Abstract: Pre-training a vision-language model and then fine-tuning it on downstream tasks have become a popular paradigm. However, pre-trained vision-language models with the Transformer architecture ...
Abstract: This work explores capabilities of the pre-trained CLIP vision-language model to identify satellite images affected by clouds. Several approaches to using the model to perform cloud presence ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results