search

cure-lab / MagicDrive

[ICLR24] Official implementation of the paper “MagicDrive: Street View Generation with Diverse 3D Geometry Control”
https://gaoruiyuan.com/magicdrive/
GNU Affero General Public License v3.0
419 stars 22 forks source link

Detection Results using BEVFusion of Table I? #6

Closed colian closed 5 months ago

colian commented 5 months ago

Hi, nice work!

Do the results of Table I all use the checkpoint officially released by bevfusion? Or the checkpoint is trained by urself?

Looking forward to your reply.

flymin commented 5 months ago

In Table 1, we use all perception models only for testing. Therefore, we adopt the officially released ckpts without training.

colian commented 5 months ago

Thanks for your reply!

Since the officially released ckpt of BEVFusion is in resolution of (256,704), and you resized the synthesis image (both (224,400) and (272,736)) into (256,704) resolution for evaluation?

flymin commented 5 months ago

Yes, we kept the original data processing pipeline. Actually, we upscale and pad each generated images to 1600x900 and save them to disk, so that we can run the original code without any change.

colian commented 5 months ago

Thanks for your quick reply!

LeonisAlpha commented 3 months ago

Yes, we kept the original data processing pipeline. Actually, we upscale and pad each generated images to 1600x900 and save them to disk, so that we can run the original code without any change.

Thanks for your reply!

May I ask if you could explain the specific operation of pads and upsamples here?