How to get FPS for a model?

[kirkzZ]

How to get FPS for a model?

Note

We only consider adding new features if they are relevant to many users.

If you request implementation of research papers -- we only consider papers that have enough significance and prevalance in the object detection field.

We do not take requests for most projects in the projects/ directory, because they are research code release that is mainly for other researchers to reproduce results.

"Make X faster/accurate" is not a valid feature request. "Implement a concrete feature that can make X faster/accurate" can be a valid feature request.

Instead of adding features inside detectron2,
you can implement many features by extending detectron2.
The projects/ directory contains many of such examples.

[zhanghang1989]

For fps on GPU, it is shown in the training log. We don't have open source version of benchmarking on devices.

[kirkzZ]

hello, thanks for your answering, the time is the fps?
eta: 9:58:17 iter: 174159 total_loss: 0.9185 loss_cls: 0.229 loss_box_reg: 0.2562 loss_mask: 0.2784 loss_rpn_cls: 0.04507 loss_rpn_loc: 0.0934 time: 0.3831 data_time: 0.0075 lr: 0.01 max_mem: 3067M

[zhanghang1989]

you may infer the fps based on fps = batch_size / time

[kirkzZ]

If I set the SOLVER.IMS_PER_BATCH', '8' ,so the fps = 8/time

[zhanghang1989]

You may also need to divide it by number of gpus, if more than one is using.

[kirkzZ]

Thanks,in a word:
if I set num_gpus=4 and SOLVER.IMS_PER_BATCH', '8',
and get the log:eta: 9:58:17 iter: 174159 total_loss: 0.9185 loss_cls: 0.229 loss_box_reg: 0.2562 loss_mask: 0.2784 loss_rpn_cls: 0.04507 loss_rpn_loc: 0.0934 time: 0.3831 data_time: 0.0075 lr: 0.01 max_mem: 3067M

so the fps=8/(4*0.3831)