Apparatus and methods for detecting adult videos

• US 8,358,837 B2
• Filed: 05/01/2008
• Issued: 01/22/2013
• Est. Priority Date: 05/01/2008
• Status: Active Grant

• Current Assignee: Verizon Patent and Licensing Incorporated (Verizon Communications Inc.)
• Original Assignee: Yahoo! Inc. (Apollo Global Management, Inc.)
• Inventors: Shakya, Subodh, Zhang, Ruofei
• Primary Examiner(s): Alavi, Amir
• Application Number:
  US12/113,835
• Publication Number:
  US 20090274364A1
• Time in Patent Office: 1,727 Days
• Field of Search: 382/155, 382/165, 382/170, 382/190, 382260-264, 707/4, 707/5, 707/9, 707/10, 709/206, 709/217, 709/218, 709/219, 709/225, 715/208, 715/271, 726/1, 726/22
• US Class Current: 382/165
• CPC Class Codes: G06V 20/40 in video content extracting overlay text G06V20/62; video retrieval G06F16/70; processing of video elementary streams in video servers H04N21/234; processing of video elementary streams in video clients H04N21/44
• CPC Class Codes: G06V 40/10 Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands

Abstract

Disclosed are apparatus and methods for detecting whether a video is adult or non-adult. In certain embodiments, a learning system is operable to generate one or more models for adult video detection. The model is generated based on a large set of known videos that have been defined as adult or non-adult. Adult detection is then based on this adult detection model. This adult detection model may be applied to selected key frames of an unknown video. In certain implementations, these key frames can be selected from the frames of the unknown video. Each key frame may generally correspond to a frame that contains key portions that are likely relevant for detecting pornographic or adult aspects of the unknown video. By way of examples, key frames may include moving objects, skin, people, etc. In alternative embodiments, a video is not divided into key frames and all frames are analyzed by a learning system to generate a model, as well as by an adult detection system based on such model.

Claims

1. A method for detecting pornographic or adult videos, comprising:

for an unknown video having a plurality of frames, defining a plurality of key frames selected from the frames of the unknown video, wherein each key frame corresponds to a frame that contains features that are likely relevant for detecting pornographic or adult aspects of the unknown video; and

analyzing the key frames using an adult detection model that was generated by a learning process based on a training set of images and their associated adult indicators that each specifies whether the associated known image is an adult or non- adult image, whereby the analysis results in an adult indicator that specifies whether the unknown video is an adult video, a non-adult video, or a suspected adult video.

2. The method as recited in claim 1, wherein defining the key frames comprises:

determining one or more portions of each frame that are significantly different from corresponding portions of a plurality of adjacent frames; and

defining the key frames based on the significantly different one or more portions of each frame.

3. The method as recited in claim 2, wherein analyzing the key frames comprises:

analyzing one or more of the significantly different portions of each key frame with the adult detection model to thereby determine an adult indicator for such one or more of the significantly different portions of such each key frame being adult or non- adult; and

determining the adult indicator of the unknown video based on the adult indicators for the key frames.

4. The method as recited in claim 3, wherein an adult indicator is determined for each significantly different portion of each key frame that is determined to include a moving object.

5. The method as recited in claim 1, further comprising:

prior to analyzing the key frames of the unknown video, executing the learning process based on one or more key frame features extracted from each known image and the each known image'"'"'s associated adult indicator so as to generate the adult detection model that is to be used for the unknown video.

6. The method as recited in claim 5, further comprising extracting a plurality of key frame features from the key frames of the unknown video, wherein the analyzing of the key frames of the unknown video is based on the extracted key frame features for such unknown video, and wherein a same type of features are used for analysis of the key frames of the unknown video and by the learning process.

7. The method as recited in claim 5, further comprising:

after analyzing the key frames of the unknown video so that the unknown video is defined as a new known video, including the key frames and associated adult indicators in the training set of known images; and

executing the learning process based on each known image, including the key frames, and each known image'"'"'s adult indicator, including the key frames'"'"' adult indicators, so as to generate a new adult detection model to be used for adult detection of new unknown videos.

8. The method as recited in claim 7, further comprising manually correcting one or more adult indicators of the known images, which include the key frames of the new known video, prior to executing the learning process on such known images.

9. The method as recited in claim 1, wherein the adult indicator specifies that the unknown video is an adult video when an adult indicator field is set, and wherein the adult indicator specifies that the unknown video is a suspected adult video when a suspected adult indicator field is set.

10. The method as recited in claim 1, wherein the analysis results in a confidence level corresponding to the adult indicator, wherein the confidence level corresponding to the adult indicator comprises a value within a first range of values when the unknown video is an adult video, wherein the confidence level corresponding to the adult indicator is within a second range of values when the unknown video is a suspected adult video, and wherein the confidence level corresponding to the adult indicator is within a third range of values when the unknown values is a non-adult video.

11. An apparatus comprising at least a processor and a memory, wherein the processor and/or memory are configured to perform operations, comprising:

for an unknown video having a plurality of frames, defining a plurality of key frames selected from the frames of the unknown video, wherein each key frame corresponds to a frame that contains features that are likely relevant for detecting pornographic or adult aspects of the unknown video; and

analyzing the key frames using an adult detection model that was generated by a learning process based on a training set of images and their associated adult indicators that each specifies whether the associated known image is an adult or non-adult image, whereby the analysis results in an adult indicator that specifies whether the unknown video is an adult video, a non-adult video, or a suspected adult video.

