Saved in:
Bibliographic Details
Main Authors: Guo, Guangyu, Zhang, Dingwen, Han, Longfei, Liu, Nian, Cheng, Ming-Ming, Han, Junwei
Format: Preprint
Published: 2021
Subjects:
Online Access:https://arxiv.org/abs/2112.09532
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866929415236091904
author Guo, Guangyu
Zhang, Dingwen
Han, Longfei
Liu, Nian
Cheng, Ming-Ming
Han, Junwei
author_facet Guo, Guangyu
Zhang, Dingwen
Han, Longfei
Liu, Nian
Cheng, Ming-Ming
Han, Junwei
contents Previous knowledge distillation (KD) methods mostly focus on compressing network architectures, which is not thorough enough in deployment as some costs like transmission bandwidth and imaging equipment are related to the image size. Therefore, we propose Pixel Distillation that extends knowledge distillation into the input level while simultaneously breaking architecture constraints. Such a scheme can achieve flexible cost control for deployment, as it allows the system to adjust both network architecture and image quality according to the overall requirement of resources. Specifically, we first propose an input spatial representation distillation (ISRD) mechanism to transfer spatial knowledge from large images to student's input module, which can facilitate stable knowledge transfer between CNN and ViT. Then, a Teacher-Assistant-Student (TAS) framework is further established to disentangle pixel distillation into the model compression stage and input compression stage, which significantly reduces the overall complexity of pixel distillation and the difficulty of distilling intermediate knowledge. Finally, we adapt pixel distillation to object detection via an aligned feature for preservation (AFP) strategy for TAS, which aligns output dimensions of detectors at each stage by manipulating features and anchors of the assistant. Comprehensive experiments on image classification and object detection demonstrate the effectiveness of our method. Code is available at https://github.com/gyguo/PixelDistillation.
format Preprint
id arxiv_https___arxiv_org_abs_2112_09532
institution arXiv
publishDate 2021
record_format arxiv
spellingShingle Pixel Distillation: A New Knowledge Distillation Scheme for Low-Resolution Image Recognition
Guo, Guangyu
Zhang, Dingwen
Han, Longfei
Liu, Nian
Cheng, Ming-Ming
Han, Junwei
Computer Vision and Pattern Recognition
Machine Learning
Previous knowledge distillation (KD) methods mostly focus on compressing network architectures, which is not thorough enough in deployment as some costs like transmission bandwidth and imaging equipment are related to the image size. Therefore, we propose Pixel Distillation that extends knowledge distillation into the input level while simultaneously breaking architecture constraints. Such a scheme can achieve flexible cost control for deployment, as it allows the system to adjust both network architecture and image quality according to the overall requirement of resources. Specifically, we first propose an input spatial representation distillation (ISRD) mechanism to transfer spatial knowledge from large images to student's input module, which can facilitate stable knowledge transfer between CNN and ViT. Then, a Teacher-Assistant-Student (TAS) framework is further established to disentangle pixel distillation into the model compression stage and input compression stage, which significantly reduces the overall complexity of pixel distillation and the difficulty of distilling intermediate knowledge. Finally, we adapt pixel distillation to object detection via an aligned feature for preservation (AFP) strategy for TAS, which aligns output dimensions of detectors at each stage by manipulating features and anchors of the assistant. Comprehensive experiments on image classification and object detection demonstrate the effectiveness of our method. Code is available at https://github.com/gyguo/PixelDistillation.
title Pixel Distillation: A New Knowledge Distillation Scheme for Low-Resolution Image Recognition
topic Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2112.09532