How to calculate mAP and FPS using SSD-MobileNet?

Hello everyone!

I re-train SSD using this tutorial jetson-inference/pytorch-ssd.md at master · dusty-nv/jetson-inference · GitHub .

When I ran the following: python3 eval_ssd.py --net=mb1-ssd --trained_model=models/weeds/mb1-ssd-Epoch-94-Loss-inf.pth --dataset=data/weeds_crops_flips_voc --label_file=models/weeds/labels.txt
I got this messages:

process image 108
Load Image: 0.009654 seconds.
Inference time:  0.028708696365356445
Prediction: 0.196644 seconds.
process image 109
Load Image: 0.009927 seconds.
Inference time:  0.029082536697387695
Prediction: 0.246111 seconds.


Average Precision Per-class:
Ambr_box: 0.46374420458803794
Can_box: 0.18316823301583887

Average Precision Across All Classes: 0.3234562188019384

How to find out mAP and FPS from this?
I’m new to deep learning, please help me

Thank you!

Hi,

The fps can be approximated by the inverse of prediction time.
The average precision should be the per-class mAP already.

You can find the detailed calculation fumula below:

Thanks.

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.