Xdecoder 10.5 !link! Online

:

Traditional computer vision pipelines often rely on separate models for detection, segmentation, and classification. XDecoder breaks this siloed approach by employing a unified architecture that treats different tasks as specialized decoding processes. By utilizing a transformer-based decoder, the model can generate masks and semantic labels simultaneously. This, combined with text-guided, promptable queries, allows the system to be highly flexible, adapting to new tasks without retraining. Key Advantages xdecoder 10.5