TAEHV is a Tiny AutoEncoder for Hunyuan Video (and other similar video models). TAEHV can encode and decode latents into videos more cheaply (in time & memory) than the full-size video VAEs, at the ...
Abstract: Hyperspectral unmixing (HU) aims to decompose mixed pixels into their constituent spectral signatures and estimate their corresponding fractional abundances. Recently, the nonlinear spectral ...
Abstract: Deep learning methods exhibit limited performance in few-shot specific emitter identification tasks due to the scarcity of labeled training samples. To address this challenge, this letter ...