Creating a marketplace for pre-used farming equipment using AI


Founded: 2015

AI problem

Multi-label image classification



Read about how our user E-Farm, the first full-transaction online platform for pre-owned farm equipment, uses the Peltarion platform to better organise the equipment sold through their online marketplace.

02/ Reduce, reuse, recycle

Until very recently, buying and selling pre-owned farming equipment was a time-consuming and drawn-out process. As a prospective buyer, you would first have to go to a local supplier and then ask them in turn to make calls to see if anyone knew if any equipment of the kind was being sold. Because of the high purchasing costs (tractors are often priced in the range of €25,000-€50,000), it is also risky to buy things from abroad or places far away, where you aren’t able to inspect the equipment yourself prior to making a purchase.

03/ Creating a marketplace

E-Farm changed this, by creating the first full-transaction online platform for pre-owned farm equipment and solving the quality inspection problem that arises when buying from abroad. With tens of thousands of products on offer and local quality inspection services, they tie together the ecosystem of sellers and buyers through their online platform.

But they were still up against a few crucial problems. First, various modifications are often made to farm equipment after it has been bought; common additions to tractors include for example attaching a front loader, a front connector, or a power take off. Due to E-Farm’s global reach, it was also difficult to make sure that these labels were kept consistent across the platform. So E-Farm approached Peltarion with this problem. They had 10,000 images ready to use and wanted to see whether they could get started with this type of model on the platform. 

04/ Scoping out the problem

The 10,000 images E-Farm wanted to train the model on featured tractors that had up to three types of modifications made to them. The images were labeled using a platform for image tagging, so the whole dataset was consistently labeled. With such a well-defined problem and a large labeled dataset, this was set up with remarkable speed. Within two weeks of contacting the team, they had a well-performing model trained on their own data live and ready to deploy. A farmer wanting to sell his tractor would therefore simply be able to pick up a phone camera and take a few photos, send it into the platform and the system would add the labels automatically before the tractor’s advertisement was made live on the platform. 

05/ Implementation

The model is currently on route to becoming an integral part of E-Farm’s platform. It is expected to go live within just a few weeks. Watch this space for more updates!

Want to replicate this use case in your business? Get in touch

06/ What else can be done