Disclaimer: I'm redirecting efforts to pyglide and may be slow to address bugs here. I also recommend looking at @crowsonkb's v-diffusion-pytorch. See captions and more generations in the Gallery.

Abstract: Intensity inhomogeneity, hidden details, poor image contrast due to low capturing device quality, limited user experience, and inappropriate environment setting during data acquisition are ...

Model based image search (i.e. using machine learning based models trained on image similarity rather than traditional Lucene based search on captions) Unsupervised or self supervised, because ...

Abstract: Domain adaptation is crucial in aerial imagery, as the visual representation of these images can significantly vary based on factors such as geographic location, time, and weather conditions ...