matterport / Mask_RCNN

Mask R-CNN for object detection and instance segmentation on Keras and TensorFlow
Other
24.66k stars 11.7k forks source link

How to use other backbones in this implementations #1339

Open crazyxprogrammer opened 5 years ago

crazyxprogrammer commented 5 years ago

I have annotated my huge traning datasets using VGG anotator tools to train MaskRCNN model implemented by matterplot but in matterplot implementation only two backbones are use ResNet 50 and ResNet 101(FPN).But now I want to use other backbones also in mask RCNN.So how I will use different backbones like inception v2 in this implementation

baichii commented 5 years ago

if your dataset is big enough, you can change the code in resnet_graph to fit for inception v2, and make sure your output's shape is the same before, but its really a huge work

crazyxprogrammer commented 5 years ago

if you already done this then please share ur code because I am beginer in coding

CSU-NXY commented 5 years ago

You can have a look at this repository which added Mobilenet to Mask RCNN. It may give you some help.

ectg commented 5 years ago

Has anyone used resnet18 as the backbone instead of resent101 or resnet50? I am not quite sure how to change the resnet_graph function to this end.

tongxiangwang commented 4 years ago

Did you do it? Now I have this problem?

jatinrastogi commented 3 years ago

Did anyone use resnet18? If yes please tell

duxuan11 commented 2 years ago

I can use resnet 18 and resnet 101 as backbone

prayagpawar commented 1 year ago

I can use resnet 18 and resnet 101 as backbone

even i have the same question

prayagpawar commented 1 year ago

Did anyone use resnet18? If yes please tell

not yet refer this https://github.com/alexander-pv/maskrcnn_tf2 let me know if u can implement this?