Overview of OpenVINO™ Toolkit Intel’s Pre-Trained Models¶
OpenVINO toolkit provides a set of Intel’s pre-trained models that you can use for learning and demo purposes or for developing deep learning software. Most recent version is available in the repo on GitHub. The table Intel’s Pre-Trained Models Device Support summarizes devices supported by each model.
The models can be downloaded via Model Downloader.
Object Detection Models¶
Model Name |
Framework |
Accuracy |
GFlops |
mParams |
---|---|---|---|---|
TensorFlow |
38.74% |
849.91 |
52.79 |
|
Caffe |
94.10% |
2.835 |
1.053 |
|
Caffe |
83.00% |
1.067 |
0.588 |
|
PyTorch |
84.52% |
0.982 |
1.021 |
|
PyTorch |
86.74% |
0.786 |
1.828 |
|
PyTorch |
91.94% |
1.767 |
1.842 |
|
PyTorch |
92.89% |
2.406 |
1.851 |
|
PyTorch |
93.57% |
2.853 |
2.392 |
|
PyTorch |
94.27% |
339.60 |
69.92 |
|
Caffe |
80.14% |
12.427 |
3.244 |
|
Caffe |
88.62% |
2.300 |
0.723 |
|
Caffe |
83.8% |
7.140 |
1.951 |
|
TensorFlow |
80.74% |
8.225 |
2.001 |
|
Caffe |
72.4% |
7.140 |
1.951 |
|
Caffe |
90.5% |
7.138 |
1.951 |
|
PyTorch |
0.244 |
0.786 |
1.817 |
|
PyTorch |
0.299 |
1.768 |
1.817 |
|
PyTorch |
0.328 |
3.143 |
1.817 |
|
PyTorch |
0.408 |
6.519 |
2.394 |
|
PyTorch |
0.442 |
404.264 |
71.565 |
|
Caffe |
88% |
2.836 |
1.165 |
|
Caffe |
90% |
3.974 |
1.650 |
|
Caffe |
90.6% |
2.798 |
1.079 |
|
PyTorch |
0.254 |
0.786 |
1.817 |
|
PyTorch |
0.323 |
1.768 |
1.817 |
|
PyTorch |
0.36 |
3.143 |
1.817 |
|
Caffe |
65.12% |
3.964 |
1.178 |
|
PyTorch |
62.55% |
3.560 |
2.887 |
|
Keras |
48.89% |
65.98 |
61.92 |
|
PyTorch |
0.165 |
0.787 |
1.821 |
|
PyTorch |
0.226 |
1.770 |
1.821 |
|
PyTorch |
0.248 |
3.146 |
1.821 |
|
PyTorch |
0.336 |
6.550 |
2.416 |
|
PyTorch |
0.274 |
1.811 |
2.327 |
|
TensorFlow |
99.65% |
0.349 |
0.634 |
|
PyTorch |
0.715 |
3.598 |
3.212 |
|
PyTorch |
80.0% |
0.986 |
1.338 |
|
TensorFlow |
63.9% |
29.421 |
50.6451 |
|
TensorFlow |
63.71% |
29.421 |
50.6451 |
|
TensorFlow |
62.9% |
29.421 |
50.6451 |
|
TensorFlow |
35.37% |
6.9949 |
15.8587 |
|
TensorFlow |
36.33% |
6.9949 |
15.8587 |
|
TensorFlow |
35.32% |
6.9949 |
15.8587 |
|
Keras |
88.64% |
5.424 |
11.229 |
Object Recognition Models¶
Model Name |
Framework |
Accuracy |
GFlops |
mParams |
---|---|---|---|---|
Caffe |
95.80% |
0.094 |
2.138 |
|
Caffe |
0.105 |
1.911 |
||
Caffe |
88.58% |
0.328 |
1.218 |
|
Caffe |
81.15% |
0.126 |
0.626 |
|
PyTorch |
82.71% |
0.462 |
11.177 |
|
Caffe |
70.20% |
0.126 |
2.483 |
|
PyTorch |
0.0705 |
0.021 |
0.191 |
|
Caffe |
0.042 |
4.595 |
||
PyTorch |
