Intel® FPGA AI Suite: IP Reference Manual

ID 768974
Date 12/01/2023
Public
Document Table of Contents

2.1. Supported Models

The Intel® FPGA AI Suite supports the following pretrained models from OpenVINO™ Model Zoo:

Table 2.   OpenVINO™ Model Zoo 2022.3.1 LTS

Model Zoo 2022.3.1 LTS path

Model

Framework

public/mobilenet-v1-1.0-224

MobileNet V1

Caffe

public/mobilenet-v2

MobileNet V2

Caffe

public/mobilenet-v2-1.4-224

MobileNet V2

TensorFlow

public/mobilenet-v3-large-1.0-224-tf

MobileNet V3

TensorFlow

public/resnet-50-tf

ResNet-50

TensorFlow

intel/unet-camvid-onnx-0001

UNet

PyTorch

public/yolo-v3-tf

YOLO v3

TensorFlow

public/yolo-v3-tiny-tf

TinyYOLO v3

TensorFlow

public/squeezenet1.1 SqueezeNet v1.1 Caffe
public/i3d_rgb_tf Inflated 3D (I3D) TensorFlow
  Multilayer Perceptrons (MLPs)  

After installation, the location $COREDLA_ROOT/example_graphs/MLP/ contains an example MLP graph.

In addition, the Intel® FPGA AI Suite supports certain signed INT8 symmetric quantized models that use neural network compression framework (NNCF) flows.

Customized models that are similar to the supported models or are derived from the above models are not supported, although in some cases they might work without modification.

The supplied example architectures (or IP Configurations) support all of the above models, except for the Small and Small_Softmax architectures that support only ResNet-50, MobileNet V1, and MobileNet V2.