3D Box Regression problem

[weisili2016]

hi, in your code , global object orientation is regressed by CNN , from only the contents of the 2D detection result, but , I found that the author of the paper[https://arxiv.org/abs/1612.00496] suggested that ,local orientation is regressed. Did I get it wrong?

[experiencor]

Can you point me to the lines to those 2 statements? It's so long that I forget most of the details.

[weisili2016]

@experiencor 3D Box Regression.ipynb

///////////Preprocessing stuff/////////////// def compute_anchors(angle): ///input angle anchors = [] wedge = 2.*np.pi/BIN l_index = int(angle/wedge) r_index = l_index + 1 ...... What does the input angle mean? global object orientation or local orientation

[experiencor]

I think it's the local orientation described in Figure 3 of the paper.