SizeCNN - deep learning methods to size the occluded crop

Summary

We provide two methods that can better estimate the size of occluded objects. Both methods use ORCNN (https://github.com/waiyulam/ORCNN), which is an extended Mask R-CNN network that outputs two masks for each object:

1. The regular visible mask (purple mask below)
2. An additional amodal mask of the visible and invisible pixels (green mask below)
   

Installation

See INSTALL.md

Getting started

We provide two methods that can be used to estimate the diameter of occluded crops:

1. ORCNN_PostProcessing.md
2. ORCNN_Regression.md

To compare the sizing performance of both methods, we also provide also a "base-line" method (using Mask R-CNN & Post-processing):

1. MRCNN_PostProcessing.md

License

Our CNNs were forked from ORCNN (https://github.com/waiyulam/ORCNN), which was forked from Detectron2 (https://github.com/facebookresearch/detectron2). As such, our CNNs will be released under the Apache 2.0 license.

Acknowledgements

The methods were developed by Pieter Blok (pieter.blok@wur.nl)