Different input resolution?

j96w / DenseFusion

"DenseFusion: 6D Object Pose Estimation by Iterative Dense Fusion" code repository

https://sites.google.com/view/densefusion

MIT License

1.1k stars 300 forks source link

Different input resolution? #151

Open TommasoBendinelli opened 4 years ago

TommasoBendinelli commented 4 years ago

Hi there, I managed to train the model correctly on my dataset and I pretty happy about results. I would like now to use as input an image with higher resolution. In what way should I change the network architecture?

j96w commented 4 years ago

Hi, a very interesting question. To be honest, I'm not sure which part of the network needs to be modified accordingly. Do you have some findings to share? For example, the efficiency problem after increased the resolution? Thanks.

TommasoBendinelli commented 4 years ago

Not really, I used the standard format 640x480. But it is definitely something to investigate. Especially for small pieces, better resolution might imply better performances.