5,337
edits
Visual Learning and Recognition: Difference between revisions
Visual Learning and Recognition (view source)
Revision as of 14:25, 17 October 2020
, 17 October 2020→R-CNN
(→R-CNN) |
|||
| Line 520: | Line 520: | ||
#* All of the region proposals likely contain an object. | #* All of the region proposals likely contain an object. | ||
# For each bounding box: | # For each bounding box: | ||
#* Dilate the proposal. | #* Dilate the proposal on each side by <math>p=16</math> pixels. | ||
#* Crop it out and scale to <math>227 \times 227</math>. | #* Crop it out and scale to <math>227 \times 227</math>. | ||
#* | #* Pass it through a CNN (5 conv + 2 FC) to get <math>4096</math> dim features. | ||
#* Do classification using an SVM. | |||
# Do object proposal refinement to predict object bounding box. | # Do object proposal refinement to predict object bounding box. | ||