ONNX support for Object Detection

examples/onnx/image_classification/README.md

@@ -0,0 +1,25 @@
+# Instructions
+This examples uses the model fron the CNTK model from the [Deep Learning - CNTK for Land Cover Classification](https://www.arcgis.com/home/item.html?id=e8bc272d1ce2456fa4b87c9af749a57f) example.
+
+The CNTK model file was converted to ONNX using the [CNTK to ONNX Export](https://github.com/onnx/tutorials/blob/master/tutorials/CntkOnnxExport.ipynb) instructions. 
+
+``` 
+import cntk as C
+model_path = "trained.model"
+z = C.Function.load(model_path, device=C.device.cpu())
+z.save("trained.onnx", format=C.ModelFormat.ONNX)
+```
+
+## Running the Model
+
+Step 0. Download the test deep learning model and image [here](https://www.arcgis.com/home/item.html?id=e8bc272d1ce2456fa4b87c9af749a57f). You can also use your own trained model and test image.
+
+Step 1. Open "Classify Pixels Using Deep Learning" geoprocessing tool.
+
+Step 2. Fill in the parameters.
+
+![](https://github.com/gbrunner/raster-deep-learning/blob/master/docs/img/onnx_landclassificationexampletool.png)
+
+Step 3. Run the tool.
+
+![](https://github.com/gbrunner/raster-deep-learning/blob/master/docs/img/onnx_landclassificationexample.png)

examples/onnx/image_classification/trained.emd

@@ -0,0 +1,44 @@
+{
+    "Framework":"Onnx",
+    "ModelConfiguration":"CntkLandClassification",
+    "ModelType":"ImageClassification",
+    "ModelFile":"trained.onnx",
+    "ExtractBands":[0,1,2,3],
+    "DataRange":[0.0,1.0],
+    "ImageHeight":256,
+    "ImageWidth":256,
+    "ModelPadding":64,
+
+    "Classes": [
+        {
+            "Value":0,
+            "Name":"Evergreen Forest",
+            "Color":[0, 51, 0]
+         },
+         {
+            "Value":1,
+            "Name":"Grassland/Herbaceous",
+            "Color":[241, 185, 137]
+         },
+         {
+            "Value":2,
+            "Name":"Bare Land",
+            "Color":[236, 236, 0]
+         },
+         {
+            "Value":3,
+            "Name":"Open Water",
+            "Color":[0, 0, 117]
+         },
+         {
+            "Value":4,
+            "Name":"Scrub/Shrub",
+            "Color":[102, 102, 0]
+         },
+         {
+            "Value":5,
+            "Name":"Impervious Surface",
+            "Color":[236, 236, 236]
+         }
+    ]
+}

examples/onnx/object_detection/README.md

@@ -0,0 +1,41 @@
+# Instructions
+The ONNX model was converted from the [XView](http://xviewdataset.org/) TensorFlow model, multires.pb. For more tools to work with the XView dataset, please see the [XView Github page](https://github.com/DIUx-xView).
+
+# XView Model Conversion Using ```tf2onnx```
+On Windows, the ONNX model was generated with
+
+```
+tensorflow=1.13.1
+onnx=1.6.0
+tf2onnx=1.5.3
+```
+
+and
+
+```
+tensorflow=1.14.0
+onnx=1.6.0
+tf2onnx=1.5.3
+```
+
+Some errors are seen when converting to ONNX that will effect it's accuracy.
+
+Conversion to ONNX can be done through the command line using the following statement:
+
+```
+python -m tf2onnx.convert --input C:\XVIEW\multires.pb --inputs "image_tensor:0" --outputs "detection_boxes:0,detection_scores:0,detection_classes:0" --output C:\XVIEW\ONNX\saved_model.onnx --opset 10
+```
+
+# Running the Model
+
+Step 0. Download the test deep learning model and the XView training or validation images.
+
+Step 1. Open "Detect Object Using Deep Learning" geoprocessing tool.
+
+Step 2. Fill in the parameters.
+
+![](https://github.com/gbrunner/raster-deep-learning/blob/master/docs/img/onnx_objectdetectiontool.png)
+
+Step 3. Run the Model.
+
+![](https://github.com/gbrunner/raster-deep-learning/blob/master/docs/img/onnx_objectdetection.png)