Case Study: Almond achieves adaptable robotic automation with Roboflow Vision AI

Almond boosts robotic picking accuracy to 81% with Roboflow

Almond, a robotics company focused on AI-powered automation for manufacturers, faced the challenge of teaching its robots to perform complex tasks like high-mix picking, where different shaped objects are randomly piled together. This was not feasible with traditional hard-coded robots, which often require expensive facility alterations. To overcome this, Almond turned to vendor Roboflow to develop a sophisticated multi-stage computer vision pipeline using solutions like AI-assisted labeling and hosted model training.

Using Roboflow, Almond built purpose-built vision models, including one trained with the RF-DETR architecture which significantly outperformed a YOLO model, achieving 81% accuracy. Roboflow’s platform streamlined the process of labeling thousands of images and training state-of-the-art models. The solution allows Almond's robots to run on a mix of local and cloud inference, resulting in highly adaptable automation that works immediately without costly facility modifications.

Almond