There are now three crops with image libraries available in that WG Image Library that AgTech startups can use for initial R&D work and prototyping for automation robots. If you’re working with romaine, iceberg, or cabbage, there are high-resolution images tagged and annotated available on Github for you to download and begin using. Also of note, broccoli is in the final stages and will be posted later this week or early next week.
What this means for startups:
A few of the early steps for many specialty crop automation startups are to buy or build an image capture platform, then go out and capture images of the crop you’re working with, then go through the iterative process of optimizing the imaging process to capture the right images with the right light, then pick the images you want to include for R&D purposes (and remove the rest from the data set), then annotate (shorthand – physically tracing the actual products you are capturing so they can be differentiated from other items in the image) all of the selected images and adding tags (what type of crop, other data that helps build better AI).
For the crops in the WG Image Library, you can skip all of the above steps and just go right to the Github link and pull the image library that we’ve built for everyone to use for free. Startups have told us the initial creation can take several months and usually requires engineering time (and that engineering time is kind of pricey). So we built the image library to help save time and money for early stage efforts. Let us know what you think – and please share widely!
This effort is user driven, so if you have crops you want to see added, please email firstname.lastname@example.org and we’ll get them in the queue.
Also, if you’re at a University with an ag department and want to work on the next set of images, we are looking for partners to help. Please email and we’ll get the conversation started.