Stories from our clients

The company owns orange plantation in Sumatera island in Indonesia where farmers are growing local orange products called Medan oranges. Their brand is called "Darifarm" oranges. They face difficulty of conducting manual observation of orange quality while the fruits are hanging on the tree. Farmers need to estimate the right time to harvest by knowing the grade of oranges during its first 6 months. However due to low accuracy, the number is often different from tonnage after harvest. 

The solution we developed for Agrari was computer vision-based using Object Detection model to classify oranges based on its color and smoothness (smooth, mild, and severe bronzing) and the model was trained on photographs taken by farmers from their mobile phone cameras. The detection turns this information into percentage of tonnage based on its grades and sizes that is useful for farmers. 

Company profile: https://agrari.id/en

BIG is a software company that helps its clients to implement digital transformation. One of its focus is involved in agriculture use cases such as oil palm agribusiness. Oil palm agribusiness is a major business player in Indonesia's renewable energy sector. Challenge in the business operation is having a standard opertion procedure to calculate the yield of palm oil fresh fruit bunch (FFB) and classify the ripeness of the fruits for selection before going to the mill. 

To meet with these needs, we developed an object detection model that can detect and classify oil palm fruit in 2 situations; fruits hanging on tree and fruits collected on the ground. Our model is capable to classify multiple levels of ripeness, which is useful for yield and quality report before transported to the oil palm mills. 

Company profile: https://berjaya-inovasi.com/

Being competitive among geoscience software providers for side scan sonar data processing is the ambition of this company. They decided to extend their software called "WADUGS" by adding AI to do automatic seafloor mapping, a very time-consuming interpretation process if done manually. Boulders are objects that are spread on the seafloor and needs to be mapped for geohazard identification such as for installation of wind turbines in the North Sea. The requirement posed is unique since the data is geospatial.

We developed a computer vision-based object detection model which is able to detect 10,000+ of boulders on the seafloor using side-scan sonar survey dataset. The model is super-fast and robust which takes only few seconds to run on massive area of survey instead of days or weeks of manual interpretation by experts. 

Company profile: https://wadugs.com/