search

ShuangXieIrene / ssds.pytorch

Repository for Single Shot MultiBox Detector and its variants, implemented with pytorch, python3.
MIT License
568 stars 165 forks source link

Any one test the accuracy on VOC dataset? #41

Open blueardour opened 5 years ago

blueardour commented 5 years ago

Hi, everyone, I test the pre-trained model on the VOC dataset, and found the mAP didn't match the expected ones. Here is my result: yolo_v2_mobilenetv2_voc: 0.721956 yolo_v3_mobilenetv2_voc: 0.766846 ssd_lite_mobilenetv2_voc: 0.708510 rfb_lite_mobilenetv2_voc: 0.713691 fssd_lite_mobilenetv2_voc: 0.744321 It can be seen, yolov2/3 perform better than the author's result, however the other three are bad than the expected result.

Have anyone met this problem?

Brozhao commented 5 years ago

Did you test the mAP of SSD+VGG+VOC ? Mine is just 0.47

Brozhao commented 5 years ago

also my ssd_lite_mobilenetv2_voc is just 0.66

blueardour commented 5 years ago

I only focus on mobilenetv2. Which pytorch version and python version do you employ?

Brozhao commented 5 years ago

Python 3.5 torch 3.1 When I tried the other SSD net,it is right.So the version I think is doesn`t matter.Did you used the pretraining model?

blueardour commented 5 years ago

Yes, pre-trained model was used. My environment is python2.7/3.6 and pytorch 0.4.1

Brozhao commented 5 years ago

Could you tell me the learning rate and max epoch of yours ? Thanks a lot!

rw1995 commented 5 years ago

Could you tell me the learning rate and max epoch of yours ? Thanks a lot!

do you slove?