Coarse-to-Fine Domain Incremental Learning with Attentive Distillation for Mining Footprint Segmentation in Multispectral Imagery

Name: Coarse-to-Fine Domain Incremental Learning with Attentive Distillation for Mining Footprint Segmentation in Multispectral Imagery
Uploaded: 2026-10-21T02:20:11.955195+00:00
Duration: 478 s
Description: MineC2FNet is a coarse-to-fine learning method for mining footprint segmentation that uses abundant coarse data to improve fine-grained segmentation. It employs an attentive distillation architecture and a composite loss function to selectively transfer general knowledge from the coarse domain to the fine domain, resulting in superior accuracy and reduced false positives.

MineC2FNet is a coarse-to-fine learning method for mining footprint segmentation that uses abundant coarse data to improve fine-grained segmentation. It employs an attentive distillation architecture and a composite loss function to selectively transfer general knowledge from the coarse domain to the fine domain, resulting in superior accuracy and reduced false positives.

Open the full episode