Distributional Instance Segmentation: Modeling Uncertainty and High Confidence Predictions with Latent-MaskRCNN

Project Page | Paper | Dataset

Pytorch implementation of our autoregressive model formulation for 3D bounding-box estimation & detection.

Distributional Instance Segmentation: Modeling Uncertainty and High Confidence Predictions with Latent-MaskRCNN
YuXuan Liu^1,2, Nikhil Mishra^1,2, Pieter Abbeel^1,2, Xi Chen^1
1Covariant.ai, ²UC Berkeley

Latent-MaskRCNN

This code is based on Detectron2

Installation see INSTALL.md

For our Latent-MaskRCNN models see configs/latent.

The MaskRCNN decoder and confidence mask implementation is in detectron2/modeling/meta_arch/rcnn.py

Our encoder and prior models are in detectron2/modeling/latent.py.

License

Shield:

This work, including the paper, code, weights, and dataset, is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.