Merge pull request webmachinelearning#55 from Honry/add-ssd

Add SSD MobileNet V1 coco examples

object_detection/README.md

@@ -2,7 +2,41 @@
 
 Object detection is the process of detecting instances of semantic objects of a pre-defined classes in digital images and videos.
 
-This example shows how the Tiny YOLO v2 model from [YAD2K project](https://github.com/allanzelener/YAD2K) and [ONNX model zoo](https://github.com/onnx/models/blob/master/vision/object_detection_segmentation/tiny-yolov2/model/tinyyolov2-8.tar.gz) may be implemented by using the WebNN API. (Note: these two models are all trained on the [Pascal VOC](http://host.robots.ox.ac.uk/pascal/VOC/) dataset.)
+This example shows how the Tiny YOLO v2 model from [YAD2K project](https://github.com/allanzelener/YAD2K) and [ONNX model zoo](https://github.com/onnx/models/blob/master/vision/object_detection_segmentation/tiny-yolov2/model/tinyyolov2-8.tar.gz), [SSD MobileNet V1 models](http://download.tensorflow.org/models/object_detection/ssd_mobilenet_v1_coco_2018_01_28.tar.gz) may be implemented by using the WebNN API. (Note: Tiny YOLO V2 models are trained on the [Pascal VOC](http://host.robots.ox.ac.uk/pascal/VOC/) dataset, SSD MobileNet V1 models are trained on the [COCO](https://cocodataset.org/#home) dataset.)
+
+
+### How to Generate SSD MobileNet V1 models
+
+Since the original [SSD MobileNet V1 model](http://download.tensorflow.org/models/object_detection/ssd_mobilenet_v1_coco_2018_01_28.tar.gz) contains customized preprocess and postprocess graphs, we only implement a cut model with WebNN API, you can generate the cut model via following commands:
+
+Here is the converter command for removing the preprocess and postprocess graphs by using tensorflow's [`optimize_for_inference`](https://github.com/tensorflow/tensorflow/blob/master/tensorflow/python/tools/optimize_for_inference.py) tool:
+
+```
+python3 -m tensorflow.python.tools.optimize_for_inference \
+--input=./frozen_inference_graph.pb \
+--output=./frozen_inference_graph_stripped.pb --frozen_graph=True \
+--input_names=Preprocessor/sub \
+--output_names='concat,concat_1' \
+--alsologtostderr
+```
+
+Use ['TensorFlow Lite converter'](https://www.tensorflow.org/lite/convert) tool to convert frozen graph to tflite model:
+
+```
+tflite_convert \
+--graph_def_file=./frozen_inference_graph_stripped.pb \
+--output_file=./ssd_mobilenet_v1_coco.tflite \
+--input_format=TENSORFLOW_GRAPHDEF --output_format=TFLITE \
+--input_shapes=1,300,300,3 --input_arrays=Preprocessor/sub \
+--enable_v1_converter \
+--output_arrays='concat,concat_1' \
+--inference_type=FLOAT --logtostderr
+```
+
+Use ['tflite2onnx'](https://github.com/jackwish/tflite2onnx) tool to convert tflite model to onnx model:
+```
+tflite2onnx ssd_mobilenet_v1_coco.tflite ssd_mobilenet_v1_coco.onnx
+```
 
 ### Usage
 

object_detection/index.html

@@ -32,6 +32,9 @@
           <label class="btn btn-outline-info btn-sm mr-2 active">
             <input type="radio" name="model" id="tinyyolov2" autocomplete="off" checked>Tiny Yolo V2
           </label>
+          <label class="btn btn-outline-info btn-sm mr-2">
+            <input type="radio" name="model" id="ssdmobilenetv1" autocomplete="off">SSD MobileNet V1
+          </label>
         </div>
       </div>
     </div>

object_detection/labels/README.md

@@ -1 +1 @@
-The label file `pascal.classes.txt` comes from [YAD2K](https://github.com/allanzelener/YAD2K), which is licensed under MIT.
+The label files `pascal_classes.txt` and `coco_classes.txt` come from [YAD2K](https://github.com/allanzelener/YAD2K), which is licensed under MIT.

object_detection/labels/coco_classes.txt

@@ -0,0 +1,91 @@
+???
+person
+bicycle
+car
+motorcycle
+airplane
+bus
+train
+truck
+boat
+traffic light
+fire hydrant
+???
+stop sign
+parking meter
+bench
+bird
+cat
+dog
+horse
+sheep
+cow
+elephant
+bear
+zebra
+giraffe
+???
+backpack
+umbrella
+???
+???
+handbag
+tie
+suitcase
+frisbee
+skis
+snowboard
+sports ball
+kite
+baseball bat
+baseball glove
+skateboard
+surfboard
+tennis racket
+bottle
+???
+wine glass
+cup
+fork
+knife
+spoon
+bowl
+banana
+apple
+sandwich
+orange
+broccoli
+carrot
+hot dog
+pizza
+donut
+cake
+chair
+couch
+potted plant
+bed
+???
+dining table
+???
+???
+toilet
+???
+tv
+laptop
+mouse
+remote
+keyboard
+cell phone
+microwave
+oven
+toaster
+sink
+refrigerator
+???
+book
+clock
+vase
+scissors
+teddy bear
+hair drier
+toothbrush