helping consumers understand what they are actually buying. We survey methods built on deep learning and image processing, covering logo recognition, OCR-based licence number validation, and brand name matching on the authenticity side, and ingredient analysis with health scoring on the nutrition side. Across these methods we found a consistent pattern — each solution tackles one problem in isolation, and none of them connect authenticity checking with nutritional guidance in a single tool. Real-world datasets are also scarce, and most systems struggle with pack-aging that does not follow a clean, standard layout. EatSafe is our attempt to close that gap. It brings together CNN-based logo detection, OCR-driven label reading, and a personalised health scoring engine so that a consumer can scan a product and find out both whether it is genuine and whether it is actually good for them.

Introduction

Packaged foods have become increasingly popular due to their convenience, but they present two major challenges: counterfeit products that imitate trusted brands and pose health risks, and complex nutritional labels that are difficult for consumers to understand. Existing solutions typically address either food authenticity or nutritional analysis, but not both simultaneously. To overcome this limitation, the study proposes EatSafe, an AI-powered framework that verifies product authenticity while providing personalized nutritional guidance in a single system.

The literature review reveals that research has focused on three separate areas: product authenticity verification, nutritional analysis, and Optical Character Recognition (OCR). Authentication methods have evolved from traditional image processing and machine learning to deep learning models such as SSD and physical security techniques like Copy Detection Patterns (CDPs). While these methods effectively identify counterfeit products, they do not evaluate nutritional quality. Conversely, nutritional analysis systems use OCR and Natural Language Processing (NLP) to extract ingredient information, identify allergens, explain additives, or estimate nutritional content, but they assume that the product being analyzed is genuine. OCR serves as the common enabling technology for both tasks, yet no existing solution integrates authenticity verification and nutritional assessment into one pipeline.

EatSafe addresses this gap through a sequential workflow. Users capture an image of a packaged food product, which is preprocessed through resizing, noise reduction, contrast enhancement, and perspective correction to improve detection accuracy. The authenticity verification module then uses a MobileNetV2-based convolutional neural network to detect and compare brand logos, applies OCR to extract the brand name and FSSAI license number, and validates this information against a product database and the official FSSAI registry. Products that fail any verification step are flagged as suspicious before nutritional analysis proceeds.

For genuine products, the nutritional analysis module extracts key nutritional values, including sugar, fat, sodium, protein, and calories, using OCR. A personalized health score ranging from 1 to 10 is calculated using a weighted scoring model that considers dietary guidelines and user-specific health conditions such as diabetes or hypertension. Products with low scores trigger health warnings, and the system recommends healthier alternatives within the same product category.

The final output provides four key pieces of information: the product's authenticity status with supporting evidence, a complete nutritional breakdown, an easy-to-understand health score explanation, and personalized dietary recommendations based on the user's health profile.

Experimental evaluation was conducted using a dataset of 200 product images (160 for training and 40 for testing). The system achieved 87.5% accuracy in logo detection, 82.5% accuracy in OCR-based brand name recognition, 90% accuracy in FSSAI license validation, 85% accuracy in nutritional information extraction, and 88% accuracy in health score prediction. The results demonstrate that combining AI-based computer vision, OCR, and nutritional analysis into a unified framework enables reliable counterfeit detection while helping consumers make healthier food choices. Overall, EatSafe provides a practical, user-friendly solution that enhances consumer safety, improves nutritional awareness, and supports informed purchasing decisions.

Conclusion

This paper presented EatSafe, a unified pipeline that ad-dresses two problems consumers face with packaged food — whether a product is genuine and whether it is nutritionally sound — from a single product photograph. Existing tools handle one or the other but never both together, and that gap is what EatSafe was built to fill. The system combines CNN-based logo detection, OCR-driven label reading, and a personalised health scoring engine in one coherent framework. On a test set of 40 images drawn from a 200-image dataset, the authenticity module achieved 87.5% accuracy and nutrition ex-traction reached 85.0%, with OCR performance on blurred or curved labels identified as the clearest area for improvement. The dataset size means these figures should be read as a proof of concept rather than a production benchmark, and expanding to at least 2,000 images across diverse packaging types is the most important next step. With broader data coverage, mobile optimisation, and integration with live FSSAI registry APIs, EatSafe has a realistic path toward becoming a practical in-store tool that helps consumers make safer and better-informed food choices every day.

