Commits

Branch selector

 master

User selector

All users

Datepicker

All time

Commit History

Commits on Dec 16, 2024

• Fix a typo in a reference to psbody-mesh package.

  radekd91

  committed
  6167d7f

  Copy full SHA for 6167d7f

Commits on Dec 6, 2024

• Disable confusing print out of stack when spectre not available.

  radekd91

  committed
  f9e30ee

  Copy full SHA for f9e30ee

• Fix a crash in video face reconstruction demo.

  radekd91

  committed
  1b08738

  Copy full SHA for 1b08738

Commits on Nov 11, 2024

• Add missing file

  radekd91

  committed
  5e7673a

  Copy full SHA for 5e7673a

• Update README

  radekd91

  committed
  e91d76b

  Copy full SHA for e91d76b

Commits on Mar 28, 2024

• Update video resampling script

  radekd91

  committed
  75f8f76

  Copy full SHA for 75f8f76

Commits on Mar 24, 2024

• MEAD processing scripts and readme

  radekd91

  committed
  53beb5d

  Copy full SHA for 53beb5d

• Make data preloading robust towards repeated sample indices passed to the dataset (only load one time)

  radekd91

  committed
  be206ac

  Copy full SHA for be206ac

• Refactoring

  radekd91

  committed
  8f577c5

  Copy full SHA for 8f577c5

• Refactoring in positional encoding

  radekd91

  committed
  1d91a26

  Copy full SHA for 1d91a26

• Bug fix in rotation losses

  radekd91

  committed
  28f063d

  Copy full SHA for 28f063d

• Update requirements

  radekd91

  committed
  cbf9247

  Copy full SHA for cbf9247

Commits on Feb 2, 2024

• Update psbody-mesh dependency to a private fork that is robust towards OpenGL failure. It won't support viewer but at least the Mesh class will work

  radekd91

  committed
  8829f9e

  Copy full SHA for 8829f9e

• Hotfix that should fix the loading issues.

  radekd91

  committed
  1fe9aa0

  Copy full SHA for 1fe9aa0

Commits on Jan 25, 2024

• Update gitignore

  radekd91

  committed
  669249f

  Copy full SHA for 669249f

• Update docker startup scripts

  radekd91

  committed
  81299e5

  Copy full SHA for 81299e5

• Experimental pytorch2 install script

  radekd91

  committed
  178fdbc

  Copy full SHA for 178fdbc

• Add the option to not load segmentations along with the video

  radekd91

  committed
  dd92e49

  Copy full SHA for dd92e49

• Fix typo

  radekd91

  committed
  4b9ede3

  Copy full SHA for 4b9ede3

• Fix typo

  radekd91

  committed
  571e0e6

  Copy full SHA for 571e0e6

• Fix typo

  radekd91

  committed
  12f77ab

  Copy full SHA for 12f77ab

• Fix typo

  radekd91

  committed
  1a1d0f3

  Copy full SHA for 1a1d0f3

• Add user argument to docker building script

  radekd91

  committed
  74ce26d

  Copy full SHA for 74ce26d

• Minor fixes

  radekd91

  committed
  efa04d0

  Copy full SHA for efa04d0

• Minor fixes

  radekd91

  committed
  43692d0

  Copy full SHA for 43692d0

• Faster reconstruction rendering with pytorch3d. Decouple aligned and non-aligned stage face detector into to objects

  radekd91

  committed
  936cebb

  Copy full SHA for 936cebb

• Update dockerfiles - switch to a non-root user to avoid output file permission issues

  radekd91

  committed
  4a4661b

  Copy full SHA for 4a4661b

Commits on Jan 15, 2024

• Fix data downloading script by adding the metadata file

  radekd91

  committed
  f71557e

  Copy full SHA for f71557e

• Update version

  radekd91

  committed
  f975d6a

  Copy full SHA for f975d6a

• Add experimental upgrade to Python 3.9 and pytorch 2.0. Possible to install from teh new conda environment and requirements files. Also added the related docker version.

  radekd91

  committed
  4cc0453

  Copy full SHA for 4cc0453

• Refactor docker location

  radekd91

  committed
  32fdfb3

  Copy full SHA for 32fdfb3

• Get rid of useless imports

  radekd91

  committed
  3064891

  Copy full SHA for 3064891

Commits on Jan 8, 2024

• Update MICA input processing with a mediapipe landmark option.

  radekd91

  committed
  cd9b771

  Copy full SHA for cd9b771

• Minor refactor

  radekd91

  committed
  6e8c8a0

  Copy full SHA for 6e8c8a0

• Bugfix - VideoFaceDetection dataset used for processing videos was not robust to multiple faces in the video. This should now be fixed.

  radekd91

  committed
  f3aa77e

  Copy full SHA for f3aa77e

Pagination

Previous

Next