Abstract
Deep learning models have grown increasingly complex, with input data sizes scaling accordingly. Despite substantial advances in specialized deep learning hardware, data loading continues to be a major bottleneck that limits training and inference speed. To address this challenge, we propose training models directly on encoded JPEG features, reducing the computational overhead associated with full JPEG decoding and significantly improving data loading efficiency. While prior works have focused on recognition tasks, we investigate the effectiveness of this approach for the restoration task of single-image super-resolution (SISR). We present a lightweight super-resolution pipeline that operates on JPEG discrete cosine transform (DCT) coefficients in the frequency domain. Our pipeline achieves a 2.6x speedup in data loading and a 2.5x speedup in training, while preserving visual quality comparable to standard SISR approaches.
Abstract (translated)
深度学习模型变得越来越复杂，输入数据的大小也随之增加。尽管在专门用于深度学习的硬件方面取得了重大进展，但数据加载仍然是限制训练和推理速度的主要瓶颈。为了解决这一挑战，我们提出了一种直接在编码JPEG特征上训练模型的方法，这减少了完全解码JPEG所带来的计算开销，并显著提高了数据加载效率。虽然先前的研究主要集中在识别任务上，但我们研究了这种方法对单图像超分辨率（SISR）恢复任务的有效性。我们介绍了一个轻量级的超分辨率管道，该管道操作于JPEG离散余弦变换（DCT）系数在频域中。我们的管道实现了数据加载速度提高2.6倍和训练速度提高2.5倍，并且保持了与标准SISR方法相当的视觉质量。
URL
https://arxiv.org/abs/2512.04284
