Skip to content
Open access

Object Detection for Pepper Harvesting: A Comparative Evaluation of Backbone-Adapted, Contemporary, and Open-Vocabulary Models

Oct 2026 · Agriculture · Vol 16, pp. 2243 · 0 citations · 64 references

TL;DR

Three complementary object-detection paradigms for multiclass pepper harvesting using 1004 RGB images annotated for human, pepper, pepper group, box, and pushcart are investigated, demonstrating that pretrained visual–semantic representations can transfer effectively to selected categories in field-acquired agricultural scenes, although performance remained strongly class-dependent.

Abstract

Reliable object detection is essential for agricultural perception under illumination changes, occlusion, and complex backgrounds. This study investigates three complementary object-detection paradigms for multiclass pepper harvesting using 1004 RGB images annotated for human, pepper, pepper group, box, and pushcart. First, Faster R-CNN, SSD, and YOLOv7 were evaluated with alternative CNN backbones to examine the effect of feature extraction and training configuration. Second, the contemporary YOLO12n, YOLO26n, and RT-DETR-L detectors were evaluated under domain-specific closed-set training. Finally, YOLOE-26n was assessed under a zero-shot open-vocabulary setting without fine-tuning on the pepper-harvesting dataset. Among the backbone-adapted models, YOLOv7–GoogLeNet achieved 99.6% test mAP@0.5. Across five independent training seeds, YOLO12n achieved the highest mean performance among the contemporary detectors with 86.01±0.53% mAP@0.5 and 71.00±0.62% mAP@0.5:0.95, while YOLO26n provided the most favorable efficiency profile at 47.06 FPS with 2.38 million parameters. Despite receiving no pepper-domain training, YOLOE-26n achieved 41.89% mAP@0.5 and reached 92.43% AP@0.5 for human, demonstrating that pretrained visual–semantic representations can transfer effectively to selected categories in field-acquired agricultural scenes, although performance remained strongly class-dependent. Quantitative and qualitative analyses further highlighted differences in object recovery, localization, computational efficiency, and failure modes among the evaluated paradigms. Overall, the study provides a common experimental reference for understanding the trade-offs among backbone adaptation, contemporary closed-set detection, and zero-shot semantic recognition in pepper-harvesting perception. Because the three paradigms followed different training and benchmarking protocols, their results are interpreted within their respective experimental settings rather than as a strict cross-paradigm ranking.

Read PDF

Similar papers

Sep 2026

Selective Cotton Boll Localization for Robotic Harvesting: Evaluation of Deep Learning Vision Models Under Field Conditions

This study developed and evaluated a deep-learning-based perception framework for selective robotic cotton picking. The dataset contained 1,008 annotated field images collected using three cameras under varying natural lighting and weather conditions. Object-detection models from the YOLOv8 through YOLOv13 families wer...

Thevathayarajh Thayananthan, Xin Zhang, Isuru Laddusinghe Badu et al. · 0 citations
Open access Nov 2026

Lightweight visual detection framework for complex background grape leaf disease detection

Accurate crop disease detection is essential for precision agriculture, yet reliable visual identification remains challenging under complex backgrounds, varying illumination, and limited computational resources. This study proposes a lightweight grape leaf disease detection framework, ULR-YOLO, based on YOLO11n. The f...

Chen-Yu Zhao, Ming-Yu Bai, Haoriqin Wang et al. · 0 citations
Open access Sep 2026

Multi-scale attention-based feature learning for jujube fruit bruise detection and variety classification

Automated fruit classification plays a central role in post-harvest quality assessment, particularly in reducing dependence on manual inspection within smart agriculture systems. In this work, a deep learning framework, MSA-TNet, is introduced to address jujube fruit classification across two related tasks, binary br...

Yonis Gulzar, Sharyar Wani, B. Nath et al. · 0 citations
Conference Open access Sep 2026

Leaf Diseases and Weed Detection using Image Processing: A Comprehensive Survey

The comprehensive survey investigates all the image processing systems and deep learning architectures from the conventional machine vision system to the latest generation of CNN architectures like ResNet, EfficientNet, VGG, MobileNet, DenseNet and YOLO for the detection of various plant leaf diseases and weed identifi...

Sriya Mohanasundaran, C. Arun · 0 citations
Open access Oct 2026

Modern Approaches to Visual Content Detection Using YOLO, RF-DETR, and CNN Models

The obtained results confirm that transformer-based object detectors provide higher detection accuracy and stronger generalization capability than convolution-based approaches for complex destructive visual content, making the proposed system a promising solution for automated content moderation and information securit...

Sh. Temirgaziyeva, A. Turarbek, A. S. Dokuz et al. · 0 citations

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.