[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-31261":3,"doc-seo-31261":26},{"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,"file_id":15,"file_url":16,"file_type":17,"file_size":18,"view_count":19,"is_deleted":4,"is_public":19,"is_downloadable":19,"audit_status":19,"page_count":20,"language":21,"table_of_contents":22,"faqs":23,"seo_title":13,"seo_description":14,"update_tm":24,"read_time":25},31261,687197207057,"Sage","https://ap-avatar.wpscdn.com/davatar_29158cc5080c5b710cf443261637dec0",6,"Technology","Optimized embedded AI efficient implementation of CNNs on ESP32-CAM","An lightweight embedded artificial intelligence system enables real-time image classification on ultra-low-cost microcontrollers using the ESP32-CAM platform. The approach integrates a quantized convolutional neural network trained on Fashion-MNIST, reaching 92.3% accuracy. System-level engineering addresses severe constraints (520 KB SRAM, no FPU, limited stack) through solutions for TensorFlow Lite buffer overflows, quantized memory misalignment, and real-time task arbitration among AI inference, camera, and WiFi. A responsive web interface provides live streaming and on-device predictions, delivering near higher-end accuracy at nearly six times lower cost and nearly four times lower power for edge robotics and TinyML.","cbCaid2wFT52vf5a","https://ap.wps.com/l/cbCaid2wFT52vf5a","pdf",1364453,1,17,"English","# Abstract\n# Introduction\n## Problem and motivation\n## Proposed lightweight CNN implementation\n## System-level engineering contributions","[{\"question\":\"What embedded AI system is presented, and what is its target task?\",\"answer\":\"The paper presents a lightweight embedded AI system for real-time image classification on the ESP32-CAM microcontroller.\"},{\"question\":\"How is the CNN model optimized to run under hardware constraints?\",\"answer\":\"It uses full-integer quantization, weight pruning, and efficient memory management to support real-time inference within tight SRAM/stack limits.\"},{\"question\":\"What engineering challenges are solved to make on-device inference reliable?\",\"answer\":\"The work addresses TensorFlow Lite buffer overflows, quantized memory misalignment, and real-time scheduling between AI inference, camera capture, and WiFi communication.\"}]",1779224583,43,{"code":4,"msg":27,"data":28},"ok",{"site_id":29,"language":30,"slug":31,"title":13,"keywords":32,"description":14,"schema_data":33,"social_meta":85,"head_meta":87,"extra_data":89,"updated_unix":24},105,"en","optimized-embedded-ai-efficient-implementation-of-cnns-on-esp32-cam","",{"@graph":34,"@context":84},[35,52,67],{"@type":36,"itemListElement":37},"BreadcrumbList",[38,42,46,49],{"item":39,"name":40,"@type":41,"position":19},"https://docshare.wps.com","Home","ListItem",{"item":43,"name":44,"@type":41,"position":45},"https://docshare.wps.com/document/","Document",2,{"item":47,"name":12,"@type":41,"position":48},"https://docshare.wps.com/document/technology/",3,{"item":50,"name":13,"@type":41,"position":51},"https://docshare.wps.com/document/optimized-embedded-ai-efficient-implementation-of-cnns-on-esp32-cam/31261/",4,{"url":50,"name":13,"@type":53,"author":54,"headline":13,"publisher":56,"fileFormat":59,"description":14,"dateModified":60,"datePublished":61,"encodingFormat":59,"isAccessibleForFree":62,"interactionStatistic":63},"DigitalDocument",{"name":9,"@type":55},"Person",{"url":39,"name":57,"@type":58},"DocShare","Organization","application/pdf","2026-05-21","2026-05-19",true,{"@type":64,"interactionType":65,"userInteractionCount":19},"InteractionCounter",{"@type":66},"ViewAction",{"@type":68,"mainEntity":69},"FAQPage",[70,76,80],{"name":71,"@type":72,"acceptedAnswer":73},"What embedded AI system is presented, and what is its target task?","Question",{"text":74,"@type":75},"The paper presents a lightweight embedded AI system for real-time image classification on the ESP32-CAM microcontroller.","Answer",{"name":77,"@type":72,"acceptedAnswer":78},"How is the CNN model optimized to run under hardware constraints?",{"text":79,"@type":75},"It uses full-integer quantization, weight pruning, and efficient memory management to support real-time inference within tight SRAM/stack limits.",{"name":81,"@type":72,"acceptedAnswer":82},"What engineering challenges are solved to make on-device inference reliable?",{"text":83,"@type":75},"The work addresses TensorFlow Lite buffer overflows, quantized memory misalignment, and real-time scheduling between AI inference, camera capture, and WiFi communication.","https://schema.org",{"og:url":50,"og:type":86,"og:title":13,"og:site_name":57,"og:description":14},"article",{"robots":88,"canonical":50},"index,follow",{"doc_id":7,"site_id":29}]