0.174 |
0.735 |
||
PyTorch |
2.167 |
23.510 |
||
PyTorch |
1.034 |
21.797 |
||
PyTorch |
0.139 |
1.882 |
Reidentification Models¶
Model Name |
Framework |
Accuracy |
GFlops |
mParams |
---|---|---|---|---|
PyTorch |
0.9947 |
0.588 |
1.107 |
|
PyTorch |
86.1% |
0.174 |
0.183 |
|
PyTorch |
92.9% |
0.564 |
0.595 |
|
PyTorch |
94.8 % |
1.170 |
1.234 |
|
PyTorch |
96.2 % |
1.993 |
2.103 |
Semantic Segmentation Models¶
Model Name |
Framework |
Accuracy |
GFlops |
mParams |
---|---|---|---|---|
PyTorch |
0.901 |
4.770 |
0.184 |
|
Caffe |
0.6907 |
58.572 |
6.686 |
|
PyTorch |
71.95% |
260.1 |
31.03 |
|
PyTorch |
75.42% |
75.818 |
26.704 |
|
PyTorch |
75.87% |
75.818 |
26.704 |
|
PyTorch |
75.79% |
75.818 |
26.704 |
Instance Segmentation Models¶
Model Name |
Framework |
Accuracy |
GFlops |
mParams |
---|---|---|---|---|
PyTorch |
40.8% |
423.08 |
48.373 |
|
PyTorch |
45.8% |
828.63 |
101.24 |
|
PyTorch |
38.85% |
147.24 |
49.833 |
|
PyTorch |
32.9% |
13.97 |
10.567 |
|
PyTorch |
35.0% |
29.33 |
13.567 |
Human Pose Estimation Models¶
Model Name |
Framework |
Accuracy |
GFlops |
mParams |
---|---|---|---|---|
Caffe |
42.8% |
15.435 |
4.099 |
|
PyTorch |
45.6% |
5.9206 |
8.1504 |
|
PyTorch |
51.1% |
8.844 |
8.1504 |
|
PyTorch |
54.3% |
14.3253 |
8.1504 |
Image Processing¶
Model Name |
Framework |
Accuracy |
GFlops |
mParams |
---|---|---|---|---|
PyTorch |
11.654 |
0.030 |
||
PyTorch |
16.062 |
0.030 |
||
PyTorch |
1.379 |
0.030 |
Text Detection¶
Model Name |
Framework |
Accuracy |
GFlops |
mParams |
---|---|---|---|---|
TensorFlow |
82.12% |
51.256 |
6.747 |
|
TensorFlow |
79.43% |
23.305 |
4.328 |
|
PyTorch |
88.45% |
7.718 |
2.259 |
Text Recognition¶
Model Name |
Framework |
Accuracy |
GFlops |
mParams |
---|---|---|---|---|
TensorFlow |
0.8818 |
1.485 |
5.568 |
|
PyTorch |
0.8887 |
0.2726 |
1.4187 |
|
text-recognition-0015 - encoder |
PyTorch |
12.4 |
398 |
|
text-recognition-0015 - decoder |
PyTorch |
0.03 |
4.33 |
|
TensorFlow |
98.83% |
0.792 |
5.555 |
|
PyTorch |
117.136 |
117.136 |
15.31 |
|
PyTorch |
75.31% |
134.513 |
17.270 |
|
formula-recognition-medium-scan-0001 - encoder |
PyTorch |
81.5% |
16.56 |
1.86 |
formula-recognition-medium-scan-0001 - decoder |
PyTorch |
81.5% |
1.69 |
2.56 |
PyTorch |
70.5% |
12.8447 |
0.2017 |
|
PyTorch |
70.5% |
0.2017 |
2.5449 |
Text Spotting¶
Model Name |
Framework |
Accuracy |
GFlops |
mParams |
---|---|---|---|---|
text-spotting-0005 - detector |
PyTorch |
71.29% |
184.495 |
27.010 |
text-spotting-0005 - recognizer-encoder |
PyTorch |
71.29% |
