AquaProbe: Efficient Adaptation of Frozen Vision Models for Underwater Images
Overview
This paper introduces AquaProbe, a method for adapting frozen vision models to underwater imagery by measuring domain gaps based on water type and imaging range. Using paired clean and physically rendered images, the approach allocates a fixed LoRA budget according to the measured divergence profile. The proportional LoRA rank allocation achieves a segmentation mean Intersection-over-Union (mIoU) of 70.1, outperforming uniform allocation by 1.9 points with the same parameter budget of 0.59 million adapter parameters. A three-way router conditioned on estimated condition clusters further achieves 70.6 mIoU on the standard SUIM split and 63.4 AUDRC under held-out degradation.
Generated from attributed reports · Updated 2 hours ago
Event evidence and corrections
0 attributed source owners. Ownership does not establish independent confirmation. Quantities are reported separately and are never added together.
Reported quantity · mIoU: 70.1 other · Basis not reported
“ion-over-Union (mIoU) for the proportional LoRA rank allocation, based on the measured profile, is 70.1; this is a better result than uniform LoRA rank allocation by 1.9 mIoU points while having the same parameter budget of 0.59 million adapter parameters. The three-way router, conditioned on estimated condition clusters, achieves 70.6 mIoU on the standard SUIM split and 63.4 AUDRC under held-out degradation.”
Exact source · revision 1Source owner not reported
Reported quantity · mIoU improvement: 1.9 other · Basis not reported
“ion-over-Union (mIoU) for the proportional LoRA rank allocation, based on the measured profile, is 70.1; this is a better result than uniform LoRA rank allocation by 1.9 mIoU points while having the same parameter budget of 0.59 million adapter parameters. The three-way router, conditioned on estimated condition clusters, achieves 70.6 mIoU on the standard SUIM split and 63.4 AUDRC under held-out degradation.”
Exact source · revision 1Source owner not reported
Reported quantity · mIoU on SUIM split: 70.6 other · Basis not reported
“ion-over-Union (mIoU) for the proportional LoRA rank allocation, based on the measured profile, is 70.1; this is a better result than uniform LoRA rank allocation by 1.9 mIoU points while having the same parameter budget of 0.59 million adapter parameters. The three-way router, conditioned on estimated condition clusters, achieves 70.6 mIoU on the standard SUIM split and 63.4 AUDRC under held-out degradation.”
Exact source · revision 1Source owner not reported
Reported quantity · AUDRC score: 63.4 other · Basis not reported
“ion-over-Union (mIoU) for the proportional LoRA rank allocation, based on the measured profile, is 70.1; this is a better result than uniform LoRA rank allocation by 1.9 mIoU points while having the same parameter budget of 0.59 million adapter parameters. The three-way router, conditioned on estimated condition clusters, achieves 70.6 mIoU on the standard SUIM split and 63.4 AUDRC under held-out degradation.”
Exact source · revision 1Source owner not reported
Report timeline
Follow attributed reports and material updates.
- arXiv Robotics — research abstractsSignalAquaProbe: Efficient Adaptation of Frozen Vision Models for Underwater Images
This paper introduces AquaProbe, a method for adapting frozen vision models to underwater imagery by measuring domain gaps based on water type and imaging range. Using paired clean and physically rendered images, the approach allocates a fixed LoRA budget according to the measured divergence profile. The proportional LoRA rank allocation achieves a segmentation mean Intersection-over-Union (mIoU) of 70.1, outperforming uniform allocation by 1.9 points with the same parameter budget of 0.59 million adapter parameters. A three-way router conditioned on estimated condition clusters further achieves 70.6 mIoU on the standard SUIM split and 63.4 AUDRC under held-out degradation.
Event coverage history
There is not enough continuous observation data to show a trend.
Timezone · UTC
Article dates follow your selected timezone. Briefing editions use Hong Kong time (UTC+8).