Try using the Emotion API from Android

This article was written on the blog before, but since it is a blog that no one comes to, I thought that Qiita would be better and posted it again.

I participated in the Ishinomaki Hackathon held from July 28th to July 30th.

Therefore, I sometimes used the Emotion API of Microsoft Cognitive Services from Android, so I will write it as a memo at that time.

I referred to the following site for implementation. http://qiita.com/kosfuji/items/575408ae17113d7b58e9 http://qiita.com/a_nishimura/items/19cf3f60ad1dd3f66a84

The official implementation sample had java code, but since Android 6.0 the Apache HTTP Client is no longer supported, so I had to use HttpUrlConnection. # Preparation

Here because you need a Subscription Key to use the API Log in with your Microsoft account from> to get your Subscription Key.

Since HTTP communication is performed, add INTERNET permission.

<uses-permission android:name="android.permission.INTERNET" />

First, create a class that inherits AsyncTask to set up a thread for communication.

public class ConnectToEmotionAPI extends AsyncTask<Void, Void, JSONObject> {
    @Override
    protected void onPreExecute() {
    }
    @Override
    protected JSONObject doInBackground(Void... params) {
    }
    @Override
    protected void onPostExecute(JSONObject result) {
    }
}

Since the result of analyzing the image from the Emotion API is returned in JSON, the return value of doInBackground should be JSONObject.

Next, we will write the communication process in doInBackground.

HttpURLConnection con = null;
URL url = null;
String urlStr = "https://westus.api.cognitive.microsoft.com/emotion/v1.0/recognize";
String key = "{Your Key}";                  //Subscription Key
DataOutputStream os = null;
BufferedReader reader;
JSONObject json = null;

try {
            url = new URL(urlStr);
            con = (HttpURLConnection)url.openConnection();
            con.setReadTimeout(10000);
            con.setConnectTimeout(20000);
            con.setRequestMethod("POST");
            con.setDoInput(true);
            con.setDoOutput(true);
            //Request header settings
            con.setRequestProperty("Content-Type", "application/octet-stream");
            con.setRequestProperty("Ocp-Apim-Subscription-Key", key);

            //Creating a request body
            Resources r = main_.getResources();
            Bitmap bmp = BitmapFactory.decodeResource(r, R.drawable.face_small);

            //Convert images to binary data
            byte[] byteArray;
            ByteArrayOutputStream bos = new ByteArrayOutputStream();
            bmp.compress(Bitmap.CompressFormat.JPEG, 100, bos);
            byteArray = bos.toByteArray();

            os = new DataOutputStream(con.getOutputStream());
            for(int i =  0 ; i < byteArray.length;i++){
                   os.writeByte(byteArray[i]);
            }

            //Connect to API
            con.connect();
            os.close();
            
            int status = con.getResponseCode();

            switch (status) {
                case HttpURLConnection.HTTP_OK:
                    InputStream in = con.getInputStream();
                    reader = new BufferedReader(new InputStreamReader(in));
                    String line;
                    String readStr = new String();

                    while (null != (line = reader.readLine())){
                        readStr += line;
                    }
                    Log.d("EmotionAPI","read string: " + readStr);

                    in.close();

                    json = new JSONArray(readStr).getJSONObject(0);
                    break;

                case HttpURLConnection.HTTP_UNAUTHORIZED:
                    break;
                default:
                    break;
            }

} catch (MalformedURLException e){
            e.printStackTrace();
} catch (JSONException e){
            e.printStackTrace();
} catch (IOException e){
            e.printStackTrace();
}

return json;

When sending an image, we use ByteArrayOutputStream to send it as binary data.

Since the returned JSON is an array format, use JSONArray to parse the JSON.

Finally, the JSON obtained by onPostExecute is parsed for each object name.

    @Override
    protected void onPostExecute(JSONObject result) {
        super.onPostExecute(result);

        JSONObject jsonData;
        String[] str = new String[2];
        try {
            jsonData = result.getJSONObject("scores");
            str[0] = jsonData.getString("happiness");
            str[1] = jsonData.getString("anger");
        } catch (Exception e){
            e.printStackTrace();
        }

        if (isSmile(str[0])) {
            Log.d("EmotionAPI","It ’s a nice smile!");
        } else if (isAnger(str[1])) {
            Log.d("EmotionAPI","Don't get so angry~");
        } else {
            Log.d("EmotionAPI","It's boring. Please react something");
        }
    }

    public boolean isSmile(String strValue){

        double value = Double.parseDouble(strValue);
        if (value > 0.5) return true;
        else return false;
    }

    public boolean isAnger(String strValue){

        double value = Double.parseDouble(strValue);
        if (value > 0.5) return true;
        else return false;
    }

The code implemented in the hackathon is published on GitHub, so please see that for the entire implementation. Github EmotionAPI Sample