Abstract:Deep learning, particularly Convolutional Neural Networks (CNNs), has gained significant attention for its effectiveness in computer vision, especially in agricultural tasks. Recent advancements in instance segmentation have improved image classification accuracy. In this work, we introduce a comprehensive dataset for training neural networks to detect weeds and soy plants through instance segmentation. Our dataset covers various stages of soy growth, offering a chronological perspective on weed invasion's impact, with 1,000 meticulously annotated images. We also provide 6 state of the art models, trained in this dataset, that can understand and detect soy and weed in every stage of the plantation process. By using this dataset for weed and soy segmentation, we achieved a segmentation average precision of 79.1% and an average recall of 69.2% across all plant classes, with the YOLOv8X model. Moreover, the YOLOv8M model attained 78.7% mean average precision (mAp-50) in caruru weed segmentation, 69.7% in grassy weed segmentation, and 90.1% in soy plant segmentation.

What problem does this paper attempt to address?

### What problems does this paper attempt to solve? This paper aims to solve the problem of weed detection in soybean crops, especially accurately identifying and segmenting soybean plants and weeds during different stages of soybean growth (from seedlings to harvest). Specifically, the paper addresses the following challenges by introducing a new dataset named **GrowingSoy**: 1. **Lack of high - quality labeled data**: Most of the existing agricultural datasets only contain labels for image classification, lacking high - quality, high - resolution images and corresponding pixel - level labels required for instance segmentation. This limits the application of deep - learning models in weed - detection tasks. 2. **Lack of the time dimension**: Existing datasets usually do not cover the entire life cycle of crop growth and cannot provide a time - series perspective on the impact of weed invasion on crops. This is crucial for understanding how weeds affect crop growth over time. 3. **Insufficient model validation**: There is a lack of a comprehensive dataset to validate the performance of different deep - learning models in weed - detection tasks, especially during different stages of soybean growth. To solve these problems, the authors constructed a new dataset containing 1,000 high - quality images and provided instance - segmentation labels for each image. These images cover all stages of soybean growth and record the presence of weeds. In addition, the authors also trained and validated this dataset using six state - of - the - art neural - network models, demonstrating the performance of the models in weed and soybean - segmentation tasks. ### Main contributions - **New dataset**: Provided a soybean instance - segmentation dataset containing 1,000 high - resolution images, covering the entire soybean - growth process from seedlings to maturity and recording the invasion of weeds. - **Model validation**: By training and evaluating six state - of - the - art neural - network models (including YOLOv5 and YOLOv8), the validity of the dataset was verified, and the performance of the models in weed and soybean - segmentation tasks at different growth stages was demonstrated. - **Time dimension**: The time - dimension feature of the dataset enables researchers to track changes in weed invasion, thereby better understanding the impact of weeds on crops and formulating more effective management strategies. ### Conclusion This research not only fills the gap in high - quality instance - segmentation datasets in the field of agricultural computer vision but also provides valuable benchmarks and tools for future weed - detection and crop - monitoring.

From Seedling to Harvest: The GrowingSoy Dataset for Weed Detection in Soy Crops via Instance Segmentation

Weed Recognition at Soybean Seedling Stage Based on YOLOV8nGP + NExG Algorithm

Semantic Segmentation for Weed Detection in Corn.

Weed25: A deep learning dataset for weed identification

Efficient Crop Segmentation Net and Novel Weed Detection Method

YOLOWeeds: A novel benchmark of YOLO object detectors for multi-class weed detection in cotton production systems

Deep Residual CNN with Contrast Limited Adaptive Histogram Equalization for Weed Detection in Soybean Crops

Instance segmentation for the fine detection of crop and weed plants by precision agricultural robots

Deep Object Detection of Crop Weeds: Performance of YOLOv7 on a Real Case Dataset from UAV Images

weedNet: Dense Semantic Weed Classification Using Multispectral Images and MAV for Smart Farming

Focus on the Crop Not the Weed: Canola Identification for Precision Weed Management Using Deep Learning

Weed detection in soybean crops using custom lightweight deep learning models

A self-supervised overlapped multiple weed and crop leaf segmentation approach under complex light condition

Weed database development: An updated survey of public weed datasets and cross-season weed detection adaptation

Transferring learned patterns from ground-based field imagery to predict UAV-based imagery for crop and weed semantic segmentation in precision crop farming

Semi-Supervised Weed Detection for Rapid Deployment and Enhanced Efficiency

WeedMap: A Large-Scale Semantic Weed Mapping Framework Using Aerial Multispectral Imaging and Deep Neural Network for Precision Farming

Semantic Segmentation of Crop and Weed using an Encoder-Decoder Network and Image Enhancement Method under Uncontrolled Outdoor Illumination

Development of Weed Detection Method in Soybean Fields Utilizing Improved DeepLabv3+ Platform

Research on weed identification in soybean fields based on the lightweight segmentation model DCSAnet

A powerful image synthesis and semi-supervised learning pipeline for site-specific weed detection