out of memory error

hello dears ? i am trying to use Faster R-CNN which used a resnet50 as a pretrained model and has a network input size [224 224 3] however my video frames has a high resolution 1900 * 1800 when I am using the model Object Detection Using Faster R-CNN Deep Learning - MATLAB & Simulink (mathworks.com) which reduced video frame size after preprocessing to 224 * 224 , I didnt get good detector after traning , even in some videos it dosnt recognize anything then I tried training without reszing to 224* 224 i got out of memory however i had GPU of 8 GB .so what are your advice to do with high reolsution videos to get good detectors ????

Answer 1

[yanqi liu]

yes，sir，may be set the MiniBatchSize=1 before train

[tara saber]

Dear sir i changed minibatchsize to 1 also i got same error. My video frames has a high resolution when i reduced to 224×224 as pretrained resnet50 it is dector has very less accuarcy and cant recognize anything what is your suggestion please help ....? Thanks dear yanqi liu