2.082 |
1.328 |
text-spotting-0005 - recognizer-decoder |
PyTorch |
71.29% |
0.106 |
0.283 |
Action Recognition Models¶
Model Name |
Framework |
Accuracy |
GFlops |
mParams |
---|---|---|---|---|
driver-action-recognition-adas-0002 - encoder |
PyTorch |
0.676 |
2.863 |
|
driver-action-recognition-adas-0002 - decoder |
PyTorch |
0.147 |
4.205 |
|
action-recognition-0001 - encoder |
PyTorch |
7.340 |
21.276 |
|
action-recognition-0001 - decoder |
PyTorch |
0.147 |
4.405 |
|
PyTorch |
0.847 |
6.660 |
4.133 |
|
PyTorch |
98.00% |
4.227 |
4.113 |
|
PyTorch |
97.14% |
3.636 |
11.173 |
Image Retrieval¶
Model Name |
Framework |
Accuracy |
GFlops |
mParams |
---|---|---|---|---|
TensorFlow |
0.834 |
0.613 |
2.535 |
Compressed models¶
Model Name |
Framework |
Accuracy |
GFlops |
mParams |
---|---|---|---|---|
PyTorch |
70.69% |
1.002 |
7.446 |
|
PyTorch |
61.71% |
Question Answering¶
Model Name |
Framework |
Accuracy |
GFlops |
mParams |
---|---|---|---|---|
PyTorch |
87.20% |
246.93 |
333.96 |
|
PyTorch |
86.36% |
246.93 |
333.96 |
|
PyTorch |
90.5% |
246.93 |
333.96 |
|
PyTorch |
85.04% |
23.9 |
57.94 |
|
PyTorch |
85.4% |
23.9 |
41.1 |
|
PyTorch |
84.4% |
23.9 |
41.1 |
|
PyTorch |
87.6% |
23.9 |
41.1 |
Machine Translation¶
Model Name |
Framework |
Accuracy |
GFlops |
mParams |
---|---|---|---|---|
PyTorch |
21.6% |
23.17 |
69.29 |
|
PyTorch |
22.8% |
23.17 |
69.29 |
|
PyTorch |
17.7% |
23.19 |
77.47 |
|
PyTorch |
21.4% |
23.19 |
77.47 |
Text To Speech¶
Model Name |
Framework |
Accuracy |
GFlops |
mParams |
---|---|---|---|---|
text-to-speech-en-0001 - duration-prediction |
PyTorch |
15.84 |
13.569 |
|
text-to-speech-en-0001 - regression |
PyTorch |
7.65 |
4.96 |
|
text-to-speech-en-0001 - generation |
PyTorch |
48.38 |
12.77 |
|
text-to-speech-en-multi-0001 - duration-prediction |
PyTorch |
28.75 |
26.18 |
|
text-to-speech-en-multi-0001 - regression |
PyTorch |
7.81 |
5.12 |
|
text-to-speech-en-multi-0001 - generation |
PyTorch |
48.38 |
12.77 |
Noise suppression¶
Model Name |
Framework |
Accuracy |
GFlops |
mParams |
---|---|---|---|---|
PyTorch |
1.2 |
7.22 |
Time Series Forecasting¶
Model Name |
Framework |
Accuracy |
GFlops |
mParams |
---|---|---|---|---|
PyTorch |
0.40 |
2.26 |
See Also¶
Legal Information¶
Caffe, Caffe2, Keras, MXNet, PyTorch, and TensorFlow are trademarks or brand names of their respective owners. All company, product and service names used in this website are for identification purposes only. Use of these names,trademarks and brands does not imply endorsement.