Emad Barsoum

icon location Engineering Manager, Computer Vision and Deep Learning at Microsoft. Worked on emotion recognition for Cognitive Service and depth pipeline for Kinect

Results 7 comments of Emad Barsoum

Port FasterRCNN models to remove dependence on custom Python layer and export model to ONNX

@GreenShadeZhang use FasterRCNN from PyTorch, this one is exportable to ONNX. The one in CNTK has a lot of Python custom code.

Horovod + DALI

I think this should be part of bigger discussion, should Horovod have data pipeline that work across frameworks and optimized for distributed training?

Inconsistencies between MaxPool, AveragePool, and LpPool specs

storage_order specify if the tensor format is row major or column major, this shouldn't be part of the spec. Regarding dilation, because if you compute average or norm, should you...

Is it trainnable in ml.net

No, you can't train ONNX in ML.NET.

Script to convert Mask R-CNN model to ONNX

@spandantiwari

Trained model

For the frequency look at '_skip_frame' in the reader (https://github.com/ebarsoum/hpgan/blob/master/src/braniac/readers/body.py). It is set in the training script: https://github.com/ebarsoum/hpgan/blob/master/src/train_hpgan.py. For Human3.6m, I skip frame so the frequency is 25 fps. Sadly,...

Cross Op

Which model need it?