Transformers for detection

Query-to-Attention

Mechanism where object queries guide the model's attention to relevant regions of the image, unlike global attention, improving prediction efficiency and specialization.

← Tillbaka