L’\u00e9quipe ARMEDIA organise un s\u00e9minaire le jeudi 14 d\u00e9cembre 2017 autour du th\u00e8me \u00ab\u00a0deep learning\u00a0\u00bb, ayant pour conf\u00e9rencier invit\u00e9 Nicolas Granger, actuellement en troisi\u00e8me ann\u00e9e de doctorat \u00e0 Telecom SudParis, sous la direction de Mounim A. El Yacoubi.
\nNicolas pr\u00e9sentera ses travaux de recherche sur la reconnaissance de s\u00e9quences de gestes dans un flux vid\u00e9o continu, par des mod\u00e8les statistiques temporels de type HMMs, LSTM\/GRU combin\u00e9s avec des CNN\/MLP, donnant lieu soit \u00e0 des HMMs hybrides, soit \u00e0 des r\u00e9seaux de neurones dits \u00abend-to-end\u00bb.<\/p>\n
Le s\u00e9minaire se d\u00e9roule sur le site de NanoInnov avec une transmission en visioconf\u00e9rence en salle H218 de l’Etoile (sauf perturbations de r\u00e9seau).<\/p>\n
Pour les personnes souhaitant se rendre sur le site de NanoInnov, pensez \u00e0 vous inscrire aupr\u00e8s de Moun\u00eem pour des questions de s\u00e9curit\u00e9 d’acc\u00e8s sur le site, avant Mercredi \u00e0 12h00
\n(mounim.el_yacoubi@telecom-sudparis.eu).<\/p>\n
Le d\u00e9tail du s\u00e9minaire est pr\u00e9cis\u00e9 ci-dessous : Titre: \u00abGesture sequence recognition by hybrid and end-to-end deep learning models\u00bb. L’\u00e9quipe ARMEDIA organise un s\u00e9minaire le jeudi 14 d\u00e9cembre 2017 autour du th\u00e8me \u00ab\u00a0deep learning\u00a0\u00bb, ayant pour conf\u00e9rencier invit\u00e9 Nicolas Granger, actuellement en troisi\u00e8me ann\u00e9e de doctorat \u00e0 Telecom SudParis, sous la direction de Mounim A. El Yacoubi. Nicolas pr\u00e9sentera ses travaux de recherche sur la reconnaissance de s\u00e9quences de gestes dans un flux vid\u00e9o […]<\/p>\n","protected":false},"author":1,"featured_media":975,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"ocean_post_layout":"","ocean_both_sidebars_style":"","ocean_both_sidebars_content_width":0,"ocean_both_sidebars_sidebars_width":0,"ocean_sidebar":"","ocean_second_sidebar":"","ocean_disable_margins":"enable","ocean_add_body_class":"","ocean_shortcode_before_top_bar":"","ocean_shortcode_after_top_bar":"","ocean_shortcode_before_header":"","ocean_shortcode_after_header":"","ocean_has_shortcode":"","ocean_shortcode_after_title":"","ocean_shortcode_before_footer_widgets":"","ocean_shortcode_after_footer_widgets":"","ocean_shortcode_before_footer_bottom":"","ocean_shortcode_after_footer_bottom":"","ocean_display_top_bar":"default","ocean_display_header":"default","ocean_header_style":"","ocean_center_header_left_menu":"","ocean_custom_header_template":"","ocean_custom_logo":0,"ocean_custom_retina_logo":0,"ocean_custom_logo_max_width":0,"ocean_custom_logo_tablet_max_width":0,"ocean_custom_logo_mobile_max_width":0,"ocean_custom_logo_max_height":0,"ocean_custom_logo_tablet_max_height":0,"ocean_custom_logo_mobile_max_height":0,"ocean_header_custom_menu":"","ocean_menu_typo_font_family":"","ocean_menu_typo_font_subset":"","ocean_menu_typo_font_size":0,"ocean_menu_typo_font_size_tablet":0,"ocean_menu_typo_font_size_mobile":0,"ocean_menu_typo_font_size_unit":"px","ocean_menu_typo_font_weight":"","ocean_menu_typo_font_weight_tablet":"","ocean_menu_typo_font_weight_mobile":"","ocean_menu_typo_transform":"","ocean_menu_typo_transform_tablet":"","ocean_menu_typo_transform_mobile":"","ocean_menu_typo_line_height":0,"ocean_menu_typo_line_height_tablet":0,"ocean_menu_typo_line_height_mobile":0,"ocean_menu_typo_line_height_unit":"","ocean_menu_typo_spacing":0,"ocean_menu_typo_spacing_tablet":0,"ocean_menu_typo_spacing_mobile":0,"ocean_menu_typo_spacing_unit":"","ocean_menu_link_color":"","ocean_menu_link_color_hover":"","ocean_menu_link_color_active":"","ocean_menu_link_background":"","ocean_menu_link_hover_background":"","ocean_menu_link_active_background":"","ocean_menu_social_links_bg":"","ocean_menu_social_hover_links_bg":"","ocean_menu_social_links_color":"","ocean_menu_social_hover_links_color":"","ocean_disable_title":"default","ocean_disable_heading":