Agrisynth

Over 30 years ago, the concept of Precision Agriculture was born. Farming has made advances, moving away from treating whole fields but the dream of farming at the organism level (crop plant, weed, pest or disease lesion) remains elusive. To operate at this level of precision, we need to recognise everything in a crop scene image at a pixel level. Artificial Intelligence (AI) solutions are far from achieving this, preventing us from achieving organism-level farming. We haven’t yet been able to train agricultural AI solutions to this level for two reasons. First, we have difficulties gathering Real World images in sufficient numbers and quality to form a relevant training dataset. Second, we have never been able to annotate (label) all objects in those images at a pixel level to train a robust AI solution. AgriSynth is breaking down this barrier by producing AI solutions that classify every object at pixel-level in complex crop scene images. We will do this by training AI solutions with synthetic image datasets. They are annotated to 100% accuracy at pixel-level, even at higher resolutions that we can only dream of today. We can generate millions of such images if required. We solve a major bottleneck within agriculture for robotic companies who enter niche markets where they can generate commercially acceptable AI solutions. We also enable researchers to accurately assess field trials that currently rely on subjective assessments.