Imagenet example

examples/imagenet/cortex.yaml

@@ -0,0 +1,7 @@
+- kind: deployment
+  name: imagenet
+
+- kind: api
+  name: classifier
+  model: s3://cortex-examples/imagenet/1566492692
+  request_handler: imagenet.py

examples/imagenet/imagenet.py

@@ -0,0 +1,26 @@
+import requests
+import numpy as np
+import base64
+from PIL import Image
+from io import BytesIO
+import math
+
+
+labels = requests.get(
+    "https://storage.googleapis.com/download.tensorflow.org/data/ImageNetLabels.txt"
+).text.split("\n")
+
+
+def pre_inference(sample, metadata):
+    if "url" in sample:
+        image = requests.get(sample["url"]).content
+    elif "base64" in sample:
+        image = base64.b64decode(sample["base64"])
+
+    decoded_image = np.asarray(Image.open(BytesIO(image)), dtype=np.float32) / 255
+    return {"images": [decoded_image.tolist()]}
+
+
+def post_inference(prediction, metadata):
+    classes = prediction["response"]["classes"]
+    return {"class": labels[np.argmax(classes)]}

examples/imagenet/load_inception_from_tf_hub.py

@@ -0,0 +1,22 @@
+import tensorflow as tf
+import tensorflow_hub as hub
+from tensorflow.python.saved_model.signature_def_utils_impl import predict_signature_def
+
+export_dir = "/model/15692343342"
+builder = tf.saved_model.builder.SavedModelBuilder(export_dir)
+with tf.Session(graph=tf.Graph()) as sess:
+    module = hub.Module("https://tfhub.dev/google/imagenet/inception_v3/classification/3")
+    input_params = module.get_input_info_dict()
+    image_input = tf.placeholder(
+        name="images", dtype=input_params["images"].dtype, shape=input_params["images"].get_shape()
+    )
+    sess.run([tf.global_variables_initializer(), tf.tables_initializer()])
+    classes = module(image_input)
+
+    signature = predict_signature_def(inputs={"images": image_input}, outputs={"classes": classes})
+
+    builder.add_meta_graph_and_variables(
+        sess, ["serve"], signature_def_map={"predict": signature}, strip_default_attrs=True
+    )
+
+builder.save()