Merge pull request ARM-software#1 from ARM-software/ssd_mobilenet

Added SSD MobileNet v1 FP32 & UINT8

.gitignore

@@ -0,0 +1 @@
+.DS_STORE

MODEL_LIST.md

@@ -16,3 +16,9 @@
 | [DS-CNN Large INT8](models/keyword_spotting/ds_cnn_large/tflite_int8) | INT8 | TensorFlow Lite |
 | [DS-CNN Medium INT8](models/keyword_spotting/ds_cnn_medium/tflite_int8) | INT8 | TensorFlow Lite |
 | [DS-CNN Small INT8](models/keyword_spotting/ds_cnn_small/tflite_int8) | INT8 | TensorFlow Lite |
+
+## Object Detection
+| Model Name |  Datatype  |  Framework |
+|-----------------|---------|--------------|
+| [SSD MobileNet v1 FP32](models/object_detection/ssd_mobilenet_v1/tflite_fp32) | FP32 | TensorFlow Lite |
+| [SSD MobileNet v1 UINT8](models/object_detection/ssd_mobilenet_v1/tflite_uint8) | UINT8 | TensorFlow Lite |

models/object_detection/ssd_mobilenet_v1/tflite_fp32/README.md

@@ -0,0 +1,36 @@
+# SSD MobileNet v1 FP32
+
+## Description
+SSD MobileNet v1 is a object detection network, that localizes and identifies objects in an input image. This is a TF Lite floating point version that takes a 300x300 input image and outputs detections for this image. This model is trained by Google.
+
+## License
+[Apache-2.0](https://spdx.org/licenses/Apache-2.0.html)
+
+## Network Information
+| Network Information |  Value         |
+|---------------------|------------------|
+|  Framework          | TensorFlow Lite |
+|  SHA-1 Hash         | 5bd511fc17ec7bfe9cd0f51bdec1537b874f52d2 |
+|  Size (Bytes)       | 27286108 |
+|  Provenance         | http://download.tensorflow.org/models/object_detection/ssd_mobilenet_v1_coco_2018_01_28.tar.gz |
+|  Paper              | https://arxiv.org/abs/1512.02325 |
+
+## Accuracy
+Dataset: Coco Validation 2017
+
+| Metric | Value |
+|--------|-------|
+| mAP | 0.21 |
+
+## Network Inputs
+| Input Node Name |  Shape  | Description |
+|-----------------|---------|-------------|
+| normalized_input_image_tensor | (1, 300, 300, 3) | A float input image. |
+
+## Network Outputs
+| Output Node Name |  Shape  | Description |
+|------------------|---------|-------------|
+| TFLite_Detection_PostProcess | () | An array of num_detection box boundaries for each input in the format (y1, x1, y2, x2) scaled from 0 to 1. |
+| TFLite_Detection_PostProcess:1 | () | COCO detection classes for each object. 1=person, 11=fire hydrant. |
+| TFLite_Detection_PostProcess:2 | () | Detection scores for each object. |
+| TFLite_Detection_PostProcess:3 | () | The number of objects detected in each image. |

models/object_detection/ssd_mobilenet_v1/tflite_fp32/recreate_model/README.md

@@ -0,0 +1,10 @@
+# SSD MobileNet v1 FP32 Model Re-Creation
+This folder contains a script that allows for the model to be re-created from scratch.
+
+## Requirements
+The script in this folder requires that the following must be installed:
+- Python 3.7
+- protoc
+
+## Running The Script
+To run the script, run the following in a terminal: `./recreate_model.sh`

models/object_detection/ssd_mobilenet_v1/tflite_fp32/recreate_model/recreate_model.sh

@@ -0,0 +1,38 @@
+# Copyright (C) 2020 Arm Limited or its affiliates. All rights reserved.
+#
+# SPDX-License-Identifier: Apache-2.0
+#
+# 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
+#
+# 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.
+
+#!/usr/bin/env bash
+
+python3.7 -m venv venv
+source venv/bin/activate
+
+pip install -r requirements.txt
+
+git clone https://github.com/tensorflow/models.git
+pushd models/research
+
+export PYTHONPATH=`pwd`:`pwd`/slim:$PYTHONPATH
+protoc object_detection/protos/*.proto --python_out=.
+
+wget http://download.tensorflow.org/models/object_detection/ssd_mobilenet_v1_coco_2018_01_28.tar.gz
+tar -xvf ssd_mobilenet_v1_coco_2018_01_28.tar.gz
+
+python object_detection/export_tflite_ssd_graph.py --pipeline_config_path=object_detection/samples/configs/ssd_mobilenet_v1_coco.config --trained_checkpoint_prefix=ssd_mobilenet_v1_coco_2018_01_28/model.ckpt --output_directory=. --add_postprocessing_op=true
+tflite_convert --graph_def_file=tflite_graph.pb --output_file=ssd_mobilenet_v1.tflite --input_shapes=1,300,300,3 --input_arrays=normalized_input_image_tensor --output_arrays=TFLite_Detection_PostProcess,TFLite_Detection_PostProcess:1,TFLite_Detection_PostProcess:2,TFLite_Detection_PostProcess:3 --change_concat_input_ranges=false --allow_custom_ops
+
+mv ssd_mobilenet_v1.tflite ../..
+
+popd

models/object_detection/ssd_mobilenet_v1/tflite_fp32/recreate_model/requirements.txt

