{"id":36019,"date":"2025-10-14T14:27:23","date_gmt":"2025-10-14T12:27:23","guid":{"rendered":"https:\/\/www.graviton.at\/letterswaplibrary\/iam-looking-for-human3-6m-but-official-cite-is-not-respond-for-3-weeks\/"},"modified":"2025-10-14T14:27:23","modified_gmt":"2025-10-14T12:27:23","slug":"iam-looking-for-human3-6m-but-official-cite-is-not-respond-for-3-weeks","status":"publish","type":"post","link":"https:\/\/www.graviton.at\/letterswaplibrary\/iam-looking-for-human3-6m-but-official-cite-is-not-respond-for-3-weeks\/","title":{"rendered":"I&#8217;am Looking For Human3.6m, But Official Cite Is Not Respond For 3 Weeks"},"content":{"rendered":"<p><!-- SC_OFF --><\/p>\n<div class=\"md\">\n<h3>\u2753[HELP] 4D-Humans \/ HMR2.0 Human3.6M eval images missing \u2014 can\u2019t find official dataset<\/h3>\n<p>I\u2019m trying to reproduce <strong>HMR2.0 \/ 4D-Humans<\/strong> evaluation on <strong>Human3.6M<\/strong>, using the official config and <code>h36m_val_p2.npz<\/code>.<\/p>\n<p>Training runs fine, and 3DPW evaluation works correctly \u2014<br \/> but <strong>H36M eval completely fails<\/strong> (black crops, sky-high errors).<\/p>\n<p>After digging through the data, it turns out the problem isn\u2019t the code \u2014<br \/> it\u2019s that the <code>h36m_val_p2.npz<\/code> expects full-resolution images (~1000\u00d71000)<br \/> with names like:<\/p>\n<p>&#8220;`<\/p>\n<p>S9_Directions_1.60457274_000001.jpg<\/p>\n<p>&#8220;`<\/p>\n<p>But there\u2019s <strong>no public dataset<\/strong> that matches both naming and resolution:<\/p>\n<table>\n<thead>\n<tr>\n<th>Source<\/th>\n<th>Resolution<\/th>\n<th>Filename pattern<\/th>\n<th>Matches npz?<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>HuggingFace \u201cHuman3.6M_hf_extracted\u201d<\/td>\n<td>256\u00d7256<\/td>\n<td><code>S11_Directions.55011271_000001.jpg<\/code><\/td>\n<td>\u2705 name, \u274c resolution<\/td>\n<\/tr>\n<tr>\n<td>MKS0601 3DMPPE<\/td>\n<td>1000\u00d71000<\/td>\n<td><code>s_01_act_02_subact_01_ca_01_000001.jpg<\/code><\/td>\n<td>\u2705 resolution, \u274c name<\/td>\n<\/tr>\n<tr>\n<td>4D-Humans auto-downloaded <code>h36m-train\/*.tar<\/code><\/td>\n<td>1000\u00d71000<\/td>\n<td><code>S1_Directions_1_54138969_001076.jpg<\/code><\/td>\n<td>close, but <code>_<\/code> vs <code>.<\/code> mismatch<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p>So the official evaluation <code>.npz<\/code> points to a Human3.6M image set that doesn\u2019t seem to exist publicly. The repo doesn\u2019t provide a download script for it, and even the HuggingFace or MKS0601 versions don\u2019t match.<\/p>\n<hr \/>\n<h3>My question<\/h3>\n<p>Has anyone successfully run HMR2.0 or 4D-Humans <strong>H36M evaluation<\/strong> recently?<\/p>\n<ul>\n<li>Where can we <strong>download the official full-resolution images<\/strong> that match <code>h36m_val_p2.npz<\/code>?<\/li>\n<li>Or can someone confirm the <strong>exact naming \/ folder structure<\/strong> used by the authors?<\/li>\n<\/ul>\n<p>I\u2019ve already <strong>registered on the official Human3.6M website<\/strong> and requested dataset access,<br \/> but it\u2019s been <strong>weeks with no approval or response<\/strong>, and I\u2019m stuck. <\/p>\n<p>Would appreciate any help or confirmation from anyone who managed to get the proper eval set.<\/p>\n<\/div>\n<p><!-- SC_ON -->   submitted by   <a href=\"https:\/\/www.reddit.com\/user\/Last_Raise4834\"> \/u\/Last_Raise4834 <\/a> <br \/> <span><a href=\"https:\/\/www.reddit.com\/r\/datasets\/comments\/1nyhfz6\/iam_looking_for_human36m_but_official_cite_is_not\/\">[link]<\/a><\/span>   <span><a href=\"https:\/\/www.reddit.com\/r\/datasets\/comments\/1nyhfz6\/iam_looking_for_human36m_but_official_cite_is_not\/\">[comments]<\/a><\/span><\/p><div class='watch-action'><div class='watch-position align-right'><div class='action-like'><a class='lbg-style1 like-36019 jlk' href='javascript:void(0)' data-task='like' data-post_id='36019' data-nonce='65e0e39b87' rel='nofollow'><img class='wti-pixel' src='https:\/\/www.graviton.at\/letterswaplibrary\/wp-content\/plugins\/wti-like-post\/images\/pixel.gif' title='Like' \/><span class='lc-36019 lc'>0<\/span><\/a><\/div><\/div> <div class='status-36019 status align-right'><\/div><\/div><div class='wti-clear'><\/div>","protected":false},"excerpt":{"rendered":"<p>\u2753[HELP] 4D-Humans \/ HMR2.0 Human3.6M eval images missing \u2014 can\u2019t find official dataset I\u2019m trying to reproduce&#8230;<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[85],"tags":[],"class_list":["post-36019","post","type-post","status-publish","format-standard","hentry","category-datatards","wpcat-85-id"],"_links":{"self":[{"href":"https:\/\/www.graviton.at\/letterswaplibrary\/wp-json\/wp\/v2\/posts\/36019","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.graviton.at\/letterswaplibrary\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.graviton.at\/letterswaplibrary\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.graviton.at\/letterswaplibrary\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.graviton.at\/letterswaplibrary\/wp-json\/wp\/v2\/comments?post=36019"}],"version-history":[{"count":0,"href":"https:\/\/www.graviton.at\/letterswaplibrary\/wp-json\/wp\/v2\/posts\/36019\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.graviton.at\/letterswaplibrary\/wp-json\/wp\/v2\/media?parent=36019"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.graviton.at\/letterswaplibrary\/wp-json\/wp\/v2\/categories?post=36019"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.graviton.at\/letterswaplibrary\/wp-json\/wp\/v2\/tags?post=36019"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}