Academic Journal of Engineering and Technology Science, 2026, 9(2); doi: 10.25236/AJETS.2026.090207.
Sean Gu1, Chi Shing Lin2
1Beanstalk International Bilingual School Shunyi Campus, Beijing, China
2Shenzhen College of International Education, Shenzhen, China
This study addresses the low adoption of carbon footprint assessment for non-standard products (e.g., local foods, customized goods) due to the lack of standardized databases and accounting models. Against the backdrop of global low-carbon trends and corporate compliance needs, the research aims to overcome challenges such as high costs and poor scalability in traditional methods. An integrated intelligent system was developed, combining hardware (electronic scales, cameras) with AI technologies: Baidu OCR for text extraction, a SentenceTransformer-based vector database for querying carbon factors, and the Deepseek large language model for semantic reasoning and calculation. A Gradio-based interface enables end-to-end operation. Tests showed text recognition accuracy of 90.2% (net content 95.0%, ingredient list 92.5%). Carbon footprint estimates had an average absolute error of 4.7% (range: 1.2%–7.3%) against expert benchmarks. The system enables rapid, manual-modeling-free on-site assessment, generating standardized reports and tailored reduction strategies. This research fills a market gap by offering a universal estimation framework for non-standard products. It demonstrates the potential of combining OCR, vector databases, and LLMs for efficient carbon factor inference, while edge AI integration enhances practicality. Applicable to green procurement, supply chain management, and ESG reporting, the system empowers SMEs to conduct carbon accounting cost-effectively. It also supports the broader adoption of carbon labeling in food, retail, and manufacturing, driving green supply chain development and corporate low-carbon transition with significant social and commercial value.
Non-standard products; Carbon footprint accounting; AI empowerment; OCR recognition; Large language model; Vector database
Sean Gu, Chi Shing Lin. Carbon Footprint Manager: An AI-assisted Carbon Footprint Calculation System for Non-standard Products. Academic Journal of Engineering and Technology Science (2026), Vol. 9, Issue 2: 55-62. https://doi.org/10.25236/AJETS.2026.090207.
[1] Zhu Liyang. Accelerating the Construction of Waste Recycling System to Assist in Achieving the Goals of Carbon Peak and Carbon Neutrality - Policy Interpretation No. 4 [J]. Resource Recycling, 2024, (02): 56-57.
[2] Shu Yinbiao, Zhao Liang, Yi Jun, et al. Methodological System and Technology Research on Product Carbon Footprint Accounting [J/OL]. Proceedings of the CSEE, 1-12 [2026-02-26]. https://link.cnki.net/urlid/11.2107.tm.20260121.1826.008.
[3] Wilting C H ,Veld ' I D .An overview of accounting approaches for national greenhouse gas emissions; Comparing Consumption and Production Footprints in the European Union [J]. Economic Systems Research, 2025, 37(3): 402-426.
[4] Xie Y, Zhou B, Wang Z, et al. Industrial Carbon Footprint (ICF) Calculation Approach Based on Bayesian Cross-Validation Improved Cyclic Stacking [J]. Sustainability, 2023, 15(19).
[5] Zhang Y, Yao S, Chen Y. Developing a carbon footprint calculation method for product life cycle based on low-carbon design: A case study of the STAGE Bluetooth speaker [J]. PLoS one, 2025, 20(8): e0327576.
[6] Feng J , Jiang Y , Xu J ,et al.Photonic-assisted division calculation of microwave signal envelopes and its applications[J].Optics Express, 2024, 32(24):43692.DOI:10.1364/oe.542473.
[7] Wang X C , Li W K , Zhang Y T .Deep Neural Network-Assisted Efficient Calculation on CO2 Pulsed Discharges Under Martian Pressure for Various Pulse Rise Rates[J].IEEE Transactions on Plasma Science, 2024(5):52.DOI:10.1109/TPS.2024.3404074.
[8] Zhang Y , Liu X , Wang W .Theoretical Calculation Assisted by Machine Learning Accelerate Optimal Electrocatalyst Finding for Hydrogen Evolution Reaction[J].ChemElectroChem, 2024, 11(13). DOI:10.1002/celc.202400084.