Grape FD Leaf Detection

Project overview

The Grape FD Leaves Detection project involved annotating grapevine images to identify leaves affected by Flavescence doree (FD) and leaves suspected of showing FD symptoms. The dataset was prepared for computer vision model training to detect and distinguish FD-affected grape leaves, and it stands as one of the most challenging and important projects undertaken.

Work performed

• Reviewed grapevine images and identified leaves showing FD symptoms.
• Created accurate bounding box annotations around FD-affected leaves.
• Leaves showing suspected FD symptoms were labelled using the susp_FD_leaves class.
• Differentiated target leaves from healthy leaves and surrounding grapevine vegetation.
• Consistent class labelling was maintained throughout the dataset.
• Reviewed annotations and corrected missing or inaccurate bounding boxes.
• Prepared the dataset for FD leaf detection and classification model training.

What made it hard

• Differentiating FD-affected leaves from healthy grape leaves.
• Leaves with subtle or unclear symptoms were hard to identify.
• Distinguishing confirmed FD leaves from suspected FD leaves.
• Handling overlapping leaves and dense grapevine vegetation.
• Partially visible leaves had to be annotated accurately.
• Maintaining consistency between the FD_leaves and susp_FD_leaves classes.

Result

A high-quality annotated dataset was prepared for grapevine FD leaf detection, with separate classes for FD-affected and suspected FD leaves.

Final deliverable

The final dataset contained accurate and consistent bounding box annotations for the FD_leaves and susp_FD_leaves classes, making it suitable for computer vision model training and grapevine disease monitoring applications.