Real-Time Pineapple Bud Counting for Yield Estimation with YOLO11 and StrongSORT

View at publisher Summarise with VIRAT
Authors: Pradei Sangkhro, Bidyarani Langpoklakpam, Lithungo K. Murry, R. Kumar
Cited by: 0
References: 14

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Pineapple is a significant crop in Northeast India, but its cultivation remains highly labour-intensive and financially expensive. Reliable and low-cost yield estimation would help farmers with effective planning; however, manual counting is impractical. This paper proposes a real-time pipeline that integrates YOLO11s with StrongSORT to detect and count pineapple buds using drone and handheld footage. The model is trained on field-acquired data, achieving a mAP@0.5 of 88.6%, precision of $84.3 \%$, and recall of $86 \%$ on the test set. The detection-only pipeline operates at a median of $\mathbf{6 1 . 7 ~ F P S}$, while the full detect-track-count pipeline operates at 31.4 FPS, supporting near-edge deployment. Tracking and counting across three test videos yielded accuracies of $96 \%, 73.5 \%$, and $96.2 \%$. Most counting errors stemmed from repeated revisits and tracker re-initialisation rather than detector failure. The proposed system offers a practical and deployable solution for reliable, low-cost pineapple yield estimation.

Fields of Study & Subjects

Smart Agriculture and AI Advanced Neural Network Applications Video Surveillance and Tracking Methods

Publication details

Journal / source
IEEE International Students' Conference on Electrical, Electronics and Computer Science (SCEECS)
Pages
1–6
Publication date
31 January 2026
Language
EN