Annotating media content for automatic content understanding
A system for annotating frames in a media stream 114 includes a pattern recognition system (PRS) 108 to generate PRS output metadata for a frame; an archive 106 for storing ground truth metadata (GTM); a device to merge the GTM and PRS output metadata and thereby generate proposed annotation data (PAD) 110 ; and a user interface 109 for use by the human annotator HA 118 . The user interface 104 includes an editor 111 and an input device 107 used by the HA 118 to approve GTM for the frame. An optimization system 105 receives the approved GTM and metadata output by the PRS 108 , and adjusts input parameters for the PRS to minimize a distance metric corresponding to a difference between the GTM and PRS output metadata.
1. A non-transitory, machine-readable storage medium, comprising executable instructions that, when executed by a processing system including a processor, facilitate performance of operations, comprising:
performing pattern recognition on video frames of a media stream to generate pattern recognition metadata associated with the video frames of the media stream;
comparing the pattern recognition metadata associated with the video frames of the media stream and ground-truth metadata associated with the video frames of the media stream to generate a single distance metric, wherein the ground-truth metadata and the pattern recognition metadata comprise a first type of metadata; and
adjusting a set of input parameters associated with the pattern recognition according to the single distance metric, wherein the single distance metric is weighted according to the first type of metadata.
2. The non-transitory, machine-readable storage medium of claim 1 , wherein the operations further comprise generating proposed annotation data associated with the video frames of the media stream.
3. The non-transitory, machine-readable storage medium of claim 2 , wherein the generating the proposed annotations data comprises merging the pattern recognition metadata associated with the video frames of the media stream with the ground-truth metadata associated with the video frames of the media stream.
4. The non-transitory, machine-readable storage medium of claim 2 , wherein the operations further comprise receiving approval of the proposed annotation data via a user interface to generate the ground-truth metadata associated with the video frames of the media stream.
5. The non-transitory, machine-readable storage medium of claim 2 , wherein the operations further comprise presenting the proposed annotation data for the video frames via a user interface, wherein the user interface includes an editor and an input device.
6. The non-transitory, machine-readable storage medium of claim 5 , wherein the presenting of the proposed annotation data comprises overlaying the proposed annotation data onto the media stream via the user interface.
7. The non-transitory, machine-readable storage medium of claim 1 , wherein the adjusting the set of input parameters associated with the pattern recognition minimizes the single distance metric.
8. The non-transitory, machine-readable storage medium of claim 1 , wherein the first type of metadata is associated with spatial position.
9. The non-transitory, machine-readable storage medium of claim 1 , wherein the operations further comprise storing the ground-truth metadata associated with the video frames of the media stream in an archive.
10. The non-transitory, machine-readable storage medium of claim 1 , wherein the operations further comprise correcting a time delay between third party metadata and the media stream via an alignment process.
11. A method, comprising:
comparing, by a processing system including a processor, pattern recognition metadata associated with video frames of a media stream and ground-truth metadata associated with the video frames of the media stream to generate a single distance metric;
adjusting, by the processing system, a set of input parameters associated with pattern recognition according to the single distance metric, wherein the single distance metric is weighted according to a first type of metadata; and
merging, by the processing system, the pattern recognition metadata associated with the video frames of the media stream with the ground-truth metadata associated with the video frames of the media stream to generate proposed annotation data associated with the video frames of the media stream.
12. The method of claim 11 , further comprising performing, by the processing system, pattern recognition on the video frames of the media stream to generate the pattern recognition metadata associated with the video frames of the media stream.
13. The method of claim 11 , wherein the adjusting the set of input parameters associated with the pattern recognition minimizes the single distance metric.
14. The method of claim 11 , further comprises storing, by the processing system, the ground-truth metadata associated with the video frames of the media stream in an archive, wherein the ground-truth metadata and the pattern recognition metadata comprise the first type of metadata.
15. The method of claim 11 , further comprising comprise presenting, by the processing system, the proposed annotation data for the video frames via a user interface, wherein the user interface includes an editor and an input device, wherein the presenting of the proposed annotation data comprises overlaying the proposed annotation data onto the media stream via the user interface.
16. The method of claim 11 , further comprising receiving, by the processing system, approval of the proposed annotation data via a user interface to generate the ground-truth metadata associated with the video frames of the media stream.
17. The method of claim 11 , further comprising correcting, by the processing system, a time delay between third party metadata and the media stream via an alignment process.
18. A system, comprising:
a pattern recognition system, including a processing system comprising a processor, that facilitates generating pattern recognition metadata associated with video frames of a media stream according to a set of input parameters associate with the pattern recognition system; and
an optimization system, including the processing system, that facilitates comparing the pattern recognition metadata associated with the video frames of the media stream and ground-truth metadata associated with the video frames of the media stream to generate a single distance metric, and wherein the set of input parameters associated with the pattern recognition system are adjusted according to the single distance metric.
19. The system of claim 18 , further comprising an encoder system, including the processing system, that facilitates generating proposed annotation data associated with the video frames of the media stream by merging the pattern recognition metadata with the ground-truth metadata associated with the video frames of the media stream.
20. The system of claim 19 , further comprising receiving approval of the proposed annotation data via a user interface to generate the ground-truth metadata associated with the video frames of the media stream.