Lab Automation Cameras, with Vision Models Trained on Your Lab

Lab Automation Cameras AI Vision Solution

Liquid handlers, sample sorters, and lab robots repeat the same steps all day, and many of them assume every tube is in its slot, every cap is off, and every tip box is full. A camera with a vision model checks those assumptions before and during a run: tubes and caps in racks, labware placement on the deck, tip presence, liquid levels, colony counts on plates, and the labels on every sample. The right camera depends on where it goes.

• Benchtop and sample intake: Roboflow AI1 (opens in a new tab), an 8 MP camera with an 8-LED ring light and an NVIDIA Jetson Orin NX, mounts above a bench or deck, runs detection and segmentation models on the camera, and sends results over MQTT, OPC UA, or Modbus TCP
• Inside an instrument: compact USB or GigE cameras from The Imaging Source or Basler paired with an NVIDIA Jetson, or Luxonis OAK cameras that run the model on the camera itself
• Sample labels: label reading cameras that read the printed labels on tubes and plates

Whichever camera you choose, the model comes from your own images. Label photos of your racks, plates, and labware in Roboflow, train a model such as RF-DETR, and deploy it with Roboflow Inference. When you add new labware or a new assay, add examples and retrain.

‍

Already have cameras on your instruments? Roboflow Inference also runs models on GigE Vision, USB, and RTSP streams from the cameras you already have, so you can add vision to the instruments and stations you run today.