ProgDTD：使用双尾端下降训练的渐进学习图像压缩（英文版）

英文标题：ProgDTD: Progressive Learned Image Compression with Double-Tail-Drop Training中文摘要：本文介绍的 ProDTD 是一种训练方法，可将非渐进式图像压缩方法转化为渐进式图像压缩方法，适用于基于 CNN 的学习图像压缩模型，不需要额外的参数，并具有定制化的渐进程度。将 ProDTD 应用于 hyperprior 模型，实验结果显示它在 MS-SSIM 和准确性方面表现出与非渐进式及其他先进渐进式模型相当的性能。英文摘要：Progressive compression allows images to start loading as low-resolutionversions, becoming clearer as more data is received. This increases userexperience when, for example, network connections are slow. Today, mostapproaches for image compressio