Techniques for face detection and tracking
Techniques are disclosed that involve face detection. For instance, face detection tasks may be decomposed into sets of one or more sub-tasks. In turn the sub-tasks of the sets may be allocated across multiple image frames. This allocation may be based on a resource budget. In addition, face tracking tasks may be performed.
1. A method, comprising:
storing a plurality of image frames in a memory; and
performing a face detection task on the plurality of image frames, the face detection task comprising multiple successive sets of tasks;
wherein the successive sets employ progressively increasing detection region size;
detecting whether sufficient computing resources exist for a sub-task;
merging and filtering faces in the data structure in response to the determination that sufficient resources do not exist for the sub-task;
wherein each of the multiple sets employs one or more detection regions having the corresponding detection region size; and
wherein at least one of the multiple sets employs two or more partially overlapping detection regions having the corresponding detection region size.
2. The method of claim 1 , further comprising allocating a resource budget for face detection sub-task performance in each of the plurality of image frames.
3. The method of claim 1 , wherein said performing comprises, for each sub-task, scanning a window within a detection region, the detection region having the detection region size of the corresponding set.
4. The method of claim 3 , wherein said performing comprises, for each sub-task, employing a plurality of classification stages, each classification stage operating on multiple windows concurrently.
5. The method of claim 1 , wherein said performing comprises performing at least one of the sub-tasks in a corresponding one of the plurality of frames.
6. The method of claim 1 , further comprising:
performing a face tracking task on the plurality of image frames.
7. The method of claim 6 , further comprising maintaining information regarding one or more previously detected faces.
8. An apparatus, comprising:
a face tracking module to track one or more previously detected faces within an image frame; and
a face detection module to perform one or more face detection tasks within the image frame;
wherein each face detection task scans a window within a detection region; and
wherein the successive sets employ progressively increasing detection region size, detect whether sufficient computing resources exist for a sub-task, and merge and filter faces in the data structure in response to the determination that sufficient resources do not exist for the sub-task; wherein each of the multiple sets employs one or more detection regions having the corresponding detection region size; and wherein at least one of the multiple sets employs two or more partially overlapping detection regions having the corresponding detection region size.
9. The apparatus of claim 8 , further comprising a storage medium to store information regarding the one or more previously detected faces.
10. The apparatus of claim 9 , wherein the storage medium further stores information regarding any faces detected by one or more face detection tasks.
11. The apparatus of claim 8 , further comprising an image sensor to generate the image frame.
12. The apparatus of claim 8 , further comprising a display to output the image frame.
13. The apparatus of claim 8 , wherein the detection module includes a plurality of classification stages, each classification stage including multiple classification module to operate on multiple windows concurrently.
14. The apparatus of claim 8 , wherein each of the face detection tasks are included in one of a plurality of task sets; and
wherein each of the task sets employs a corresponding window size.
15. A non-transitory computer readable medium having stored thereon instructions that, when executed by a machine, cause the machine to:
perform a face detection task on the plurality of image frames, the face detection task comprising multiple successive sets of tasks;
wherein the successive sets employ progressively increasing detection region size;
detect whether sufficient computing resources exist for a sub-task;
merge and filter faces in the data structure in response to the determination that sufficient resources do not exist for the sub-task;
wherein each of the multiple sets employs one or more detection regions having the corresponding detection region size; and
wherein at least one of the multiple sets employs two or more partially overlapping detection regions having the corresponding detection region size.
16. The article of claim 15 , wherein the instructions, when executed by a machine, cause the machine to:
allocate a resource budget for face detection task performance in each of the plurality of image frames.
17. The article of claim 15 , wherein said performing comprises, for each task, scanning a window within a detection region, the detection region having the detection region size of the corresponding set.
18. The method of claim 1 including using overlapping detection regions.