Open GiulioRomualdi opened 1 year ago
As explained in https://github.com/robotology/walking-teleoperation/pull/114, this PR implements a CNN to classify the roughness of an object using the readouts of the skin on the robot palm. The neural network is loaded in a json file and is loaded using frugally-deep. Since frugally-deep is a header-only library I decided to automatically clone it with fetch content. The CNN takes as input the calibrated skin data (provided by the skinmanager) and treat it as a grayscale image. The code used to generate the model can be found at https://github.com/paolo-viceconte/rock_classifier/commit/381fcf4d68ddbdf0096b3624fb9c469174285536 (@paolo-viceconte)
json
frugally-deep
skinmanager
Moreover, this PR removes the fingertip actuation (@RiccardoGrieco) in https://github.com/robotology/walking-teleoperation/pull/114/commits/63aff826e30aa1c114e289f91a581013d72c0db5
As explained in https://github.com/robotology/walking-teleoperation/pull/114, this PR implements a CNN to classify the roughness of an object using the readouts of the skin on the robot palm. The neural network is loaded in a
jsonfile and is loaded usingfrugally-deep. Sincefrugally-deepis a header-only library I decided to automatically clone it with fetch content. The CNN takes as input the calibrated skin data (provided by theskinmanager) and treat it as a grayscale image. The code used to generate the model can be found at https://github.com/paolo-viceconte/rock_classifier/commit/381fcf4d68ddbdf0096b3624fb9c469174285536 (@paolo-viceconte)Moreover, this PR removes the fingertip actuation (@RiccardoGrieco) in https://github.com/robotology/walking-teleoperation/pull/114/commits/63aff826e30aa1c114e289f91a581013d72c0db5