"default","ocean_post_title":"","ocean_post_subheading":"","ocean_post_title_style":"","ocean_post_title_background_color":"","ocean_post_title_background":0,"ocean_post_title_bg_image_position":"","ocean_post_title_bg_image_attachment":"","ocean_post_title_bg_image_repeat":"","ocean_post_title_bg_image_size":"","ocean_post_title_height":0,"ocean_post_title_bg_overlay":0.5,"ocean_post_title_bg_overlay_color":"","ocean_disable_breadcrumbs":"default","ocean_breadcrumbs_color":"","ocean_breadcrumbs_separator_color":"","ocean_breadcrumbs_links_color":"","ocean_breadcrumbs_links_hover_color":"","ocean_display_footer_widgets":"default","ocean_display_footer_bottom":"default","ocean_custom_footer_template":"","ocean_post_oembed":"","ocean_post_self_hosted_media":"","ocean_post_video_embed":"","ocean_link_format":"","ocean_link_format_target":"self","ocean_quote_format":"","ocean_quote_format_link":"post","ocean_gallery_link_images":"on","ocean_gallery_id":[],"footnotes":""},"categories":[332],"tags":[],"class_list":["post-976","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-seminaires-2017-fr","entry","has-media"],"_links":{"self":[{"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/posts\/976","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/comments?post=976"}],"version-history":[{"count":1,"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/posts\/976\/revisions"}],"predecessor-version":[{"id":1568,"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/posts\/976\/revisions\/1568"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/media\/975"}],"wp:attachment":[{"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/media?parent=976"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/categories?post=976"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/tags?post=976"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
\n-Date: Le jeudi 14 d\u00e9cembre 2017 \u00e0 15h.
\n-Lieu: NanoInnov, B\u00e2t 61, salle 2053.<\/strong>
\nIl faudra passer par l\u2019accueil au b\u00e2t 862 pour r\u00e9cup\u00e9rer un badge et appeler ensuite pour que Moun\u00eem vienne vous chercher (01 70 22 02 60 – 06 83 70 12 89).<\/p>\n
\n–
\nPr\u00e9sentation en anglais par Nicolas Granger.
\n–
\nAbstract<\/strong>
\nThe latest developments of Machine Learning on sequence classification, such as speech or hand-writing recognition, have been largely dominated by the use of Neural Networks(NNs). A common approach takes the form of Hybrid Neural-Network\/Hidden Markov Model(HMM) classifiers, where NNs typically shine due to their ability to capture underlying feature representations from low-level input data, while HMMs take care of the temporal modelling. However, with the advances of Recurrent Neural Networks(RNNs) leading to LSTM\/GRU-like architectures, fully neural deep models have now become a de facto standard. These approaches, offering standalone models with end-to-end training, based on the target loss function, have obtained higher accuracy results on several sequence recognition tasks. The objective of this seminar is to give insights on the extent to which these higher accuracies are systematic, and whether there are modeling situations where end-to-end LSTML\/GRU models actually do not outperform HMM\/NN models. To this end, we will analyze the differences between Hybrid NN\/HMM and RNN based approaches, by investigating the impact of the key modelling components of each model.
\nAfter introducing the basics of sequence classification, we will present the experimental test-bed for this comparison, and then develop some of the most notable findings of this comparative study. Our experiments will be based on the Chalearn 2014 dataset, composed of recorded gesture sequences within a continuous video stream. These gestures are challenging as they are high dimensional and highly imbalanced in terms of class distribution, therefore requiring specific training adjustments as we will show.<\/p>\n","protected":false},"excerpt":{"rendered":"