I see that you published the pretrained checkpoints for camera, lidar and fusion here #19, but when I evaluate them I don't get the same mAP numbers reported in your Readme (and paper).
I get the following numbers:
Camera: 22.7 mAP (ok)
Lidar: 64.6 mAP (different, see also #54)
Fusion: 67.9 mAP (different, this is the number from paper v1)
Since the fusion mAP is the one reported in your v1 paper, I'm supposing that also the lidar checkpoint is from v1. In the Readme I can find camera and fusion pretrained checkpoints for v2/v3 paper, and I get the correct numbers in evaluation, but Lidar is missing.
Could you please release the pretrained lidar checkpoint that is used to train the fusion model that gets 69.6 mAP?
Hello,
First of all, thank you for your work!
I see that you published the pretrained checkpoints for camera, lidar and fusion here #19, but when I evaluate them I don't get the same mAP numbers reported in your Readme (and paper). I get the following numbers:
Since the fusion mAP is the one reported in your v1 paper, I'm supposing that also the lidar checkpoint is from v1. In the Readme I can find camera and fusion pretrained checkpoints for v2/v3 paper, and I get the correct numbers in evaluation, but Lidar is missing.
Could you please release the pretrained lidar checkpoint that is used to train the fusion model that gets 69.6 mAP?
Thank you!