References

[1] J. Tutt, O. Taran, R. Chaban, B. Pulfer, Y. Belousov, T. Holotyak, and S. Voloshynovskiy, ”Authentication of Copy Detection Patterns: A Pattern Reliability Based Approach,” IEEE Transactions on Information Forensics and Security, vol. 19, pp. 3124-3135, 2024. [2] A.R. Venodhani, Y. Asina Begum, E. Priya Darshini, G. Naresh kumar Reddy, G. Hanisha, and P. Sai Pavithra, ”Automated Detection of Fake Brand Logos using a Multi-Stage Image Analysis and Machine Learning Approach,” in Proc. 3rd Int. Conf. Self Sustainable Artificial Intelligence Systems (ICSSAS), 2025, pp. 1410-1416. [3] B. Wang, T. Bu, Z. Hu, L. Yang, Y. Zhao, and X. Li, ”Coarse-to-Fine Nutrition Prediction,” IEEE Transactions on Multimedia, vol. 26, pp. 3651-3663, 2024. [4] T. Castro et al., ”Comparison of Healthiness, Labelling, and Price between Private and Branded Label Packaged Foods in New Zealand (2015-2019),” Nutrients, vol. 13, no. 8, p. 2731, 2021.. [5] M. M. H. Sohan, I. Nanda, M. M. Khan, and R. Dey, ”Fake Product Review Detection Using Machine Learning,” in Proc. IEEE World AI IoT Congress (AIIoT), 2022, pp. 527-532 [6] S. S. P, S. A. M, S. G, S. V, S. Cloudin, and S. Manjunath, ”REAL TIME FAKE LOGO DETECTION,” in Proc. Int. Conf. Circuit Power and Computing Technologies (ICCPCT), 2023, pp. 878-885. [7] R. Jana, A. R. Chowdhury, and M. Islam, ”Optical Character Recogni tion from Text Image,” International Journal of Computer Applications Technology and Research, vol. 3, no. 4, pp. 239-243, 2014. [8] N. Prabha, A. M. S, A. G. V, R. M, and T. M. B, ”AI Ingredient Analyzer- Nutriknow,” International Journal of Innovative Science and Research Technology, vol. 10, no. 4, pp. 2227-2232, 2025. [9] M. Chakraborty and K. Biswas, ”Limit of Detection for Five Common Adulterants in Milk: A Study With Different Fat Percent,” IEEE Sensors Journal, vol. 18, no. 6, pp. 2395-2403, 2018. [10] M. Chimerad, P. Borjian, P. Pathak, and H. J. Cho, ”Self-Powered Flexible Edible UV Sensors Based on Sustainable Materials Applicable for Food Packaging,” IEEE Sensors Letters, vol. 8, no. 10, Art. no. 6013304, 2024. [11] van der Bend DLM, van Eijsden M, van Roost MHI, de Graaf K, Rood enburg AJC. The Nutri-Score algorithm: Evaluation of its validation pro cess. Front Nutr. 2022 Aug 15;9:974003. doi: 10.3389/fnut.2022.974003. PMID: 36046131; PMCID: PMC9421047. [12] V. C P, A. D, D. D. Kedilaya, S. S. Gondkar and S. Halhalli, ”Online Analysis of Ingredient Safety, Leveraging OCR and Machine Learning for Enhanced Consumer Product Safety,” 2024 2nd International Con ference on Artificial Intelligence and Machine Learning Applications Theme: Healthcare and Internet of Things (AIMLA), Namakkal, India, 2024, pp. 1-6, doi: 10.1109/AIMLA59606.2024.10531558. [13] S. Janokar, K. More, R. Pundir, R. Mhamulkar, S. Piraji and P. Sangvikar, ”Food Ingredient Analysis Integrating OCR and LLM for Enhanced Consumer Safety,” 2025 5th International Conference on Soft Computing for Security Applications (ICSCSA), Salem, India, 2025, pp. 594-599, doi: 10.1109/ICSCSA66339.2025.11171116. [14] S. K, A. R, C. R and N. R, ”AI-Powered Ingredient Detector for Allergies: Enhancing Food Safety Through Natural Language Processing and Computer Vision,” 2025 International Conference on Computing and Communication Technologies (ICCCT), Chennai, India, 2025, pp. 1-5, doi: 10.1109/ICCCT63501.2025.11019718. [15] K. Cooper, A. Ankrum, M. A. Clarke, J. B. Boron, J. Ponce and E. Bass, ”Measures of Reading Ease and NOVA Food Processing Classification on Ingredients Lists in the United States,” 2024 IEEE 12th International Conference on Healthcare Informatics (ICHI), Orlando, FL, USA, 2024. [16] J. N. P. Markavathi, J. G. R, A. K. S, B. Roshan and A. Bala, ”NutriScan: A Mobile Application for Health-Conscious Food Choices,” 2025 IEEE International Students’ Conference on Electrical, Electronics and Computer Science (SCEECS), Bhopal, India, 2025. [17] T. Silva, T. Carvalho, V. Filipe, L. Gonc¸alves and A. Sousa, ”Nutritional Insight: Using OCR to Decode Food Labels for Better Health,” 2024 International Conference on Graphics and Interaction (ICGI), Vila Real, Portugal, 2024, pp. 1-8, doi: 10.1109/ICGI64003.2024.10923764. [18] D. Pandey, I. Kapoor, I. Singh and V. K. Sharma, ”Food Label Analysis and Ingredient Safety,” 2025 3rd International Conference on Commu nication, Security, and Artificial Intelligence (ICCSAI), Greater Noida, India, 2025, pp. 690-694, doi: 10.1109/ICCSAI64074.2025.11063934. [19] B. Natarajan, S. Harini, M. S. Fathima and S. Subhashini, ”Nutrition and Food Suggestions for Healthcare using Chatbot,” 2024 4th International Conference on Soft Computing for Security Applications (ICSCSA), Salem, India, 2024, pp. 15-20, doi: 10.1109/ICSCSA64454.2024.00010. [20] S. Vaishnavi, C. Thanujashree, J. Sri Aishwarya and N. Nandhini, ”Nutrition Recommendation System for Sports Persons using Random Forest Algorithm,” 2024 International Conference on Smart Systems for Electrical, Electronics, Communication and Computer Engineering

Copyright

Copyright © 2026 Hrushikesh Lomte, Ayush Kumkar, Om Jagtap, Nilesh Rokade, Ms. Roshani Gawade. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.