Model Zoo
This page provides a list of commonly used TensorFlow Lite models and their converted DLAs for ready implementation. Each model archive here are given tflite, mdla2 and mdla3 file for inference. Users can only run the version specific DLA model on the supprting platform. Here is the lookup table.
Platform |
MDLA Version |
|---|---|
Genio 350 |
N/A |
Genio 510 |
MDLA 3.0 |
Genio 700 |
MDLA 3.0 |
Genio 1200 |
MDLA 2.0 |
Note
The speed statistics shown here were measured by running the inference on APU of Genio-700 with performance mode enabled. To obtain exact statistics, you should run the APP on the platform. The performance may vary between different versions of the board image and EVK.
Task |
Model Name |
Data Type |
Input Size |
Inference Time(ms) |
Download |
Classification |
MobileNetV1 |
uint8 |
224x224 |
1.0 |
|
Classification |
MobileNetV1 |
float32 |
224x224 |
3.0 |
|
Classification |
MobileNetV2 |
uint8 |
224x224 |
1.0 |
|
Classification |
MobileNetV2 |
float32 |
224x224 |
2.0 |
|
Classification |
MobileNetV3-Large |
uint8 |
224x224 |
1.0 |
|
Classification |
MobileNetV3-Large |
float32 |
224x224 |
4.0 |
|
Classification |
InceptionV3 |
uint8 |
299x299 |
6.0 |
|
Classification |
InceptionV3 |
float32 |
299x299 |
22.0 |
|
Classification |
EfficientNet |
uint8 |
224x224 |
1.0 |
|
Classification |
EfficientNet |
float32 |
224x224 |
3.0 |
|
Classification |
ResNetV1-50 |
uint8 |
224x224 |
3.0 |
|
Classification |
ResNetV1-50 |
float32 |
224x224 |
12.0 |
|
Classification |
VGG16 |
uint8 |
224x224 |
27.0 |
|
Classification |
VGG16 |
float32 |
224x224 |
54.0 |
|
Detection |
SSD-MobileNetV2 |
float32 |
300x300 |
15.0 |
|
Detection |
YoloV2 |
uint8 |
416x416 |
14.0 |
|
Detection |
YoloV2 |
uint8 |
608x608 |
27.3 |
|
Detection |
YoloV2 |
uint8 |
1920x1088 |
133.0 |
|
Detection |
YoloV3 |
uint8 |
416x416 |
28.0 |
|
Detection |
YoloV3 |
uint8 |
608x608 |
53.0 |
|
Detection |
YoloV3 |
uint8 |
1920x1088 |
278.0 |
|
Detection |
YoloV3-Tiny |
uint8 |
608x608 |
5.0 |
|
Detection |
YoloV4-Tiny |
float32 |
640x640 |
11.0 |
|
Detection |
YoloV5S |
uint8 |
320x320 |
3.0 |
|
Pose Estimation |
PoseNet-MobileNetV1 |
float32 |
257x257 |
5.0 |