PyCenterNet

PyCenterNet copied to clipboard

When I modify the loss function of loss_ bbox_ Init and loss_ bbox_ from GIoULoss to CIoULoss, which can be trained but the evaluation result is null . Can you...

wmrenr

AttributeError: 'PyCenterNetHead' object has no attribute 'init_assigner'

1

Hi I'm trying to train the PyCenterNet++ with the "[R2-101]-DCN" backbone, below is the error message and environmental details, for reference: **----------------------- environmental details -----------------------** torch.__version__ :- 1.4.0 torch.cuda.is_available() :-...

sw-ot-jaymehta

what 's the mean of loss_sem?

wmrenr

Hi, where can I find the configuration for the real time model? thanks for your works and sharing. Best, Vinh edit: I guess it's this config: pycenternet_r50_fpn_3_levels_giou_fast_adamw_mstrain_6x_coco.py ?

vinh-cao

Thank you for your great contribution! How can the central keypoints be visualized during testing？

wmrenr

Could you please provide the training log for the setting in pycenternet_r50_fpn_giou_1x_coco.py? Thank you.

goldentimecoolk

ImportError: cannot import name 'corner_pool_ext' from partially initialized module 'mmdet.ops.corner_pool' (most likely due to a circular import) I met the error when i am going to start a train.I haven't...

Chlhacker

Can you provide the final model that has been trained on the coco dataset? Thanks a lot！

ssli23

use resnet50 as the backbone training COCO, the map is only 42.5，cannot reach 46.4

5

 The configs file used is pycenternet_r50_fpn_3_levels_giou_fast_mstrain_6x_coco.py，others haven't changed anything，the result of map is only 42.5，why？？？？？

wssywh

Hi, I got an error in SEPFLoss, The input dimensions are [1814400](pred) and [3628800](target), I don't know how to solve it, can you provide any advice? Thanks! ``` Traceback (most...

Qiang-Feng98