Our Mission for embedded AI
In our mission to unleash AI on the edge, Datakalab has applied its proprietary quantization method on classical architecture and converted everythong to run in OpenVino® format for seamless compatibility with Intel® hardware such as Intel® core of 12th generation or Xeon® server
Download any of the quantized models below to put your deep learning models on Intel® hardware. Enjoy and let us know what you think!
- Intel® Deep Learning Boost (Intel® DL Boost)
- Intel® AVX-512
- Intel® AVX-2
If you are looking for further compression methods (continual learning, pruning, batch norm folding, etc.) to fit your models on Intel devices, we’d love to hear from you!
Latency on I5 12500TE® with less than 1% of accuracy drop
| Model name | <p align="center"> FPS in int8 | <p align="center"> FPS in fp32 | <p align="center"> Performance Ratio[int8/fp32] | <p align="center"> Relative Accuracy Drop[(fp32-int8)/fp32] | <p align="center"> Memory Size Ratio [fp32/int8] |
|---|---|---|---|---|---|
| Mobilenet V2 | <p align="center"> 1057 | <p align="center"> 424 | <p align="center"> x 2.5 | <p align="center"> 0,12% | <p align="center"> x 3.6 |
| Resnet50 | <p align="center"> 263 | <p align="center"> 76 | <p align="center"> x 3.5 | <p align="center"> 0,13% | <p align="center"> x 3.9 |
| Efficientnetlite b0 | <p align="center"> 1165 | <p align="center"> 659 | <p align="center"> x 1.8 | <p align="center"> 0,1% | <p align="center"> x 3.7 |
| YoloV3 | <p align="center"> 34.25 | <p align="center"> 9.75 | <p align="center"> x 3.5 | <p align="center"> 0.93% <sup>1</sup> | <p align="center"> x 3.75 |
<sup>1</sup> Accuracy Ratio on mAP results
Thanks
All those benchmarks where made using Intel® DevCloud Thanks a lot @Intel®
Contact us
Feel free to reach out by email at hello at datakalab.com