M. Szczepanski

Graph-Guided Token Merging (G2TM) is extended, a single module inserted early in a ViT-based network, beyond its original Segmenter setting, and its behavior and accuracy-efficiency trade-off are consistent across every tested architecture for a given backbone size, indicating that its effectiveness is a property of the encoder rather than the decoder.

Achieving fully integer-only inference for lightweight detection transformers could revolutionize their deployment in resource-constrained environments.