search

lzccccc / SMOKE

SMOKE: Single-Stage Monocular 3D Object Detection via Keypoint Estimation
MIT License
707 stars 177 forks source link

How DIMENSION_REFERENCE in SMOKE_HEAD is calculated? #33

Open xxradon opened 3 years ago

xxradon commented 3 years ago

I want to train smoke on different datasets,but when I pass over the whole code,I kown that num class is related with DIMENSION_REFERENCE.In kitti dataset,there is 3 values for DIMENSION_REFERENCE in smoke/config/defaults.py:

# Reference car size in (length, height, width)
# for (car, cyclist, pedestrian)
_C.MODEL.SMOKE_HEAD.DIMENSION_REFERENCE = ((3.88, 1.63, 1.53),
                                           (1.78, 1.70, 0.58),
                                           (0.88, 1.73, 0.67))

and I want to know how DIMENSION_REFERENCE in SMOKE_HEAD is calculated? Thanks.

ccerhan commented 3 years ago

You can calculate the dim references by simply averaging all dims (length, height, width) for each class in the dataset. Let's say there are 2 cars in your whole dataset Car 1 (4, 2, 2) Car 2 (5, 3, 1) dim ref of car class should be (4.5, 2.5, 1.5)

Also, the depth reference can be calculated by putting the depth information (z-location) of all objects for all classes in an array, D. The depth reference will be ( avg(D), std(D) )