search

heshuting555 / DsHmp

[CVPR-2024] Decoupling Static and Hierarchical Motion Perception for Referring Video Segmentation
78 stars 0 forks source link

Simple question regarding model training #7

Closed golddohyun closed 3 months ago

golddohyun commented 3 months ago

Hello, thank you for the excellent work!

I have some questions about model training. I downloaded the backbone weights of MaskFormer2 (model_final_86143f.pkl), and I want clarification on where in the code these weights are being used.

Is it correct to specify the path to the backbone weight .pkl file in the MODEL.WEIGHTS [path_to_weights] section of the training script?

Thanks:)

heshuting555 commented 3 months ago

The code is based on the detectron framework. You can learn about this framework. It will load backbone weight by default.

golddohyun commented 3 months ago

Thanks for the reply! So you mean that i don't have to directly write MODEL.WEIGHTS ./model_final_86143f.pkl in the training script?

heshuting555 commented 3 months ago

You need or you can put it in the config.yaml

golddohyun commented 3 months ago

ahh great all issues solved. thanks!