A two-stage 'prompt' pipeline that turns SAR image classifications into text outperforms an end-to-end model for remote sensing visual question answering, and late fusion with optical data gives the best overall accuracy.
Half a percent of labels is enough: Efficient animal detection in UA V imagery using deep CNNs and active learning,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
SAR Strikes Back: A New Hope for RSVQA
A two-stage 'prompt' pipeline that turns SAR image classifications into text outperforms an end-to-end model for remote sensing visual question answering, and late fusion with optical data gives the best overall accuracy.