Any notebook on PointsItemList

Is there any notebook available showing the usage of PointsItemList

You want lesson 3, headpose

1 Like

Thanks @muellerzr

  1. what is being done here

     def convert_biwi(coords):
         c1 = coords[0] * cal[0][0]/coords[2] + cal[0][2]
         c2 = coords[1] * cal[1][1]/coords[2] + cal[1][2]
         return tensor([c2,c1]) 
  2. what are the below fixed set of values are they camera params ?

     array([[517.679,   0.   , 320.   ],
            [  0.   , 517.679, 240.5  ],
            [  0.   ,   0.   ,   1.   ]])
  3. Are below coordinates x,y,z ?

array([187.332 , 40.3892, 893.135 ])

  1. if there were 6d pose in stead of 3d pose where other 3 were rotations across the 3 axes what could be likely approach or tweak in this notebook needed

All but 4 are explained in the lecture :slight_smile:

For 6D do you mean we have a 6D image?

Aren’t those intrinsic camera parameters? on diagonal fx, fy- focal length and cx, cy as camera center?
look here:

Yes they are. It’s a calibration given for that particular dataset

Yes basically I n trying to fit this data bunch for building car position detection competition in kaggle . Every position of car is expressed using 6 pose
X,y,z,yaw ,pitch ,roll . Last three are rotation across axes .
You can have brief check in this ,basically datasets building part