T-Rex Label

What T-Rex Label is

T-Rex Label is an annotation tool built on zero-shot detection models. You draw a box around one object, the model finds similar objects in that image, and the same prompt can be applied across other images, so a large labelling job starts from a single example rather than a trained model.

What you can do with it

• Label one object and propagate the detection
• Apply a prompt across many images
• Segment and box objects automatically
• Export in formats training pipelines accept
• Work without installing anything

Who it is for

• Computer vision engineers
• Teams building detection datasets
• Agriculture, industry and logistics projects

What to watch out for

• Zero-shot detection is convenient but imperfect; review results before they become training data
• Images you upload may contain people or sensitive locations, so check where they are processed
• Datasets built for surveillance, medical or biometric purposes carry additional legal obligations
• Automated propagation can amplify a single wrong label across a whole batch