search

naisy / train_ssd_mobilenet

Train ssd_mobilenet of the Tensorflow Object Detection API with your own data.
MIT License
61 stars 24 forks source link

mAp of mobilenet_ssd #8

Open bubble0000 opened 4 years ago

bubble0000 commented 4 years ago

Hello,I want to know about the precision that the model can achieved.I am training this model with 2500 photos,and the iteration is 300k.When I am evaluating this model,the mAP is 36%~38%,it seems that it could not grow more.So I want to know if there is any improvement ways,or maybe the precision of this model can only be this case.Thanks for answering my questions.

naisy commented 4 years ago

Hi @bubble0000,

Probably the best way is to crop the data and resize it to 300x300 so that the aspect ratio is 1:1. Please crop the image frame to be predicted to have an aspect ratio of 1:1, too.

reference: https://forums.intel.com/s/question/0D50P00004NM0RCSA1/image-aspect-ratio-how-does-it-affect-accuracy?language=en_US