pkmandke / ros-face-detection

Face Detection using dlib and ROS

Face detection in ROS

Face Detection using ROS pub/sub.

ROS Face detector

The package samples images from the webcam and displays a bounding box after detecting the largest face in the frame using dlib. The code for face detection is inspired by OpenFace.

TODO

• Add actionlib
• Add launch file (with if syntax)
• Reduce effect of face detect delay

Details

Update: 30/05/2019

Three nodes:

1. Image capture and publish. Source
2. Subscribe and plot image. Also, send the image as a (action) goal for face detection. The lag problem is solved by plotting irrespective of whether face co-ordinates are received so that only the face rectangle faces a lag. Source
3. The Face detection server. Processes requests and sends back face co-ordinates as results.Source

Other updates:

• Object oriented API for actionlib and pub/sub.
• Helper scripts.

Snapshots

Using actionlib

• Old Service based

There are 3 nodes. 2 of them are a publish-subscribe pair that transfer np.array images using floats. The subscriber node then calls a service(server) node which returns the face co-ordinates if found. These co-ordinates are used by the subscriber node to draw the bounding box.

Snapshot are in the snaps folder.