Add paddle mobilenets (#3412)

Author: eaidova

demos/classification_benchmark_demo/cpp/README.md

@@ -63,6 +63,8 @@ omz_converter --list models.lst
 * mobilenet-v2-pytorch
 * mobilenet-v3-large-1.0-224-tf
 * mobilenet-v3-small-1.0-224-tf
+* mobilenet-v3-large-1.0-224-paddle
+* mobilenet-v3-small-1.0-224-paddle
 * nfnet-f0
 * octave-resnet-26-0.25
 * regnetx-3.2gf

demos/classification_benchmark_demo/cpp/models.lst

@@ -27,6 +27,8 @@ mobilenet-v2-1.4-224
 mobilenet-v2-pytorch
 mobilenet-v3-large-1.0-224-tf
 mobilenet-v3-small-1.0-224-tf
+mobilenet-v3-large-1.0-224-paddle
+mobilenet-v3-small-1.0-224-paddle
 nfnet-f0
 octave-resnet-26-0.25
 regnetx-3.2gf

demos/classification_demo/python/README.md

@@ -67,6 +67,8 @@ omz_converter --list models.lst
 * mobilenet-v2-pytorch
 * mobilenet-v3-large-1.0-224-tf
 * mobilenet-v3-small-1.0-224-tf
+* mobilenet-v3-large-1.0-224-paddle
+* mobilenet-v3-small-1.0-224-paddle
 * nfnet-f0
 * octave-resnet-26-0.25
 * regnetx-3.2gf

demos/classification_demo/python/models.lst

@@ -29,6 +29,8 @@ mobilenet-v2-1.4-224
 mobilenet-v2-pytorch
 mobilenet-v3-large-1.0-224-tf
 mobilenet-v3-small-1.0-224-tf
+mobilenet-v3-large-1.0-224-paddle
+mobilenet-v3-small-1.0-224-paddle
 nfnet-f0
 octave-resnet-26-0.25
 regnetx-3.2gf

models/public/device_support.md

@@ -78,6 +78,8 @@
 | mobilenet-v2-pytorch | YES | YES | YES |
 | mobilenet-v3-large-1.0-224-tf | YES | YES | YES |
 | mobilenet-v3-small-1.0-224-tf | YES | YES | YES |
+| mobilenet-v3-large-1.0-224-paddle | YES | YES | YES |
+| mobilenet-v3-small-1.0-224-paddle | YES | YES | YES |
 | mobilenet-yolo-v4-syg | YES | YES |    |
 | modnet-photographic-portrait-matting | YES | YES | YES |
 | modnet-webcam-portrait-matting | YES | YES | YES |

models/public/index.md

@@ -71,7 +71,9 @@
    omz_models_model_mobilenet_v2_1_0_224
    omz_models_model_mobilenet_v2_1_4_224
    omz_models_model_mobilenet_v2_pytorch
+   omz_models_model_mobilenet_v3_large_1_0_224_paddle
    omz_models_model_mobilenet_v3_large_1_0_224_tf
+   omz_models_model_mobilenet_v3_small_1_0_224_paddle
    omz_models_model_mobilenet_v3_small_1_0_224_tf
    omz_models_model_nfnet_f0
    omz_models_model_octave_resnet_26_0_25
@@ -351,8 +353,8 @@ You can download models and convert them into OpenVINO™ IR format (\*.xml + \*
 | MobileNet V1 1.0 224        | Caffe\*<br>TensorFlow\*            | [mobilenet-v1-1.0-224](./mobilenet-v1-1.0-224/README.md)<br>[mobilenet-v1-1.0-224-tf](./mobilenet-v1-1.0-224-tf/README.md)| 69.496%/89.224%<br>71.03%/89.94% | 1.148 | 4.221 |
 | MobileNet V2 1.0 224        | Caffe\*<br>TensorFlow\*<br>PyTorch\*| [mobilenet-v2](./mobilenet-v2/README.md) <br>[mobilenet-v2-1.0-224](./mobilenet-v2-1.0-224/README.md)<br>[mobilenet-v2-pytorch](./mobilenet-v2-pytorch/README.md) | 71.218%/90.178%<br>71.85%/90.69%<br>71.81%/90.396% | 0.615~0.876 | 3.489 |
 | MobileNet V2 1.4 224        | TensorFlow\*                       | [mobilenet-v2-1.4-224](./mobilenet-v2-1.4-224/README.md) | 74.09%/91.97% | 1.183  | 6.087 |
-| MobileNet V3 Small 1.0      | TensorFlow\*                       | [mobilenet-v3-small-1.0-224-tf](./mobilenet-v3-small-1.0-224-tf/README.md) | 67.36%/87.44% | 0.1168 | 2.537 |
-| MobileNet V3 Large 1.0      | TensorFlow\*                       | [mobilenet-v3-large-1.0-224-tf](./mobilenet-v3-large-1.0-224-tf/README.md) | 75.30%/92.62% | 0.4450 | 5.4721 |
+| MobileNet V3 Small 1.0      | TensorFlow\*<br>Paddle\* | [mobilenet-v3-small-1.0-224-tf](./mobilenet-v3-small-1.0-224-tf/README.md) <br> [mobilenet-v3-small-1.0-paddle](./mobilenet-v3-small-1.0-224-paddle/README.md) | 67.36%/87.44%<br>68.21%/88.04% | 0.1168<br>0.1269 | 2.537<br>2.9339 |
+| MobileNet V3 Large 1.0      | TensorFlow\*<br>Paddle\*                   | [mobilenet-v3-large-1.0-224-tf](./mobilenet-v3-large-1.0-224-tf/README.md)<br>[mobilenet-v3-large-1.0-224-paddle](./mobilenet-v3-large-1.0-224-paddle/README.md) | 75.30%/92.62%<br>75.248%/92.32% | 0.4450<br>0.4565 | 5.4721<br>5.468 |
 | NFNet F0                    | PyTorch\*                          | [nfnet-f0](./nfnet-f0/README.md) | 83.34%/96.56% | 24.8053 | 71.4444 |
 | RegNetX-3.2GF               | PyTorch\*                          | [regnetx-3.2gf](./regnetx-3.2gf/README.md) | 78.17%/94.08% | 6.3893 | 15.2653 |
 | ResNet 26, alpha=0.25       | MXNet\*                            | [octave-resnet-26-0.25](./octave-resnet-26-0.25/README.md)     | 76.076%/92.584%| 3.768 | 15.99 |

