Skip to content
This repository has been archived by the owner on Oct 31, 2023. It is now read-only.

Latest commit

 

History

History
68 lines (60 loc) · 2.8 KB

data_analysis.md

File metadata and controls

68 lines (60 loc) · 2.8 KB
Raw

Hand Contact Probabilities

Phalange-level

If you are interested in phalange-level association (Fig 5(b) in the paper), look at scripts/show_contactmap.py. finger_idx and part_idx contain the indices of the finger and the phalange closest to each object point. Then apply_semantic_colormap_to_mesh() takes that and the thermal contact value at each point to create the color-coded contact maps shown in the README.

Fine-grained

E.g. Fig 5(a) in the paper. This is done with scripts/data_analysis/hand_contact_prob.py, for example computing the hand contact probablity for all grasps by participant #28, use intent:

$ python scripts/data_analysis/hand_contact_prob.py --p_num 28 --intent use

You can average various slices of the dataset in this script by choosing p_num, object_name, and intent:

(contactpose) $ python scripts/data_analysis/hand_contact_prob.py --help
usage: hand_contact_prob.py [-h] [--p_num P_NUM]
                            [--intent {use,handoff,use,handoff}]
                            [--object_name OBJECT_NAME]

optional arguments:
  -h, --help            show this help message and exit
  --p_num P_NUM         Participant number, comma or - separated, ignore for
                        all participants
  --intent {use,handoff,use,handoff}
                        Grasp intent
  --object_name OBJECT_NAME
                        Name of object, comma separated, ignore for all
                        objects

Active Areas

These are areas on the object surface which are highly likely to be touched. For more details see Fig. 7 in the paper. scripts/data_analysis/active_areas.py allows you to discover the active areas for various objects, corresponding to a particular part of the hand.

Use without the --show switch to calculate and save the active area information in data/contactpose_data, and use the --show switch to display the object mesh with active areas textured on it.

(contactpose) $ python scripts/data_analysis/active_areas.py --help
usage: active_areas.py [-h] --finger_idx {0,1,2,3,4} --part_idx {0,1,2,3}
                       --object_name OBJECT_NAME --intent {use,handoff}
                       [--p_nums P_NUMS] [--show]

optional arguments:
  -h, --help            show this help message and exit
  --finger_idx {0,1,2,3,4}
                        0->4 : thumb->little
  --part_idx {0,1,2,3}  0->3 : proximal to distal phalanges, 3 = finger tip
  --object_name OBJECT_NAME
  --intent {use,handoff}
  --p_nums P_NUMS       Participant numbers, comma or - separated.Skipping
                        means all participants
  --show