)]}'
{
  "commit": "3606b78108f8db4b8cfc2ae7fed8de34be8d21e2",
  "tree": "3884be5f71d0ec24c5d1fb7ab62e5b247af5355e",
  "parents": [
    "effd974b16d4ac8bdb2415b25ba39b09a0e280bf"
  ],
  "author": {
    "name": "paulwilkins",
    "email": "paulwilkins@google.com",
    "time": "Thu Apr 16 11:15:36 2015 +0100"
  },
  "committer": {
    "name": "paulwilkins",
    "email": "paulwilkins@google.com",
    "time": "Tue Apr 21 12:50:11 2015 +0100"
  },
  "message": "Modified test for auto key frame detection.\n\nThe existing test was triggering a lot of false positives on some types\nof animated material with very plain backgrounds. These were triggering\ncode designed to catch key frames in letter box format clips.\n\nThis patch tightens up the criteria and imposes a minimum requirement\non the % blocks coded intra in the first pass and the ratio between the\n% coded intra and the modified inter % after discounting neutral (flat)\nblocks that are coded equally well either way.\n\nOn a particular problem animation clip this change eliminated a large\nnumber of false positives including some cases where the old code\nselected kf several times in a row. Marginal false negatives are less\ndamaging typically to compression and in the problem clip there are now\na couple of cases where \"visual\" scene cuts are ignored because of well\ncorrelated content across the scene cut.\n\nReplaced some magic numbers related to this with #defines and added\nexplanatory comments.\n\nChange-Id: Ia3d304ac60eb7e4323e3817eaf83b4752cd63ecf\n",
  "tree_diff": [
    {
      "type": "modify",
      "old_id": "8103abc83997e335c992e28daaf2a16bec91a961",
      "old_mode": 33188,
      "old_path": "vp9/encoder/vp9_firstpass.c",
      "new_id": "456dabe7504f277121b73061974ca4ebe67cd012",
      "new_mode": 33188,
      "new_path": "vp9/encoder/vp9_firstpass.c"
    }
  ]
}