search

dvlab-research / PanopticFCN

Fully Convolutional Networks for Panoptic Segmentation (CVPR2021 Oral)
Apache License 2.0
391 stars 53 forks source link

Is this model able to convert to tensorRT for faster acceleration? #1

Closed lucasjinreal closed 3 years ago

lucasjinreal commented 3 years ago

Is this model able to convert to tensorRT for faster acceleration?

lucasjinreal commented 3 years ago

Oh. BTW, I can not found any demo script inside repo, did you ever visiualize the final result?

yanwei-li commented 3 years ago

Thanks for your attention! For faster acceleration, we will try tensorRT later. For visualization, we provide several results in the paper on COCO, Cityscapes, and Mapillary Vistas datasets. Of course, you can also visualize results using the tools in Detectron2.

lucasjinreal commented 3 years ago

@yanwei-li Would u opensource vis script as well?

yanwei-li commented 3 years ago

Thanks for your interest. Maybe you can try the visualization utils in Detectron2, which is also suitable for this project.

lucasjinreal commented 3 years ago

It seems someone already done this: https://zhuanlan.zhihu.com/p/341813709