Closed goloskokovic closed 4 years ago
This issue has been automatically marked as stale due to inactivity and will be closed in 7 days if no further activity occurs. If further support is needed, please provide an update and/or more details.
This issue has been automatically closed due to inactivity. Please reactivate if further support is needed.
Improve performance for Object Detection models on ARM architecture using XNOR-Networks: Binary Neural Networks: A Survey XNOR-Net: ImageNet Classification Using Binary Convolutional Neural Networks
@mcaraman @Andrews548
@jywu-msft @prabhat00155
System information ONNX Runtime version 1.3.1
Describe the solution you'd like Develop SOTA on-device binary ONNX model support for ARMv8 that enable business to real-time decisions, deliver more efficient experiences to customers:
Mobilenet v1, Tensorflow, Arm NN on RPi4 Execution time : 209 ms XNOR binary model, Xnor.ai platform on RPi4 Execution time : 79.5 ms
Describe alternatives you've considered OpenVINO Toolkit Xnor.ai
Additional context ONNX model: resnet18_imagenet_binarization_xnor.onnx