plugin/multilevelCropAndResizePlugin/README.md
This plugin is deprecated since TensorRT 10.12 and will be removed in a future release. Note alternatives are planned to be provided.
Table Of Contents
The MultilevelCropAndResize plugin performs the ROIAlign operation on the output feature maps from FPN (Feature Pyramid Network). It is used for MaskRCNN inference in Transfer Learning Toolkit.
This plugin supports the NCHW format. It takes 6 inputs in the following order: roi, and 5 feature_maps from FPN (Note: 5 feature_maps are required for this plugin and will not function properly with a lesser number of feature_maps).
roi is the ROI candidates from the MultilevelProposeROI plugin. Its shape is [N, rois, 4] where N is the batch_size, rois is the number of ROI candidates and 4 is the number of coordinates.
feature_maps are the output of FPN. In TLT MaskRCNN, the model we provide contains 5 feature maps from FPN's different stages.
This plugin generates one output tensor of shape [N, rois, C, pooled_size, pooled_size] where C is the channel of mutiple feature maps from FPN and pooled_size is the height(and width) of the feature area after ROIAlign.
This plugin has the plugin creator class MultilevelCropAndResizePluginCreator and the plugin class MultilevelCropAndResize.
The following parameters were used to create MultilevelCropAndResize instance:
| Type | Parameter | Description |
|---|---|---|
int | pooled_size | The spatial size of a feature area after ROIAlgin will be [pooled_size, pooled_size] |
int[3] | image_size | The size of the input image in CHW. Defaults to [3, 832, 1344] |
For terms and conditions for use, reproduction, and distribution, see the TensorRT Software License Agreement documentation.
May 2025: Add deprecation note.
March 2022: This is the second release of this README.md file.
June 2020: First release of this README.md file.
There are no known issues in this plugin.