Active Learning Strategies Compared for YOLOv8 on Lincolnbeet

TLDR:

Machine learning applications in agriculture have an incredible impact on water, herbicides, pesticides, and fertilizer usage! In many cases, a reduction of up to 50% is possible by using ML for precision spraying.

Excited to share the impressive results on active learning in agriculture that Igor Susmelj just published!

Leveraged active learning to supercharge our YOLOv8 model for lincolnbeet (sugarbeet and weed detection). Boosted mean Average Precision (mAP) by up to 14.6x compared to random image labeling. Active learning reduced annotation costs by up to 77% - perfect for optimizing your annotation budget!

It's incredible how data selection by using active learning can make such a huge impact!