[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-81982-en":3,"doc-seo-81982-105":29,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":13,"seo_description":14,"update_tm":27,"read_time":28},81982,1099514068035,"Ezra","https://ap-avatar.wpscdn.com/davatar_276721f389ce27ea32af1340a28f341c",6,"Technology","EdgeCompress Coupling Multidimensional Model Compression and Dynamic Inference for EdgeAI","EdgeCompress targets the high computational cost of convolutional neural networks that limits deployment on resource-constrained embedded devices. The framework combines dynamic image cropping with a lightweight foreground predictor to avoid redundant background computation, compound shrinking to jointly compress depth, width, and resolution based on accuracy contribution, and a dynamic inference cascade that selects model complexity per input difficulty. Experiments on ImageNet-1K show 48.8% less computation for ResNet-50 with 0.8% higher top-1 accuracy, and 4.1% accuracy gains with comparable computation versus HRank.","EdgeCompress: Coupling Multidimensional Model Compression and Dynamic Inference for EdgeAI  \nHao Kong , Di Liu , Member, IEEE, Shuo Huai , Xiangzhong Luo , Ravi Subramaniam , Christian  \nMakaya, Qian Lin, Weichen Liu , Member, IEEE  \narXiv :2607 .06982v 1 [ cs .CV] 8 Jul 2026  \nAbstract—Convolutional neural networks (CNNs) have demonstrated encouraging results in image classification tasks. However, the prohibitive computational cost of CNNs hinders the deployment of CNNs onto resource-constrained embedded devices. To address this issue, we propose EdgeCompress, a comprehensive compression framework to reduce the computational overhead of CNNs. In EdgeCompress, we first introduce dynamic image cropping, where we design a lightweight foreground predictor to accurately crop the most informative foreground object of input images for inference, which avoids redundant computation on background regions. Subsequently, we present compound shrinking to collaboratively compress the three dimensions (depth, width, and resolution) of CNNs according to their contribution to accuracy and model computation. Dynamic image cropping and compound shrinking together constitute a multi-dimensional CNN compression framework, which is able to comprehensively reduce the computational redundancy in both input images and neural network architectures, thereby improving the inference efficiency of CNNs. Further, we present a dynamic inference framework to efficiently process input images with different recognition difficulties, where we cascade multiple models with different complexities from our compression framework and dynamically adopt different models for different input images, which further compresses the computational redundancy and improves the inference efficiency of CNNs, facilitating the deployment of advanced CNNs onto embedded hardware. Experiments on ImageNet-1K demonstrate that EdgeCompress reduces the computation of ResNet-50 by 48.8% while improving the top- 1 accuracy by 0.8% . Meanwhile, we improve the accuracy by 4.1% with similar computation compared to HRank. the stateof-the-art compression framework. The source code and models are available at [https://github.com/ntuliuteam/edge-compress](https://github.com/ntuliuteam/edge-compress)  \nIndex Terms—Embedded systems, neural network compression, hardware/software co-design, dynamic neural network  \n© 2023 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works. This is the author’s accepted version of the article published in IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, vol.  \n42, no. 12, pp. 4657-4670, Dec. 2023, DOI: 10.1109/TCAD.2023.3276938 .  \nCorresponding author: Weichen Liu (email: [liu@ntu.edu.sg](liu@ntu.edu.sg))  \nHao Kong, Shuo Huai are with the School of Computer Science and Engineering, Nanyang Technological University (NTU), Singapore, and also with HP-NTU Digital Manufacturing Corporate Lab, Nanyang Technological University (NTU), Singapore.  \nDi Liu is with Department of Computer Science, Norwegian University of Science and Technology (NTNU), Trondheim, Norway.  \nXiangzhong Luo and Weichen Liu are with the School of Computer Science and Engineering, Nanyang Technological University (NTU), Singapore.  \nRavi Subramaniam, Christian Makaya, and Qian Lin are with HP Inc., Palo Alto, CA, USA.  \nFig. 1. The predictions from ResNet-50 . For easy samples, the network can still generate correct predictions at a smaller resolution (e.g., 112 × 112 for ImageNet-1K) . For hard samples, simply resizing images to a smaller resolution can lead to misclassification, while using dynamic cropping can correctly classify hard samples at a sm","cbCaigtjFo6cCcSe","https://ap.wps.com/l/cbCaigtjFo6cCcSe","pdf",7859877,1,15,"English","en",105,"# Introduction\n## Motivation and Problem Overview\n## EdgeAI and Computational Bottlenecks\n## Input Resolution Redundancy\n## Dynamic Image Cropping Idea\n## Network Architecture Compression Directions","[{\"question\":\"What problem does EdgeCompress address in deploying CNNs to EdgeAI devices?\",\"answer\":\"It addresses the prohibitive computation cost of CNNs that prevents running accurate models on resource-constrained embedded devices.\"},{\"question\":\"How does EdgeCompress reduce computation from the input images?\",\"answer\":\"It introduces dynamic image cropping using a lightweight foreground predictor to crop the most informative object, avoiding redundant computation on background regions.\"},{\"question\":\"How does EdgeCompress further reduce computation inside the CNN architecture?\",\"answer\":\"It applies compound shrinking to collaboratively compress depth, width, and resolution according to each dimension’s contribution to accuracy and compute cost.\"}]",1784177407,38,{"code":4,"msg":30,"data":31},"ok",{"site_id":24,"language":23,"slug":32,"title":13,"keywords":33,"description":14,"schema_data":34,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":27},"edgecompress-coupling-multidimensional-model-compression-and-dynamic-inference-for-edgeai","",{"@graph":35,"@context":85},[36,53,68],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,50],{"item":40,"name":41,"@type":42,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":44,"name":45,"@type":42,"position":46},"https://docshare.wps.com/document/","Document",2,{"item":48,"name":12,"@type":42,"position":49},"https://docshare.wps.com/document/technology/",3,{"item":51,"name":13,"@type":42,"position":52},"https://docshare.wps.com/document/edgecompress-coupling-multidimensional-model-compression-and-dynamic-inference-for-edgeai/81982/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-17","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"What problem does EdgeCompress address in deploying CNNs to EdgeAI devices?","Question",{"text":75,"@type":76},"It addresses the prohibitive computation cost of CNNs that prevents running accurate models on resource-constrained embedded devices.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does EdgeCompress reduce computation from the input images?",{"text":80,"@type":76},"It introduces dynamic image cropping using a lightweight foreground predictor to crop the most informative object, avoiding redundant computation on background regions.",{"name":82,"@type":73,"acceptedAnswer":83},"How does EdgeCompress further reduce computation inside the CNN architecture?",{"text":84,"@type":76},"It applies compound shrinking to collaboratively compress depth, width, and resolution according to each dimension’s contribution to accuracy and compute cost.","https://schema.org",{"og:url":51,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,113,118,123,128,131,135],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":45,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":111,"slug":112},50,"technology",{"id":114,"doc_module":4,"doc_module_name":45,"category_name":115,"show_sort_weight":116,"slug":117},7,"Healthcare",40,"healthcare",{"id":119,"doc_module":4,"doc_module_name":45,"category_name":120,"show_sort_weight":121,"slug":122},8,"Research & Report",30,"research-report",{"id":124,"doc_module":4,"doc_module_name":45,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":45,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":45,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":45,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]