Name
Abstract | ||
|---|---|---|
Convolutional Neural Networks (CNNs) have gained significant traction in the field of machine learning, particularly due to their high accuracy in visual recognition. Recent works have pushed the performance of GPU implementations of CNNs showing significant improvements in their classification and training times. With these improvements, many frameworks have become available for implementing CNNs on both CPUs and GPUs, with no support for FPGA implementations. In this work we present a modified version of the popular CNN framework Caffe, with FPGA support. This allows for classification using CNN models and specialized FPGA implementations with the flexibility of reprogramming the device when necessary, seamless memory transactions between host and device, simple-to-use test benches, and the ability to create pipelined layer implementations. To validate the framework, we use the Xilinx SDAccel environment to implement an FPGA-based Winograd convolution engine and show that it can be used alongside other layers running on a host processor to run several popular CNNs (AlexNet, GoogleNet, VGG A, Overfeat). The results show that our framework achieves 50 GFLOPS across 3×3 convolutions in the benchmarks. This is achieved within a practical framework, which will aid in future development of FPGA-based CNNs. |
Year | DOI | Venue |
|---|---|---|
2016 | 10.1109/FPT.2016.7929549 | 2016 International Conference on Field-Programmable Technology (FPT) |
Keywords | DocType | Volume |
Caffeinated FPGA,convolutional neural networks,machine learning,GPU implementation performance,CNN framework Caffe,visual recognition,Xilinx SDAccel environment,FPGA-based Winograd convolution engine,AlexNet,GoogleNet,VGG A,Overfeat,GFLOPS,FPGA-based CNN | Conference | abs/1609.09671 |
ISBN | Citations | PageRank |
978-1-5090-5603-3 | 6 | 0.60 |
References | Authors | |
0 | 6 | |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Roberto DiCecco | 1 | 7 | 0.94 |
| Griffin Lacey | 2 | 6 | 0.93 |
| Jasmina Vasiljevic | 3 | 23 | 3.17 |
| Paul Chow | 4 | 868 | 119.97 |
| Graham W. Taylor | 5 | 1523 | 127.22 |
| Shawki Areibi | 6 | 367 | 41.38 |