models/public/mobilenet-v3-large-1.0-224-paddle/README.md

@@ -0,0 +1,88 @@
+# mobilenet-v3-large-1.0-224-paddle
+
+## Use Case and High-Level Description
+
+`mobilenet-v3-large-1.0-224-paddle` is one of MobileNets V3 - next generation of MobileNets,
+based on a combination of complementary search techniques as well as a novel architecture design.
+`mobilenet-v3-large-1.0-224-paddle` is pretrained in Paddle\* framework and targeted for high resource use cases.
+For details see [paper](https://arxiv.org/abs/1905.02244) and [repository](https://github.com/PaddlePaddle/PaddleClas).
+
+## Specification
+
+| Metric                          | Value                                     |
+|---------------------------------|-------------------------------------------|
+| Type                            | Classification                            |
+| GFlops                          | 0.4565                                    |
+| MParams                         | 5.468                                     |
+| Source framework                | Paddle\*                                  |
+
+## Accuracy
+
+| Metric | Result         |
+| ------ | -------------- |
+| Top 1  | 75.248%        |
+| Top 5  | 92.32%         |
+
+## Input
+
+### Original Model
+
+Image, name: `x`, shape: `1, 3, 224, 224`, format: `B, C, H, W`, where:
+
+- `B` - batch size
+- `C` - number of channels
+- `H` - image height
+- `W` - image width
+
+Expected color order: `RGB`.
+Mean values - [123.675,116.28,103.53], scale values - [58.395, 57.12, 57.375].
+
+### Converted Model
+
+Image, name: `x`, shape: `1, 3, 224, 224`, format: `B, C, H, W`, where:
+
+- `B` - batch size
+- `C` - number of channels
+- `H` - image height
+- `W` - image width
+
+Expected color order: `BGR`.
+
+## Output
+
+### Original Model
+
+Object classifier according to ImageNet classes, name - `softmax_1.tmp_0`,  shape - `1, 1000`, output data format is `B, C` where:
+
+- `B` - batch size
+- `C` - predicted probabilities for each class in [0, 1] range
+
+### Converted Model
+
+The converted model has the same parameters as the original model.
+
+## Download a Model and Convert it into OpenVINO™ IR Format
+
+You can download models and if necessary convert them into OpenVINO™ IR format using the [Model Downloader and other automation tools](../../../tools/model_tools/README.md) as shown in the examples below.
+
+An example of using the Model Downloader:
+```
+omz_downloader --name <model_name>
+```
+
+An example of using the Model Converter:
+```
+omz_converter --name <model_name>
+```
+
+## Demo usage
+
+The model can be used in the following demos provided by the Open Model Zoo to show its capabilities:
+
+* [Classification Benchmark C++ Demo](../../../demos/classification_benchmark_demo/cpp/README.md)
+* [Classification Python\* Demo](../../../demos/classification_demo/python/README.md)
+
+## Legal Information
+
+The original model is distributed under the
+[Apache License, Version 2.0](https://raw.githubusercontent.com/PaddlePaddle/PaddleClas/release/2.3/LICENSE).

