The code is developed to simply and implement AffordanceNet in PyTorch (AffordanceNet was developed in Caffe). Here is the original paper for AffordanceNet.
I based this work on TorchVision and PyTorch-Simple-MaskRCNN.
I used pytorch-simple-affnet with the following repos:
- LabelFusion for generating real images.
- NDDS for generating synthetic images.
- arl-affpose-dataset-utils a custom dataset that I generated.
- densefusion for predicting an object 6-DoF pose.
- arl-affpose-ros-node: for deploying our network for 6-DoF pose estimation with our ZED camera.
- barrett-wam-arm for robotic grasping experiments. Specifically barrett_tf_publisher and barrett_trac_ik.
In the sample below we see the differences between traditional Object Instance Segmentation (left) and Object-based Affordance Detection (right).
