[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-128438-en":3,"doc-seo-128438-105":30,"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":4,"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":27,"seo_description":14,"update_tm":28,"read_time":29},128438,962084925290,"Ophelia","https://ap-avatar.wpscdn.com/davatar_085a072bc5b1113ac321206ff7593b45",6,"Technology","Sample Distillation for Object Detection and Image Classification","A novel distillation procedure is proposed to efficiently select informative samples for large-scale learning. Instead of training directly on a very large dataset to achieve state-of-the-art results, the method recursively reduces an initial training set using a criterion that maximizes the information content of the selected subset. Experiments on object detection and image classification show consistent gains over using a uniformly random sub-set, including improved bootstrapping negative mining and better synthetic distortions under artificial perturbations.","CORE  Metadata, citation and similar [papers at core.ac.uk](papers at core.ac.uk)  \nProvided by Infoscience- École polytechnique fédérale de Lausanne  \nJMLR: Workshop and Conference Proceedings 37:1{16, 2014 ACML 2014  \nSample Distillation for Object Detection and Image  \nClassi􀀌cation  \nOlivier Can􀀓evet [olivier.canevet@idiap.ch](olivier.canevet@idiap.ch)  \nLeonidas Lefakis [leonidas.lefakis@idiap.ch](leonidas.lefakis@idiap.ch)  \nFran􀀘cois Fleuret [francois.fleuret@idiap.ch](francois.fleuret@idiap.ch)  \nComputer Vision and Learning group, Idiap Research Institute, Martigny, Switzerland  \n􀀓  \nEcole Polytechnique F􀀓ed􀀓erale de Lausanne, Lausanne, Switzerland  \nEditor: Dinh Phung and Hang Li  \nAbstract  \nWe propose a novel approach to e􀀎ciently select informative samples for large-scale learning. Instead of directly feeding a learning algorithm with a very large amount of samples, as it is usually done to reach state-of-the-art performance, we have developed a \\distillation\" procedure to recursively reduce the size of an initial training set using a criterion that ensures the maximization of the information content of the selected sub-set.  \nWe demonstrate the performance of this procedure for two di􀀋erent computer vision problems. First, we show that distillation can be used to improve the traditional bootstrapping approach to object detection. Second, we apply distillation to a classi􀀌cation problem with arti􀀌cial distortions. We show that in both cases, using the result of a distillation process instead of a random sub-set taken uniformly in the original sample set improves performance signi􀀌cantly.  \nKeywords: Greedy Edge Expectation Maximization (GEEM), Boosting, Object detection, Image classi􀀌cation  \n1. Introduction  \nImage classi􀀌cation and object detection have reached an acceptable level of performance thanks to the use of machine learning techniques with large-scale training sets. In both cases, large amounts of data points can be produced, either through arti􀀌cial distortions of positive samples, or through the bootstrapping of negative samples, the objective being to enrich the available population, either in the neighborhood of positive examples, or at the interface between the positive and the negative population, as characterized by a predictor trained on a limited data-set.  \nIn such cases, the crux of the problem is the training itself, based on these large amounts of data. Most of the learning algorithms have a computation cost at least linear with the number of samples, and are di􀀎cult to parallelize.  \nInterestingly, while most training sets are redundant, very few methods explicitly try to leverage that redundancy to reduce the computational cost. Some well known techniques such as the stochastic gradient descent are explicitly justi􀀌ed through the redundancy of samples, but no method exists that processes the data in the same spirit as a feature-  \n􀀍c 2014 O. Can􀀓evet, L. Lefakis & F. Fleuret.  \nCanvet Lefakis Fleuret  \nselection procedure does with a feature space: By explicitly reducing the cardinality of the space, while keeping the informative content as high as possible.  \nWe propose in this article to exploit a recent machine-learning technique called \\reservoir learning\" by Lefakis and Fleuret (2013) which has been developed precisely to select jointly informative sub-sets of samples. We adapt it to a large-scale context by applying it ina recursive manner, allowing to reduce the overall computation cost by several orders of magnitude, to control its memory footprint, and to parallelize it on a multi-core architecture.  \nWe apply this \\distillation\" procedure to Boosting in two di􀀋erent contexts: Pedestrian detection in natural images, where it allows to improve the set of bootstrapped negative samples, and character recognition, where distillation is used to get a better set of synthetic distortions.  \n2. Related works  \nThough bootstrapping is widely used in computer vision (Shotton et al. , 2005)","cbCaibQDZK2obiag","https://ap.wps.com/l/cbCaibQDZK2obiag","pdf",416606,1,16,"English","en",105,"# Introduction\n## Background and motivation\n## Proposed distillation method\n# Related works\n## Hard negative mining for object detection\n## Background modeling approaches\n## Avoiding hard negatives and limitations\n## Data augmentation and jittering","[{\"question\":\"What problem does the proposed sample distillation method address?\",\"answer\":\"It addresses the training inefficiency caused by large, redundant datasets by selecting a smaller subset that preserves high information content.\"},{\"question\":\"How does distillation improve object detection performance?\",\"answer\":\"It improves bootstrapping by selecting more informative bootstrapped negative samples instead of using a random subset.\"},{\"question\":\"Where else is distillation applied besides object detection?\",\"answer\":\"It is applied to image classification using a classification problem with artificial distortions, yielding a better set of synthetic distortions and improved results.\"}]","Sample Distillation for Object Detection and Image Classification | PDF",1785947699,40,{"code":4,"msg":31,"data":32},"ok",{"site_id":24,"language":23,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"sample-distillation-for-object-detection-and-image-classification","",{"@graph":36,"@context":85},[37,54,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/technology/",3,{"item":52,"name":13,"@type":43,"position":53},"https://docshare.wps.com/document/sample-distillation-for-object-detection-and-image-classification/128438/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":23,"description":14,"dateModified":62,"datePublished":62,"encodingFormat":61,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-08-05",true,{"@type":65,"interactionType":66,"userInteractionCount":4},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"What problem does the proposed sample distillation method address?","Question",{"text":75,"@type":76},"It addresses the training inefficiency caused by large, redundant datasets by selecting a smaller subset that preserves high information content.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does distillation improve object detection performance?",{"text":80,"@type":76},"It improves bootstrapping by selecting more informative bootstrapped negative samples instead of using a random subset.",{"name":82,"@type":73,"acceptedAnswer":83},"Where else is distillation applied besides object detection?",{"text":84,"@type":76},"It is applied to image classification using a classification problem with artificial distortions, yielding a better set of synthetic distortions and improved results.","https://schema.org",{"og:url":52,"og:type":87,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":89,"canonical":52},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,113,117,122,127,130,134],{"id":20,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":111,"slug":112},50,"technology",{"id":114,"doc_module":4,"doc_module_name":46,"category_name":115,"show_sort_weight":29,"slug":116},7,"Healthcare","healthcare",{"id":118,"doc_module":4,"doc_module_name":46,"category_name":119,"show_sort_weight":120,"slug":121},8,"Research & Report",30,"research-report",{"id":123,"doc_module":4,"doc_module_name":46,"category_name":124,"show_sort_weight":125,"slug":126},9,"Religion & Spirituality",20,"religion-spirituality",{"id":125,"doc_module":4,"doc_module_name":46,"category_name":128,"show_sort_weight":125,"slug":129},"World Cup","world-cup",{"id":131,"doc_module":4,"doc_module_name":46,"category_name":132,"show_sort_weight":131,"slug":133},10,"Lifestyle","lifestyle",{"id":135,"doc_module":4,"doc_module_name":46,"category_name":136,"show_sort_weight":106,"slug":137},19,"General","general"]