Models / nvidia / SegFormer B4 (ADE20K)

SegFormer B4 (ADE20K)

JSON →
vision

A semantic segmentation model based on the SegFormer architecture with a MiT-B4 backbone, fine-tuned on the ADE20K dataset at 512x512 resolution.

imagevision
Specs
Context & limits
context window
max output
Lifecycle
Dates
released2021-06-01
Resources