aharley / track_check_repeat Goto Github PK

Hi! I was just browsing through projects that are using cc3d and I noticed a spot where you might be able to get a lot better performance. I don't know how critical this routine is to your project, but I'll offer some advice in case it is helpful.

Could be rewritten approximately like so to remove lots of redundant computation:

    labels = connected_components(mask)
    stats = cc3d.statistics(labels)
    for segid, extracted_vox in cc3d.each(labels, binary=True):
        slices = stats['bounding_boxes'][segid]
        zmin, zmax = slices[2].start, slices[2].stop
        ymin, ymax = slices[1].start, slices[1].stop
        xmin, xmax = slices[0].start, slices[0].stop
        voxel_count = stats['voxel_counts'][segid]

        if (zmax-zmin > min_side and 
            ymax-ymin > min_side and 
            xmax-xmin > min_side and 
            voxel_count > min_voxels): 

I haven't tested this snippet, it's just a guide. If this isn't helpful, please feel free to ignore this; I don't want to waste your time.

Thanks so much and good luck!
Will

Hi,
first of all let me say thank you for the great work! :) I was thinking about trying it out, but I have noticed that there is no license included in the project. Without a license, the work is under exclusive copyright by default. Could you please add one? You could check out choosealicense.com if you are unsure which license fits your needs.
Thank you very much.

Hi,
I have a few questions regarding your evaluation.

1. In the readme file from this repository, it's mentioned that the testing has been done using sequences 0009-0010. However, in your code, the evaluation code uses sequences 0010-0011. Could you specify which sequence you evaluated your trained model on?

2. Was the quantitative result in object discovery from the paper produced using the sequence 0010-0011?

3. For the object discovery, did you evaluate your detector on all types of objects(e.g. cyclist, car, van, and pedestrian) except for 'DontCare' class in KITTI?

Thanks a lot for your work!