simon comments

Repositories
Issues
Comments

Results 4 comments of


                                            simon

Extract features from various layers of MaskRCNN, such as ROI align features, pyramid features [P1,P2,P3,P4,P5]

you mean the output featuremap in every layer of mrcnn architecture?

Mask_RCNN Backbone from Resnet 101 to Resnet152 problem

would you training the mask rcnn with resnet-152 backbone model to tune its weight from begining to end instead of use the original .h5 file in matterport github of weight...

Mask_RCNN Backbone from Resnet 101 to Resnet152 problem

that coco h5 model seems to be for use of resnet101 or resnet50 model to continue training and object detection, if you want to have an actual totally resnet152 h5...

Mask_RCNN Backbone from Resnet 101 to Resnet152 problem

ok, now i clearly know your problem, i cant find the resnet152 trained on coco dataset h5 file either, so you had to train your self from begining to end...