search

HaojunYu1998 / UltraDet

144 stars 20 forks source link

Whole network information #7

Closed optimisticabouthumanity closed 7 months ago

optimisticabouthumanity commented 7 months ago

Is it possible to know how many parameters, layers and FLOPs (including backbone and flownet) do the network have as a whole?

xjtulyc commented 7 months ago

Our code is based on Detectron2, a popular computer vision library developed by Facebook AI Research. To determine the total number of parameters, layers, and FLOPs for the entire network, including both the backbone and FlowNet components, you can refer to the relevant documentation provided by Detectron2.

optimisticabouthumanity commented 7 months ago

@xjtulyc , how many parameters does only the ultradet.pth have?

xjtulyc commented 7 months ago

33.06 M, 33059164 in total.