5,337
edits
Visual Learning and Recognition: Difference between revisions
Visual Learning and Recognition (view source)
Revision as of 17:18, 22 October 2020
, 22 October 2020→Bells & Whistles
| Line 681: | Line 681: | ||
Most used: | Most used: | ||
* Multi-scale training and testing | * Multi-scale training and testing | ||
* Iterative bounding-box prediction | * Iterative bounding-box prediction + weighted NMS | ||
** Do classification, regression then repeat. | |||
* Ensemble | * Ensemble | ||
* More data | * More data | ||
| Line 689: | Line 690: | ||
* Network variants | * Network variants | ||
* NMS variants | * NMS variants | ||
;How to do ensembling for object detection? | |||
This is an open research question. | |||
You can concatenated the regions and pass them through each network. | |||
You can do weighted NMS. | |||
===Bootstrapping=== | |||
Also known as hard negative mining. | |||
# Mine hard negatives from model to fix training set. | |||
# Train model on new fixed training set. | |||
==Will be on the exam== | ==Will be on the exam== | ||