HDA dataset issues: Difference between revisions
No edit summary |
No edit summary |
||
| (3 intermediate revisions by the same user not shown) | |||
| Line 1: | Line 1: | ||
== hda_sample_train_data could be increased == | <del>== hda_sample_train_data could be increased ==</del> | ||
Person 72 appears fully visible in camera 56 in frame 259 and is not in the hda_sample_train_data <br> | <del>Person 72 appears fully visible in camera 56 in frame 259 and is not in the hda_sample_train_data <br> | ||
Person 78 appears fully visible in camera 56 in frame 12 and is not in the hda_sample_train_data | Person 78 appears fully visible in camera 56 in frame 12 and is not in the hda_sample_train_data</del> | ||
It seems that since those two people are people that at the time needed to have their face blurred out, at the time they were not added to the training set. | <del>It seems that since those two people are people that at the time needed to have their face blurred out, at the time they were not added to the training set.</del> | ||
It's probably not a good idea to change (add) stuff to hda_sample_train_data now, since there is already published results with this 250 image training set | <del>It's probably not a good idea to change (add) stuff to hda_sample_train_data now, since there is already published results with this 250 image training set</del> | ||
HDA new training set made. Previous one archived. | |||
== Annotation errors == | |||
Ped 17 on Cam 57 in frames 425 and 426 is unoccluded (but marked occluded in the annotations, and in the allD files..., and has the masks...) | Ped 17 on Cam 57 in frames 425 and 426 is unoccluded (but marked occluded in the annotations, and in the allD files..., and has the masks...) | ||
Latest revision as of 19:55, 18 November 2014
== hda_sample_train_data could be increased ==
Person 72 appears fully visible in camera 56 in frame 259 and is not in the hda_sample_train_data
Person 78 appears fully visible in camera 56 in frame 12 and is not in the hda_sample_train_data
It seems that since those two people are people that at the time needed to have their face blurred out, at the time they were not added to the training set.
It's probably not a good idea to change (add) stuff to hda_sample_train_data now, since there is already published results with this 250 image training set
HDA new training set made. Previous one archived.
Annotation errors
Ped 17 on Cam 57 in frames 425 and 426 is unoccluded (but marked occluded in the annotations, and in the allD files..., and has the masks...)