Commits

Branch selector

 new-generation

User selector

All users

Datepicker

All time

Commit History

Commits on Mar 10, 2018

• readme

  anatolix

  committed
  8ac3229

  Copy full SHA for 8ac3229

• readme

  anatolix

  committed
  1ed31a3

  Copy full SHA for 1ed31a3

• work in progress

  anatolix

  committed
  3f4e2f6

  Copy full SHA for 3f4e2f6

• Validation during training. Could be slow.

  anatolix

  committed
  c130f05

  Copy full SHA for c130f05

Commits on Jan 24, 2018

• speed test. testing model fixes

  anatolix

  committed
  7e79fd3

  Copy full SHA for 7e79fd3

Commits on Jan 20, 2018

• ddremov@ coco metric notebook

  anatolix

  committed
  eb806d4

  Copy full SHA for eb806d4

Commits on Jan 19, 2018

• forgot to add notebook

  anatolix

  committed
  19cd22f

  Copy full SHA for 19cd22f

• stupid bug fix

  anatolix

  committed
  f7470b2

  Copy full SHA for f7470b2

Commits on Jan 17, 2018

• fix: save metric one time per epoch

  anatolix

  committed
  907be30

  Copy full SHA for 907be30

• reset iterator each iteration

  anatolix

  committed
  22e2c41

  Copy full SHA for 22e2c41

• added inhouse validation metric and changed validation

  anatolix

  committed
  894256f

  Copy full SHA for 894256f

Commits on Jan 15, 2018

• adaptation to different kind of models.

  anatolix

  committed
  0a056be

  Copy full SHA for 0a056be

Commits on Jan 14, 2018

• validate pose no longer needed, train_pose could do everything

  anatolix

  committed
  ba64357

  Copy full SHA for ba64357

Commits on Jan 4, 2018

• validate pose no longer needed, train_pose could do everything

  anatolix

  committed
  489c38c

  Copy full SHA for 489c38c

Commits on Jan 3, 2018

• submodule

  anatolix

  committed
  73898d9

  Copy full SHA for 73898d9

• Large refactoring, separating my pet project from multipose training to submodule project

  anatolix

  committed
  c8f34a0

  Copy full SHA for c8f34a0

Commits on Jan 2, 2018

• Large refactoring allowing using several configuration files inside a project.

  Show description for 0393549
  anatolix

  committed
  0393549

  Copy full SHA for 0393549

Commits on Dec 24, 2017

• New hdf5 file format (old files still supported)

  Show description for 068dcfb
  anatolix

  committed
  068dcfb

  Copy full SHA for 068dcfb

Commits on Dec 19, 2017

• Simplified mask & hdf5 generation. Half code removed. Haven't tested. Don't use yet.

  anatolix

  committed
  bdccb57

  Copy full SHA for bdccb57

Commits on Dec 10, 2017

• forgot to change number of samples in datasets

  anatolix

  committed
  85ec2d5

  Copy full SHA for 85ec2d5

• Fixed bugs for generate_hdf5.py

  Show description for 2cc45e7
  anatolix

  committed
  2cc45e7

  Copy full SHA for 2cc45e7

Commits on Dec 9, 2017

• Add missed import math.

  dremovd

  committed
  a2775e0

  Copy full SHA for a2775e0

Commits on Dec 8, 2017

• Delete duplicated/unused imports.

  dremovd

  committed
  99281e4

  Copy full SHA for 99281e4

• Refactoring of process_image.py to calculate COCO keypoint metrics.

  dremovd

  committed
  221cfa3

  Copy full SHA for 221cfa3

Commits on Nov 27, 2017

• fixed learning rate decay.

  Show description for a24603d
  anatolix

  committed
  a24603d

  Copy full SHA for a24603d

Commits on Nov 25, 2017

• Found error in augmentation, hope it will fix training.

  Show description for 806f3e4
  anatolix

  committed
  806f3e4

  Copy full SHA for 806f3e4

Commits on Nov 16, 2017

• rmpe_server_tester fixed

  anatolix

  committed
  0cefcca

  Copy full SHA for 0cefcca

• documentation updated.

  anatolix

  committed
  6bfa65a

  Copy full SHA for 6bfa65a

• fixed stupid bug. documentation updated.

  anatolix

  committed
  481bfdc

  Copy full SHA for 481bfdc

• Fixed batch calculation size problem(was hardcoded 10 in some places).

  Show description for 6a2f027
  anatolix

  committed
  6a2f027

  Copy full SHA for 6a2f027

Commits on Nov 15, 2017

• Fixed batch calculation size problem(was hardcoded 10 in some places).

  Show description for 385e380
  anatolix

  committed
  385e380

  Copy full SHA for 385e380

Commits on Nov 14, 2017

• This should really fix NaN problem in my data

  anatolix

  committed
  9e5adc4

  Copy full SHA for 9e5adc4

Commits on Nov 13, 2017

• Looks like no NaN in this setup. Looks like max aggregation of exponents is the cause.

  Show description for a2fa316
  anatolix

  committed
  a2fa316

  Copy full SHA for a2fa316

• Warning added

  anatolix

  committed
  cb86759

  Copy full SHA for cb86759

• Looks like working code. In some pictures it different from C++ code but looks like C++ code is broken not this one. Try to find bug later, now just commit and try to train it

  anatolix

  committed
  8796722

  Copy full SHA for 8796722

Pagination

Previous

Next