Flymingo Computer Vision System Improves Warehouse Performance
Flymingo is an Israeli computer vision company that identifies mistakes in supply chain operational processes from existing warehouse camera images. It is the only supply chain software company known to industry analyst firm ARC Advisory Group that has AI Vision at the core of its solution. The company uses AI for image recognition in combination with a warehouse management system. The solution is used to drive compliance with standard operating procedures in warehouses. SOPs are critical for safety and reliability.
Reinforcement learning applied to warehouses
Reinforcement learning is a type of machine learning that allows AI models to improve their decision-making process based on positive, neutral, and negative feedback. For example, to train a computer system to recognize images of dogs, we first use humans to look at tens of thousands of images of animals. The humans label the images as dog, not dog, or unclear. Then, those images are shown to a computer, and the system responds by saying, “This is a dog” or “This is not a dog.” The dog-recognition algorithm then receives feedback that can be either positive (yes, you’re right, this is a dog), negative (no, this was not a dog), or neutral (we’re not sure if this is a dog). As the system repeats the feedback loop, it gets better at correctly identifying images.
In Flymingo’s solution, security cameras in the warehouse capture images of objects such as pallets, trucks, cases, people and staging areas into the system. The system identifies the type of object being viewed. By combining images, data from the WMS and contextual rules, the system can recognize whether processes are being adhered to.
How do WMS and computer vision systems work together?
Warehouse management systems are a category of application software that supports receiving, storage, picking, shipping, and value-added services. WMS solutions rely on automatic identification systems (often RF scanning devices that read barcodes) to record real-time status changes.
A WMS can achieve near perfect picking and inventory accuracy if SOPs are followed. In the warehouse, every slot in the distribution center has a barcode label. Barcode labels can also be placed on cases, pallets, and individual stock keeping units. For example, the WMS instructs a floor associate to go to location AX32 and pick two cases. The associate goes to the location, scans the barcode label on the slot to verify that they are in the correct location, and then scans the barcode labels on the two cases, proving that the correct number of cases were picked. Using a WMS in conjunction with AutoID can achieve inventory accuracy of over 99.9%.
A WMS can achieve much greater accuracy than a paper-based system, but to achieve that level of accuracy, it is essential that workers aren’t distracted and follow SOPs. For example, if a worker is going to the correct slot and a coworker walks by and starts a small conversation, they may accidentally pick up a case from the wrong slot when they return to the rack. Or, a worker may be instructed to apply a shipping label when a pallet is sitting in a designated spot near the dock. The worker clicks a button on an RF device to start the label printer. The worker must go and pick up the label, return to the pallet, and apply the label to the pallet. But if they decide to take a bathroom break before going to the label machine, another worker may pick up the label.
Flymingo’s vision system can work in conjunction with the WMS to identify when an error is likely to have occurred. In the example of applying shipper labels to staged pallets, the vision system can be set to flag a potential error if a worker doesn’t return to the pallet’s location within three minutes. The worker’s manager is alerted to the specific type of mistake and can view the footage to determine if a mistake was, in fact, made. The manager can then provide rapid vision-based coaching. Near real-time feedback improves employee performance.
Not all errors stem from mistakes. Some employees intentionally violate SOPs. For example, your warehouse may have performance targets, and employees may be instructed to go to a specific location to pick up two cases. The employee may choose to make up time by not going to that location. Instead, they may use their mobile device to indicate that the slot is out of stock, falsely claiming that they missed the pick-up. Or, a thief at a shipping dock may intentionally put a pallet in an employee’s truck. In these cases, visual intelligence can be used to detect untrustworthy employees.
It’s an impressive technology, but does it really work?
Abaline Chief Operating Officer Avi Boas spoke at the Made4net user conference Inspire 2024. Abaline is a privately owned, family-owned dealership with a fleet of 25 trucks. Abaline’s main distribution center is a 165,000-square-foot facility in Bayonne, New Jersey. The dealership also serves healthcare facilities, educational institutions, and other sectors.
Abaline uses the Made4net WMS solution. By implementing Flymingo on top of their WMS, they were able to significantly improve process adherence. Boas says that before implementing Flymingo, they might not have realized for weeks that they had shipped the wrong medical supplies to an overseas customer, at which point he and his shift managers would sometimes spend hours poring over stored security footage to determine what had gone wrong. When managers told employees about the mistake, they would often get defensive and blame the mistake on something outside of their control.
Now, mistakes are identified in near real time, and when an employee is approached, they are often able to identify the mistake they made before their manager has even spoken to them about it. In addition to improved employee relations, customer service has also improved due to improved process adherence.
