Commits · 95bf6bfbd0df9fde02530be7e4f4f937242f4c33 · Vision / GenVQA · GitLab

Snippets Groups Projects

Nov 08, 2016
- extract word vectors · 95bf6bfb
  tgupta6 authored 8 years ago
  
  95bf6bfb
Nov 03, 2016
- eval and visualize script for the interpreted evaluation · 15aa9fc0
  tgupta6 authored 8 years ago
  
  15aa9fc0
Nov 02, 2016
- Merge branch 'eval_interpret' into rel_bin_feats · 95ac8f92
  tgupta6 authored 8 years ago
  
  95ac8f92
Nov 01, 2016
- fix , bug · 88622a32
  tgupta6 authored 8 years ago
  
  88622a32
- eval_interpret · 0b89734e
  tgupta6 authored 8 years ago
  
  0b89734e
- make visualization darker · 7f8770b4
  tgupta6 authored 8 years ago
  
  7f8770b4
- answer eval · 5d6de0cd
  tgupta6 authored 8 years ago
  
  5d6de0cd
- answer eval · 416b6f96
  tgupta6 authored 8 years ago
  
  416b6f96
Oct 29, 2016
- fix bug in yes no feat · 82948cf2
  tgupta6 authored 8 years ago
  
  82948cf2
- updated constants_vision_gpu_1 · 4cf7ffc6
  tgupta6 authored 8 years ago
  
  4cf7ffc6
- add all rel and yes no num features · 262cb78b
  tgupta6 authored 8 years ago
  
  262cb78b
- zero out rel feat, yes no features already zeroed out · 96ea6292
  tgupta6 authored 8 years ago
  
  96ea6292
- ans select best model · f9226e5f
  tgupta6 authored 8 years ago
  
  f9226e5f
Oct 28, 2016
- changed k to 100 in obj atr eval · e41c45f7
  tgupta6 authored 8 years ago
  
  e41c45f7
- add attributes comparison to cmp_obj_freq_acc · 2d88e085
  tgupta6 authored 8 years ago
  
  2d88e085
- delete freq_acc_plot.py · a9e4418e
  tgupta6 authored 8 years ago
  
  a9e4418e
- created freq_acc_plot_atr and _obj · 961a76b3
  tgupta6 authored 8 years ago
  
  961a76b3
- remove bin 0-10 · c892cf89
  tgupta6 authored 8 years ago
  
  c892cf89
- changed pass to continue · 63a88bea
  tgupta6 authored 8 years ago
  
  63a88bea
- freq_acc_plot modified to read in object_labels.json · 4863d27a
  tgupta6 authored 8 years ago
  
  4863d27a
- freq_acc_plot and cmp_obj_freq_acc modified · 5a14825e
  tgupta6 authored 8 years ago
  
  5a14825e
- bucketed object accuracy in freq_acc_plot · 287e6eb1
  tgupta6 authored 8 years ago
  
  287e6eb1
- fix bug in writing attribute pred and change k to 20 in obj atr eval · 3a4eefd5
  tgupta6 authored 8 years ago
  
  3a4eefd5
Oct 27, 2016
- remove typo in obj eval · 426b9f54
  tgupta6 authored 8 years ago
  
  426b9f54
- write attribute predictions to attribute_predictions.json · 8b8bf239
  tgupta6 authored 8 years ago
  
  8b8bf239
- use eval_regions_json in obj atr select_best_mode · f8a84710
  tgupta6 authored 8 years ago
  
  f8a84710
Oct 24, 2016
- batch norm single features before adding · 725422dc
  tgupta6 authored 8 years ago
  
  725422dc
- change reduce_mean to reduce_sum · 2ba75798
  tgupta6 authored 8 years ago
  
  2ba75798
- add yes no num feat · b78a47a7
  tgupta6 authored 8 years ago
  
  b78a47a7
- added adjective scores for ques and nouns · 4ae6983f
  tgupta6 authored 8 years ago
  
  4ae6983f
- constants_crunchy modified · e95f5ec8
  tgupta6 authored 8 years ago
  
  e95f5ec8
- parsing script for vqa test and testdev · 949f26f5
  tgupta6 authored 8 years ago
  
  949f26f5
Oct 23, 2016
- changed batch size to 50 and wts to ans through obj atr in constants.crunchy · 324b3beb
  tgupta6 authored 8 years ago
  
  324b3beb
- Split question and answer noun relevance scores and use those as features · 834d6a1f
  tgupta6 authored 8 years ago
  
  834d6a1f
Oct 21, 2016
- zeroed out yes no features · 7e17e2dc
  tgupta6 authored 8 years ago
  
  7e17e2dc
- zeroed out relevance features in answer inference graph · 1fcbf4f8
  tgupta6 authored 8 years ago
  
  1fcbf4f8
- the architecture that gives 64 overall accuracy · c9c1b754
  tgupta6 authored 8 years ago
  
  c9c1b754
Oct 19, 2016
- remove obj atr last fc layer bn and relu · 67921a24
  tgupta6 authored 8 years ago
  
  67921a24
- use relevance to fuse region features · 78c53ec5
  tgupta6 authored 8 years ago
  
  78c53ec5
- train answer script prints correct answer relevances · 1a240224
  tgupta6 authored 8 years ago
  
  1a240224