Http Server

Dockerfile.server

@@ -0,0 +1,25 @@
+FROM tensorflow/tensorflow:latest-py3
+
+# Install system packages
+RUN apt-get update && apt-get install -y --no-install-recommends \
+      bzip2 \
+      g++ \
+      git \
+      graphviz \
+      libgl1-mesa-glx \
+      libhdf5-dev \
+      openmpi-bin \
+      wget && \
+    rm -rf /var/lib/apt/lists/*
+
+COPY src /src
+COPY entrypoints /src/entrypoints
+COPY models /models
+
+WORKDIR /src
+
+RUN pip install -r requirements.txt
+
+ENV PYTHONPATH='/src/:$PYTHONPATH'
+
+ENTRYPOINT ["entrypoints/entrypoint.predict.server.sh"]

entrypoints/entrypoint.predict.server.sh

@@ -0,0 +1,10 @@
+#!/bin/bash
+set -e
+
+BASE_MODEL_NAME=$1
+WEIGHTS_FILE=$2
+
+# predict
+python -m evaluater.server \
+--base-model-name $BASE_MODEL_NAME \
+--weights-file $WEIGHTS_FILE

src/evaluater/predict.py

@@ -1,4 +1,3 @@
-
 import os
 import glob
 import json

src/evaluater/server.py

@@ -0,0 +1,86 @@
+#!/usr/bin/python
+# -*- coding: utf-8 -*-
+
+import os
+from flask import Flask, request, jsonify
+from evaluater.predict import image_file_to_json, image_dir_to_json, predict
+import urllib
+import shutil
+import argparse
+from keras import backend as K
+from PIL import ImageFile, Image
+from handlers.model_builder import Nima
+from handlers.data_generator import TestDataGenerator
+
+app = Flask('server')
+
+def load_model(config):
+    global model
+    model = Nima(config.base_model_name)
+    model.build()
+    model.nima_model.load_weights(config.weights_file)
+    model.nima_model._make_predict_function()  # https://github.com/keras-team/keras/issues/6462
+    model.nima_model.summary()
+
+def main(image_source, predictions_file, img_format='jpg'):
+    # load samples
+    if os.path.isfile(image_source):
+        image_dir, samples = image_file_to_json(image_source)
+    else:
+        image_dir = image_source
+        samples = image_dir_to_json(image_dir, img_type='jpg')
+
+    ImageFile.LOAD_TRUNCATED_IMAGES = True
+    Image.MAX_IMAGE_PIXELS = None
+
+    # initialize data generator
+    data_generator = TestDataGenerator(samples, image_dir, 64, 10, model.preprocessing_function(),
+                                       img_format=img_format)
+
+    # get predictions
+    predictions = predict(model.nima_model, data_generator)
+    K.clear_session()
+
+    # calc mean scores and add to samples
+    for i, sample in enumerate(samples):
+        sample['mean_score_prediction'] = calc_mean_score(predictions[i])
+
+    if predictions_file is not None:
+        save_json(samples, predictions_file)
+
+    return samples
+
+@app.route('/prediction', methods=['POST'])
+def prediction():
+
+    global images
+
+    if request.method == 'POST':
+        images = request.json
+
+        if images:
+
+            shutil.rmtree('temp')
+            os.mkdir('temp')
+            for image in images:
+                filename_w_ext = os.path.basename(image)
+                try:
+                    urllib.request.urlretrieve(image, 'temp/'+ filename_w_ext)
+                except:
+                    print('An exception occurred :' + image)
+
+            result = main('temp', None)
+
+            return jsonify(result)
+
+        return jsonify({'error': 'Image is not available'})
+
+if __name__ == '__main__':
+
+    parser = argparse.ArgumentParser()
+    parser.add_argument('-b', '--base-model-name', help='CNN base model name', required=True)
+    parser.add_argument('-w', '--weights-file', help='path of weights file', required=True)
+    args = parser.parse_args()
+
+    load_model(args)
+    app.run(host='0.0.0.0', port=5005)

src/requirements.txt

@@ -1,3 +1,4 @@
 keras==2.1.*
 nose==1.3.*
 pillow==5.0.*
+Flask==1.0.*