How to load C3D features into numpy array？ about c3d HOT 6 CLOSED

Hi, I made a python function that might help: https://gist.github.com/gbpcosta/22f9679c1984473f2971e4d4d2d60aa9

from c3d.

@MichaelXin The format is:
5 32-bit integers: num, chanel, length, height, width. (record the size of the blob)
Then followed by the data of (num * channel * length * height * width) each data is a 32-bit float in row order.

from c3d.

@MichaelXin check out the script to read blob which preserving blob shape: https://github.com/facebook/C3D/blob/master/examples/c3d_feature_extraction/script/read_binary_blob_preserve_shape.m
You can write similar Python script. Note that MATLAB uses column order, while numpy uses row order. Thus, you don't need to reshape (as I used in MATLAB).

from c3d.

@MichaelXin Hi,I'm confused about load the "fc6-1"features.The feature's number is 4096，so why we shoud reshape the data size.
Thanks.

from c3d.

Could you please give me a python script for read_binary_blob and read_binary_bob_preserve_shape.m ?

from c3d.

@gbpcosta Thanks a lot! :)

from c3d.