Skip to content

Implementation of Robotic pick-and-place of novel objects in clutter with multi-affordance grasping and cross-domain image matching by Zeng Et Al

Notifications You must be signed in to change notification settings

sujaymanb/ClutterObjDetectionFCN

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

10 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Implementation

I use a fully convolutional network (FCN) style architecture with two trunks/inputs for the RGB and depth respectively. The encoder has 4 layers (32 filters of size 5, stride of 2 for the first 3 layers and stride 1 for the final layer). The features for the RGB and depth are concatenated channelwise and we interleave 2 conv layers with bilinear upsampling to construct the heatmaps.

The model is trained using the BCE (Binary cross entropy) loss, classifying the pixels with probabilities for pick.

Modalities: Both RGB and depth are used.

Things to try

Given more time I would try the following things:

  • Data preprocessing: Resize (downsample), crop the data
  • Data augmentation: add some noise, rotations, brightness, random cropping, etc to augment datset
  • Different architectures: try U-net architectures
  • Hyperparameter Tuning: learning rate, conv layer params (num filters, etc)

Running Instructions

Train:

  • modify params in train.py
  • Run: python train.py

Test:

  • modify params in test.py
    • select which training run_name to load the model from
    • select which checkpoint epoch_number to load the model for
  • Run: python test.py
  • output is generated in the 'test' folder for corresponding run folder in:
    • output/[run_name]/test

References

Zeng et al. "Robotic pick-and-place of novel objects in clutter with multi-affordance grasping and cross-domain image matching." 2018 IEEE international conference on robotics and automation (ICRA). IEEE, 2018.

About

Implementation of Robotic pick-and-place of novel objects in clutter with multi-affordance grasping and cross-domain image matching by Zeng Et Al

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages