-
Upsampling Strategies for Dense Mapping
A deep technical comparison of nearest neighbor, bilinear interpolation, and transposed convolution for dense prediction networks.
-
VAE-Inspired Diffusion Revolution
This blog explains how diffusion models overcome VAE blurriness by generating images through gradual denoising, while also showing the conceptual connection to VAE-style latent modeling. It compares the stability and generative strengths of diffusion models with traditional approaches and highlights why their stepwise reverse process is so powerful.
