Result of a simple experiment on KITTI dataset by adding RGB features into points

[xmyqsh]

Here is a simple experiment on KITTI dataset. By adding RGB features into points, the 3d AP increases, but the bev AP drops a lot.

Benchmark

car AP@0.70, 0.70, 0.70:
bbox AP:90.70, 88.95, 87.33
bev  AP:89.65, 84.71, 81.73
3d   AP:85.85, 76.36, 69.63
aos  AP:90.61, 88.30, 86.31

with RGB feature

car  AP@0.70,  0.70,  0.70:
bbox AP:90.63, 88.86, 87.35
bev  AP:89.75, 86.15, 83.00
3d   AP:85.75, 75.68, 68.93
aos  AP:90.48, 88.36, 86.58

Based on Painted PointPillars result with segmentation feature instead of RGB feature BEV on test set

 mAP | Car AP
Mod. | Easy | Mod. | Hard 
73.84 90.21 87.75    84.92
76.46 90.01 87.65    85.26
+2.62 -0.2   -0.1   +0.34

I address this as an overfitting problem and will test it.

Does anybody observe a similar result? How about using the Nucense dataset? How about adding augmentation on RGB?

Hope for large 3d AP gain on Pedestrian and Cyclist.

[xmyqsh]

Guess what? Benchmark trainset

car AP@0.70, 0.70, 0.70:
bbox AP:98.16, 91.20, 90.32
bev  AP:97.97, 90.36, 89.80
3d   AP:90.34, 88.32, 80.83
aos  AP:98.09, 90.94, 89.95

with RGB feature trainset

car AP@0.70, 0.70, 0.70:
bbox AP:98.32, 91.37, 91.30
bev  AP:98.23, 90.55, 90.13
3d   AP:97.91, 89.30, 86.74
aos  AP:98.14, 91.09, 90.88

[jhultman]

Hi, thanks for sharing your results. What is the difference in experimental setup between the results in your first post and the results in your follow-up post?

Did you disable database sampling at training time? Otherwise the pasted samples may have inconsistent RGB features.

[AllenPeng0209]

Wondering how to implement seg feature on this framework quickly. I try to implement while in pipelines/preprocess.py file, however it's seen like i can't inference image in subprecessor. How to modify it? Thanks.