tf-faster-rcnn icon indicating copy to clipboard operation
tf-faster-rcnn copied to clipboard

Fine tune Faster RCNN

Open satyakesav opened this issue 6 years ago • 4 comments

Hi,

I want to fine tune the pre trained model on a new data set. Does the command mentioned for training "./experiments/scripts/train_faster_rcnn.sh 0 pascal_voc vgg16" actually tunes the model instead of training from scratch?

satyakesav avatar Apr 17 '18 01:04 satyakesav

If you have a pretrained model you can replace that in data/imagenet_weights, or it will finetune from the imagenet weights.

rnsandeep avatar Apr 18 '18 10:04 rnsandeep

hi, do you know how to finetune all layers especially the conv3, conv4, conv5 before rpn? @rnsandeep @satyakesav ,it would be really helpful to me.

janevril avatar Apr 21 '18 08:04 janevril

hi, do you know how to finetune all layers especially the conv3, conv4, conv5 before rpn? @rnsandeep @satyakesav ,it would be really helpful to me.

have you solved this problem? How do you do?

Fuhan1994 avatar Apr 09 '19 05:04 Fuhan1994

hello,i wonder if this code can be fine-tuned for the last few conv layers

Fuhan1994 avatar Apr 09 '19 06:04 Fuhan1994