Abstract
This paper presents the development of a semantic segmentation pipeline designed for autonomous navigation in agricultural environments. Inspired by the NATERIDA microbots project—an AI-powered exploration robot capable of environmental monitoring and intelligent data collection—we implement a U-Net convolutional neural network trained on automatically generated pixel-level annotations. The system uses GroundingDINO for open-vocabulary object detection and the Segment Anything Model (SAM) for mask generation, eliminating the need for manual annotation. Data is collected through a simulated Pioneer 3-AT robot performing random-walk exploration in a Webots physics simulator. The model segments scenes into five terrain classes: soil, vegetation, obstacles, pedestrians, and sky. After training for 100 epochs on an NVIDIA GeForce RTX 4060 Laptop GPU with Albumentations-based data augmentation, the system achieves a validation Intersection-over-Union (IoU) of 0.4453 and a Dice coefficient of 0.5497. While these results indicate moderate performance, they demonstrate a viable pipeline for low-cost dataset creation and segmentation model training without manual labeling effort.
Files
Software
- Language
- Python
- Development status
- Research prototype
References
- Prasiddha Mainali, Prashanna Tiwari, and Prince Kumar Mandal, “NATERIDA microbots,” Devpost, Aug. 2025. Available: https://devpost.com/software/naterida-microbots-i4bv8r
- Olaf Ronneberger, Philipp Fischer, and Thomas Brox, “U-Net: Convolutional networks for biomedical image segmentation,” in Proc. MICCAI, 2015, pp. 234–241.
- Jonathan Long, Evan Shelhamer, and Trevor Darrell, “Fully convolutional networks for semantic segmentation,” in Proc. CVPR, 2015, pp. 3431–3440.
- Liang-Chieh Chen, Yukun Zhu, George Papandreou, Florian Schroff, and Hartwig Adam, “Encoder-decoder with atrous separable convolution for semantic image segmentation,” in Proc. ECCV, 2018, pp. 801–818.
- Enze Xie, Wenhai Wang, Zhiding Yu, Anima Anandkumar, Jose M. Alvarez, and Ping Luo, “SegFormer: Simple and efficient design for semantic segmentation with transformers,” in Proc. NeurIPS, 2021, pp. 12077–12090.
- Shilong Liu, Zhaoyang Zeng, Tianhe Ren, Feng Li, Hao Zhang, et al., “Grounding DINO: Marrying DINO with grounded pre-training for open-set object detection,” in Proc. ECCV, 2024.
- Alexander Kirillov, Eric Mintun, Nikhila Ravi, Hanzi Mao, Chloe Rolland, et al., “Segment Anything,” in Proc. ICCV, 2023, pp. 4015–4026.
- Olivier Michel, “Webots: Professional mobile robot simulation,” International Journal of Advanced Robotic Systems, vol. 1, no. 1, pp. 39–42, 2004.