MapTR

MapTR copied to clipboard

Dear authors, I have two simple questions: 1) why there is no hyper-parameter before Loss-focal and Loss-position when calculating Loss-instance-match? I believe the magnitude of loss-position is much larger than...

Running-George

Hi, thanks for your great work. We are training our own model with maptrv2 but the output mAP does not meet our expectations. The results on test data differ significantly...

HongbiaoZ

MapTRv2 on argoverse2 2d training config

1

Will you release argoverse2 2d training config?

imyyf

memroy -> memory

eltociear

Environment configuration

2

Hello, when I run python setup.py develop, an error will be reported: RuntimeError: Python version >= 3.9 required. Instead of asking for Python version 3.8

PlutoXN

When I generate dataset, there is an error related to the mmcv. And it still fail to run(incompatibility) even i have updated mmcv to 2.0.0rc4. Can anyone help? Error: (maptr)...

Jiayiiiiiii502

How about not sharing point-level queries in the one2many assigner

1

@LegendBC @wondervictor @outsidercsy Thanks for your code! I'm curious that sharing point-level queries could cause interaction between one2one group and one2many group in the training, which is not the desired...

JingweiZhang12

Hello, I would like to ask why the mAP result I got using the nuscenes-test dataset is significantly higher than the mAP result using the validation dataset.（About 10 higher）

colahe

Hi team: I have two question. 1. the link of gkt temporal log is not shared. 2. from the bevformer log, it doesn't use fp16 training, and when I try...

gold-mango

original dataset overlap problem

1

Hello, author, In the maptr_V2 paper, the central line exhibited excellent results. Did you split the dataset because the training and testing sets of the original dataset overlap geographically（both nuscense...

wheelandwheat