models/public/mobilenet-v3-large-1.0-224-paddle/accuracy-check.yml

@@ -0,0 +1,24 @@
+models:
+  - name: mobilenet-v3-large-1.0-224-paddle
+    launchers:
+      - framework: openvino
+        adapter: classification
+
+    datasets:
+      - name: imagenet_1000_classes
+        preprocessing:
+          - type: resize
+            size: 256
+            aspect_ratio_scale: greater
+          - type: crop
+            size: 224
+
+        metrics:
+          - name: accuracy@top1
+            type: accuracy
+            top_k: 1
+            reference: 0.75248
+          - name: accuracy@top5
+            type: accuracy
+            top_k: 5
+            reference: 0.9232

models/public/mobilenet-v3-large-1.0-224-paddle/model.yml

@@ -0,0 +1,58 @@
+# Copyright (c) 2022 Intel Corporation
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+description: >-
+  "mobilenet-v3-large-1.0-224-paddle" is one of MobileNets V3 - next generation of MobileNets,
+  based on a combination of complementary search techniques as well as a novel architecture
+  design. "mobilenet-v3-large-1.0-224-paddle" is pretraind in Paddle framework and targeted for low resource use cases.
+  For details see paper <https://arxiv.org/abs/1905.02244> and repository <https://github.com/PaddlePaddle/PaddleClas>.
+task_type: classification
+files:
+  - name: paddleclas-2.2.1-py3-none-any.whl
+    size: 306598
+    checksum: 16b0ea3a97393b17e40c657fce1b83ef78a8accef34e721cf6691151da1730efdbc7018635053636a023a2ca5abb3e4b
+    source: https://files.pythonhosted.org/packages/ba/2d/7e76250b345157c60639e9bbb301a39f1796cf882ac887a082245e54fa7d/paddleclas-2.2.1-py3-none-any.whl
+  - name: MobileNetV3_large_x1_0_pretrained.pdparams
+    size: 33273297
+    checksum: 4b5f60288b6b4e8a394c2c7b8e6aaaa17d02ab3541ce4589268626a2ec78beb5deeccc956c16b602412e7d687d2b2327
+    original_source: https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV3_large_x1_0_pretrained.pdparams
+    source: https://storage.openvinotoolkit.org/repositories/open_model_zoo/public/2022.2/mobilenet-v3-large-1.0-224-paddle/MobileNetV3_large_x1_0_pretrained.pdparams
+  - name: tools/export_model.py
+    size: 3635
+    checksum: cf7563115a0ec676d989525500f67dc30dd52f6d7771ffe91e58538fed9479a9a134bb1168311ea71fe0bf29b739c561
+    source: https://raw.githubusercontent.com/PaddlePaddle/PaddleClas/7c1f291b88914a862b824393ba00d8f54269a73e/tools/export_model.py
+  - name: MobileNetV3_large_x1_0.yaml
+    size: 2768
+    checksum: 3d7e066070fafdb3f01d49b3b6f422b96d93cf895dff89017accc583b034ad72be2c042e0db88fecee4d57fa74f4ba1f
+    source: https://raw.githubusercontent.com/PaddlePaddle/PaddleClas/d6d5efe055caa9e08d0c392fcc02b49319afe14a/ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x1_0.yaml
+postprocessing:
+  - $type: unpack_archive
+    format: zip
+    file: paddleclas-2.2.1-py3-none-any.whl
+  - $type: regex_replace
+    file: tools/export_model.py
+    pattern: sys.path.append\(os.path.abspath\(os.path.join\(__dir__, '..\/'\)\)\)
+    replacement: sys.path.append(os.path.abspath(os.path.join(__dir__, '../paddleclas')))
+input_info:
+  - name: x
+    shape: [1, 3, 224, 224]
+    layout: NCHW
+model_optimizer_args:
+  - --reverse_input_channels
+  - --scale_values=[58.395, 57.12, 57.375]
+  - --mean_values=[123.675, 116.28, 103.53]
+  - --output=softmax_1.tmp_0
+  - --input_model=$conv_dir/inference_model/inference.pdmodel
+framework: paddle
+license: https://raw.githubusercontent.com/PaddlePaddle/PaddleClas/release/2.3/LICENSE

models/public/mobilenet-v3-large-1.0-224-paddle/pre-convert.py

@@ -0,0 +1,40 @@
+#!/usr/bin/env python3
+
+# Copyright (c) 2022 Intel Corporation
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import argparse
+import subprocess # nosec - disable B404:import-subprocess check
+import sys
+
+from pathlib import Path
+
+
+def main():
+    parser = argparse.ArgumentParser()
+    parser.add_argument('input_dir', type=Path)
+    parser.add_argument('output_dir', type=Path)
+    args = parser.parse_args()
+
+    subprocess.run([sys.executable, '--',
+        str(args.input_dir / 'tools' / 'export_model.py'), '-c',
+        str(args.input_dir / 'MobileNetV3_large_x1_0.yaml'), '-o',
+        f"Global.pretrained_model={args.input_dir/ 'MobileNetV3_large_x1_0_pretrained'}",
+        '-o', f"Global.save_inference_dir={args.output_dir / 'inference_model'}",
+        '-o', 'Global.device=cpu'
+    ], check=True)
+
+
+if __name__ == '__main__':
+    main()