SSD: Single Shot MultiBox Detector for prediction with DD

[beniz]

This PR adds support for the image detection task (e.g. object detection) via SSD. Only prediction mode is covered by the present PR. Training should follow in a near futur.

Functionalities:

• multi-class object detection
• single and batch forward processing of images
• JSON output with bouding boxes resized to fit the originally submitted image
• threshold on prediction confidence via confidence_threshold output parameter
• ready-to use demo in demo/objdetect

Sample usage:

• create service

This is identical to other tasks

curl -X PUT "http://localhost:8080/services/imageserv" -d "{\"mllib\":\"caffe\",\"description\":\"image classification service\",\"type\":\"supervised\",\"parameters\":{\"input\":{\"connector\":\"image\",\"height\":300,\"width\":300},\"mllib\":{\"nclasses\":2}},\"model\":{\"repository\":\"/path/to/model\"}}"

• prediction

curl -X POST "http://localhost:8080/predict" -d '{"service":"imageserv","parameters":{"output":{"bbox":true,"confidence_threshold":0.4}},"data":["/path/to/image"]}'