@@ -0,0 +1,32 @@
+absl-py==0.11.0
+astor==0.8.1
+cached-property==1.5.2
+certifi==2020.6.20
+cycler==0.10.0
+gast==0.2.2
+google-pasta==0.2.0
+grpcio==1.33.2
+h5py==3.0.0
+importlib-metadata==2.0.0
+Keras-Applications==1.0.8
+Keras-Preprocessing==1.1.2
+kiwisolver==1.3.1
+Markdown==3.3.3
+matplotlib==3.3.2
+mock==4.0.2
+numpy==1.18.5
+opt-einsum==3.3.0
+Pillow==8.0.1
+protobuf==3.13.0
+pyparsing==2.4.7
+python-dateutil==2.8.1
+scipy==1.5.4
+six==1.15.0
+tensorboard==1.15.0
+tensorflow==1.15.4
+tensorflow-estimator==1.15.1
+termcolor==1.1.0
+tf-slim==1.1.0
+Werkzeug==1.0.1
+wrapt==1.12.1
+zipp==3.4.0

models/object_detection/ssd_mobilenet_v1/tflite_uint8/README.md

@@ -0,0 +1,41 @@
+# SSD MobileNet v1 UINT8
+
+## Description
+SSD MobileNet v1 is a object detection network, that localizes and identifies objects in an input image. This is a TF Lite quantized version that takes a 300x300 input image and outputs detections for this image. This model is trained and quantized by Google.
+
+## License
+[Apache-2.0](https://spdx.org/licenses/Apache-2.0.html)
+
+## Network Information
+| Network Information |  Value         |
+|---------------------|------------------|
+|  Framework          | TensorFlow Lite |
+|  SHA-1 Hash         | 1f9c945db9e32c33e5b91539f756a8fbef636405 |
+|  Size (Bytes)       | 6898880 |
+|  Provenance         | http://download.tensorflow.org/models/object_detection/ssd_mobilenet_v1_quantized_300x300_coco14_sync_2018_07_18.tar.gz |
+|  Paper              | https://arxiv.org/abs/1512.02325 |
+
+## Accuracy
+Dataset: Coco Validation 2017
+
+| Metric | Value |
+|--------|-------|
+| mAP | 0.18 |
+
+## Optimizations
+| Optimization |  Value  |
+|-----------------|---------|
+| Quantization | UINT8 |
+
+## Network Inputs
+| Input Node Name |  Shape  | Description |
+|-----------------|---------|-------------|
+| image_tensor | (1, 300, 300, 3) | Input RGB images (a range of 0-255 per RGB channel). |
+
+## Network Outputs
+| Output Node Name |  Shape  | Description |
+|------------------|---------|-------------|
+| TFLite_Detection_PostProcess | () | The y1, x1, y2, x2 coordinates of the bounding boxes for each detection |
+| TFLite_Detection_PostProcess:1 | () | The class of each detection |
+| TFLite_Detection_PostProcess:2 | () | The probability score for each classification |
+| TFLite_Detection_PostProcess:3 | () | A vector containing a number corresponding to the number of detections |

models/object_detection/ssd_mobilenet_v1/tflite_uint8/recreate_model/README.md

@@ -0,0 +1,9 @@
+# SSD MobileNet v1 UINT8 Model Re-Creation
+This folder contains a script that allows for the model to be re-created from scratch.
+
+## Requirements
+The script in this folder requires that the following must be installed:
+- Python 3.7
+
+## Running The Script
+To run the script, run the following in a terminal: `./recreate_model.sh`

models/object_detection/ssd_mobilenet_v1/tflite_uint8/recreate_model/recreate_model.sh

@@ -0,0 +1,32 @@
+# Copyright (C) 2020 Arm Limited or its affiliates. All rights reserved.
+#
+# SPDX-License-Identifier: Apache-2.0
+#
+# 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
+#
+# 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.
+
+#!/usr/bin/env bash
+
+python3.7 -m venv venv
+source venv/bin/activate
+
+pip install -r requirements.txt
+
+wget http://download.tensorflow.org/models/object_detection/ssd_mobilenet_v1_quantized_300x300_coco14_sync_2018_07_18.tar.gz
+tar -xvf ssd_mobilenet_v1_quantized_300x300_coco14_sync_2018_07_18.tar.gz
+
+pushd ssd_mobilenet_v1_quantized_300x300_coco14_sync_2018_07_18
+
+tflite_convert --graph_def_file=tflite_graph.pb --output_file=ssd_mobilenet_v1.tflite --input_shapes=1,300,300,3 --input_arrays=normalized_input_image_tensor --output_arrays=TFLite_Detection_PostProcess,TFLite_Detection_PostProcess:1,TFLite_Detection_PostProcess:2,TFLite_Detection_PostProcess:3 --change_concat_input_ranges=false --allow_custom_ops --inference_type=QUANTIZED_UINT8 --mean_values=128 --std_dev_values=128
+mv ssd_mobilenet_v1.tflite ..
+
+popd

models/object_detection/ssd_mobilenet_v1/tflite_uint8/recreate_model/requirements.txt

@@ -0,0 +1,22 @@
+absl-py==0.11.0
+astor==0.8.1
+cached-property==1.5.2
+gast==0.2.2
+google-pasta==0.2.0
+grpcio==1.33.2
+h5py==3.0.0
+importlib-metadata==2.0.0
+Keras-Applications==1.0.8
+Keras-Preprocessing==1.1.2
+Markdown==3.3.3
+numpy==1.18.5
+opt-einsum==3.3.0
+protobuf==3.13.0
+six==1.15.0
+tensorboard==1.15.0
+tensorflow==1.15.4
+tensorflow-estimator==1.15.1
+termcolor==1.1.0
+Werkzeug==1.0.1
+wrapt==1.12.1
+zipp==3.4.0