Detasseling: Male & Female Tassel Detection

Project overview

The Detasseling project involved annotating tassels in corn fields to identify male and female tassels based on their position relative to the corn rows. The main objective was to identify female tassels that remained after the detasseling process and to quantify the remaining tassels that needed to be removed.

Tassels located within the designated row were considered male tassels, while tassels appearing outside the row were considered female tassels. The annotation focused particularly on the remaining female tassels after detasseling, because these indicate tassels that were not successfully removed and may require further attention.

Work performed

• Reviewed corn field images and identified tassels based on their position within or outside the crop rows.
• Tassels inside the row were treated as male.
• Tassels outside the row were treated as female.
• Annotated the remaining female tassels after the detasseling process.
• Focused on identifying female tassels that were missed or left behind during detasseling.
• Created accurate bounding boxes around the relevant tassels.
• Consistent classification was maintained across different field conditions.
• Reviewed annotations to minimise missed or incorrectly classified tassels.
• Prepared the dataset for computer vision applications related to detasseling quality assessment.

What made it hard

• Determining correctly whether a tassel was inside or outside the designated row.
• Distinguishing male and female tassels based on their position in the field.
• Small or partially visible remaining female tassels were hard to identify.
• Handling dense corn plants and overlapping tassels.
• Ensuring that remaining female tassels were not missed during annotation.
• Consistent annotations had to be maintained across different row layouts and image conditions.

Result

A high-quality annotated dataset was prepared to support detasseling quality assessment, with particular focus on identifying remaining female tassels and reducing missed tassels in corn fields.

Final deliverable

The final dataset contained consistent annotations for male and female tassels, with emphasis on identifying remaining female tassels after detasseling. The dataset was prepared for computer vision model training and automated detasseling quality monitoring.