5,337
edits
Visual Learning and Recognition: Difference between revisions
Visual Learning and Recognition (view source)
Revision as of 17:15, 20 October 2020
, 20 October 2020→RoI Pooling variants
| Line 623: | Line 623: | ||
* RoI Warp (MNC) | * RoI Warp (MNC) | ||
* RoI Align (Mask R-CNN) | * RoI Align (Mask R-CNN) | ||
===Architectures=== | |||
Network Variants | |||
* ResNet | |||
Skip-connection Variants | |||
* ION Inside Out Network | |||
* TDM: Top-down Modulation Network | |||
* FPN: Feature Pyramid Network | |||
===ION: Inside Out Network=== | |||
The key idea is that we want a feature vector which uses features from multiple scales. | |||
'''Potential Exam Question''' | |||
We want to use features from multiple levels. The RoI is fixed. | |||
The resolution, number of channels, and magnitude of features can be different. | |||
* Do L2 normalization of the features at different layers | |||
* Concatenate features | |||
* Rescale them. | |||
* Do 1x1 convolution and give it to the FC layer. | |||
==Will be on the exam== | ==Will be on the exam== | ||