Awesome

ECCV2024-AdpatICMH

[ECCV2024] Image Compression for Machine and Human Vision With Spatial-Frequency Adaptation

paper link.

Overview

The overall framework of AdpatICMH

Absctract

Image compression for machine and human vision (ICMH) has gained increasing attention in recent years. Existing ICMH methods are limited by high training and storage overheads due to heavy design of task-specific networks. To address this issue, in this paper, we develop a novel lightweight adapter-based tuning framework for ICMH, named Adapt-ICMH, that better balances task performance and bitrates with reduced overheads. We propose a spatial-frequency modulation adapter (SFMA) that simultaneously eliminates non-semantic redundancy with a spatial modulation adapter, and enhances task-relevant frequency components and suppresses task-irrelevant frequency components with a frequency modulation adapter. The proposed adapter is plug-and-play and compatible with almost all existing learned image compression models without compromising the performance of pre-trained models. Experiments demonstrate that Adapt-ICMH consistently outperforms existing ICMH frameworks on various machine vision tasks with fewer fine-tuned parameters and reduced computational complexity.

Install

git clone https://github.com/qingshi9974/ECCV2024-AdpatICMH
pip install compressai
pip install timm tqdm click

Install Detectron2 for object detection and instance segementation.

Dataset

The following datasets are used and needed to be downloaded.

• ImageNet1K
• COCO 2017 Train/Val
• Kodak

Example Usage

Specify the data paths, target rate point, corresponding lambda, and checkpoint in the config file accordingly.

Classification

python examples/classification.py -c config/classification.yaml<br> Add argument -T for evaluation.

Object Detection

python examples/detection.py -c config/detection.yaml<br> Add argument -T for evaluation.

Instance Segmentation

python examples/segmentation.py -c config/segmentation.yaml<br> Add argument -T for evaluation.

Pre-trained Weights for TIC

TODO

• Release the Pre-trained Weights for TIC-SFMA, mbt2018mean-SFMA, cheng2020anchor-SFMA
• Add the code for mbt2018mean and cheng2020anchor

Citation

If you find our project useful, please cite the following paper.

@inproceedings{li2024image,
  title={Image compression for machine and human vision with spatial-frequency adaptation},
  author={Li, Han and Li, Shaohui and Ding, Shuangrui and Dai, Wenrui and Cao, Maida and Li, Chenglin and Zou, Junni and Xiong, Hongkai},
  booktitle={European Conference on Computer Vision},
  year={2024}
}

Ackownledgement

Our work is based on the framework of CompressAI and TransTIC.