search

microsoft / CNTK

Microsoft Cognitive Toolkit (CNTK), an open source deep-learning toolkit
https://docs.microsoft.com/cognitive-toolkit/
Other
17.52k stars 4.28k forks source link

Using other base models in faster rcnn #2202

Closed fathomson closed 7 years ago

fathomson commented 7 years ago

By default the faster rcnn has a config for the AlexNet and VGG16 base models. I tried to get it to work with ResNet however I was not able to find the correct config parameters. I've used print parameter and plot() to look for the required node names, however not successfull. How can I get faster rcnn to work with i.e. ResNet base model.

pkranen commented 7 years ago

ResNet is not yet supported since currently we only have ROI max pooling and not ROI average pooling.

sooraj-citta commented 6 years ago

The documentation (https://docs.microsoft.com/en-us/cognitive-toolkit/Object-Detection-using-Faster-R-CNN) explains about only the models Alexnet & VGGnet. I can download the model file 'InceptionV3_ImageNet_CNTK' but where to get the configuration file for the inception. Any hints would be helpful!