en
0.25
0.5
0.75
1.25
1.5
1.75
2
Part-Based R-CNNs for Fine-Grained Category Detection
Published on Oct 29, 20146254 Views
Semantic part localization can facilitate fine-grained categorization by explicitly isolating subtle appearance differences associated with specific object parts. Methods for pose-normalized represent
Related categories
Chapter list
Part‐based R-CNNs for Fine-‐grained Category Detection00:00
Challenges of Fine‐grained Categorization - 100:23
Challenges of Fine‐grained Categorization - 200:54
Finding correspondence - 101:05
Finding correspondence - 201:28
Pose‐normalized correspondence01:56
Prior work on fine‐grained categorization - 102:27
Prior work on fine‐grained categorization - 202:49
Progress in deep learning03:05
Deep representa7ons for fine-grained03:43
Limitations - 104:24
Limitations - 204:49
Extend RCNN to parts05:22
Unifying correspondence and feature learning05:52
Overview of our approach06:22
Object and Part detectors06:57
Object and Part detectors - 107:23
Object and Part detectors - 207:34
Box constraint08:19
Geometric constraint: Gaussian Mixture08:42
Geometric constraint: non-‐parametric09:10
Comparison of constraints09:45
Fine-grained categorization10:27
Results10:44
Dataset: CUB‐200-201110:47
Fine-graied categorization results11:16
Does finetuning help? - 112:20
Does finetuning help? - 212:41
Part localization results13:00
Part localization samples13:31
Where doesn’t it work? - 113:59
Where doesn’t it work? - 214:24
Untitled14:38