How can I interpret a float32 as four uint8s in numpy?

Question

I tried to use python to read points cloud data files, which with extension ".pcd". They store RGBA color information with just one numpy float32. How can i convert the float32 data type into 4 uint8 numbers in python ? For example, what is the rgb value hidden in number 4.51073351e-39 ?

In [13]: pc.pc_data["rgb"]
Out[13]: 
array([  4.51073351e-39,   4.41853788e-39,   4.87845245e-39, ...,
         3.31220574e-39,   2.94594696e-39,   3.12997949e-39], dtype=float32)

Here is the link to download the dataset: http://rgbd-dataset.cs.washington.edu/dataset/rgbd-dataset_pcd_ascii/apple_1.tar

And i used pypcd to read the file: here is the pypcd github link:https://github.com/dimatura/pypcd

those values seem very very low... are you sure the format is OK? — Jean-François Fabre, Oct 25 '16 at 20:40
possible dup : http://stackoverflow.com/questions/16444726/binary-representation-of-float-in-python-bits-not-hex — corn3lius, Oct 25 '16 at 20:50
@corn3lius: None of the answers there are good practice when using numpy — Eric, Oct 25 '16 at 21:12

score 2 · Accepted Answer · answered Oct 25 '16 at 21:10

I think you want pc.pc_data["rgb"].view((np.uint8, 4)), which for your example gives:

array([[23, 30, 49,  0],
       [22, 29, 48,  0],
       [36, 31, 53,  0],
       ...
       [21, 17, 36,  0],
       [26, 20, 32,  0],
       [28, 21, 34,  0]], dtype=uint8)

score 1 · Answer 2 · answered Mar 16 '18 at 04:03

1

Really late to the party, but pypcd (author here) actually has this function as decode_rgb_from_pcl, you can call it as pypcd.decode_rgb_from_pcl(pc.pc_data['rgb']). Admittedly, I like the way @Eric implemented this better than mine :D

answered Mar 16 '18 at 04:03

Daniel Maturana

11
1

How can I interpret a float32 as four uint8s in numpy?

2 Answers2