Overview

• Introduction
• Object Classes
  • whd

Introduction

This dataset is focused on the precise identification and annotation of wheat heads in agricultural fields. The dataset consists of images capturing wheat heads during different growth stages. The primary goal is to assist in the development of algorithms for the automatic detection and analysis of wheat crops.

Object Classes

whd

Description

The "whd" class represents wheat heads. These are elongated structures with a series of seeds arranged along a central axis. Mature wheat heads are typically visible in the plant amidst surrounding leaves and stems. They exhibit a distinctive texture with a dense formation of individual grains.

Instructions

• Annotate all visible wheat heads: Draw bounding boxes around each distinct wheat head. Include the full length of the head from where the seeds start to where they end.

• Partial visibility: If a wheat head is partially visible due to occlusion by leaves or other plants, estimate its full size based on the visible portion and annotate accordingly.

• Clusters and overlaps: For clusters of overlapping wheat heads, try to distinguish and individually annotate each head as long as they are visually separable.

• Do not annotate leaves or other plant structures that are not part of the wheat head itself.

• Size consideration: Do not annotate any wheat head that is less than 10 pixels in length in the image to ensure annotations focus on clearly visible instances.

By following these instructions, annotators will consistently identify and mark wheat heads across the dataset, aiding in the dataset's utility for machine learning applications.