12. The apparatus as recited in claim 11, wherein defining the key frames comprises:

determining one or more portions of each frame that are significantly different from corresponding portions of a plurality of adjacent frames; and

defining the key frames based on the significantly different one or more portions of each frame.

13. The apparatus as recited in claim 12, wherein analyzing the key frames comprises:

analyzing one or more of the significantly different portions of each key frame with the adult detection model to thereby determine an adult indicator for such one or more of the significantly different portions of such each key frame being adult or non- adult; and

determining the adult indicator of the unknown video based on the adult indicators for the key frames.

14. The apparatus as recited in claim 13, wherein an adult indicator is determined for each significantly different portion of each key frame that is determined to include a moving object.

15. The apparatus as recited in claim 11, wherein the processor and/or memory are

further configured for performing operations, further comprising:

prior to analyzing the key frames of the unknown video, executing the learning process based on one or more key frame features extracted from each known image and the each known image'"'"'s associated adult indicator so as to generate the adult detection model that is to be used for the unknown video.

16. The apparatus as recited in claim 15, wherein the processor and/or memory

are further configured for performing operations, further comprising:

extracting a plurality of key frame features from the key frames of the unknown video, wherein the analyzing of the key frames of the unknown video is based on the extracted key frame features for such unknown video, and wherein a same type of features are used for analysis of the key frames of the unknown video and by the learning process.

17. The apparatus as recited in claim 15, wherein the processor and/or memory are further configured for performing operations, further comprising:

after analyzing the key frames of the unknown video so that the unknown video is defined as a new known video, including the key frames and associated adult indicators in the training set of known images; and

executing the learning process based on each known image, including the key frames, and each known image'"'"'s adult indicator, including the key frames'"'"' adult indicators, so as to generate a new adult detection model to be used for adult detection of new unknown videos.

18. The apparatus as recited in claim 17, wherein the processor and/or memory are further configured for supporting manually correcting one or more adult indicators of the known images, which include the key frames of the new known video, prior to executing the learning process on such known images.

19. The apparatus as recited in claim 11, wherein the adult indicator specifies that the unknown video is an adult video when the adult indicator is in a first state, that the unknown video is a non-adult video when the adult indicator is in a second state, or that the unknown video is a suspected adult video when the adult indicator is in a third state.

20. At least one non-transitory computer readable storage medium having computer program instructions stored thereon that are arranged to perform operations, comprising:

for an unknown video having a plurality of frames, defining a plurality of key frames selected from the frames of the unknown video, wherein each key frame corresponds to a frame that contains features that are likely relevant for detecting pornographic or adult aspects of the unknown video; and

analyzing the key frames using an adult detection model that was generated by a learning process based on a training set of images and their associated adult indicators that each specifies whether the associated known image is an adult or non- adult image, whereby the analysis results in an adult indicator that specifies whether the unknown video is an adult video, a non-adult video, or a suspected adult video.

21. The at least one non-transitory computer readable storage medium as recited in claim 20, wherein defining the key frames comprises:

determining one or more portions of each frame that are significantly different from corresponding portions of a plurality of adjacent frames; and

defining the key frames based on the significantly different one or more portions of each frame.

22. The at least one non-transitory computer readable storage medium as recited in claim 21, wherein analyzing the key frames comprises:

analyzing one or more of the significantly different portions of each key frame with the adult detection model to thereby determine an adult indicator for such one or more of the significantly different portions of such each key frame being adult or non- adult; and

determining the adult indicator of the unknown video based on the adult indicators for the key frames.

23. The at least one non-transitory computer readable storage medium as recited in claim 22, wherein an adult indicator is determined for each significantly different portion of each key frame that is determined to include a moving object.

24. The at least one non-transitory computer readable storage medium as recited in claim 20, wherein the computer program instructions are further arranged to perform operations, further comprising:

prior to analyzing the key frames of the unknown video, executing the learning process based on one or more key frame features extracted from each known image and the each known image'"'"'s associated adult indicator so as to generate the adult detection model that is to be used for the unknown video.

25. The at least one computer readable storage medium as recited in claim 24, wherein the computer program instructions are further arranged to operations, further comprising:

extracting a plurality of key frame features from the key frames of the unknown video, wherein the analyzing of the key frames of the unknown video is based on the extracted key frame features for such unknown video, and wherein a same type of features are used for analysis of the key frames of the unknown video and by the learning process.

26. The at least one non-transitory computer readable storage medium as recited in claim 24, wherein the computer program instructions are further arranged to perform operations, further comprising:

after analyzing the key frames of the unknown video so that the unknown video is defined as a new known video, including the key frames and associated adult indicators in the training set of known images; and

executing the learning process based on each known image, including the key frames, and each known image'"'"'s adult indicator, including the key frames'"'"' adult indicators, so as to generate a new adult detection model to be used for adult detection of new unknown videos.

27. The at least one non-transitory computer readable storage medium as recited in claim 26, wherein the computer program instructions are further arranged to perform operations, further comprising:

manually correcting one or more adult indicators of the known images, which include the key frames of the new known video, prior to executing the learning process on such known images.

28. At least one non-transitory computer readable storage medium having computer program instructions stored thereon that are arranged to perform operations, comprising:

sending a request for a plurality of videos, wherein the request is associated with a parameter that indicates that pornographic or adult videos are to be filtered from such videos; and

receiving a plurality of references to a plurality of videos from which a plurality of adult videos have been filtered using an adult detection model that was generated by a learning process based on a training set of videos that each include an adult indicator that specifies whether the each known video is an adult video, a non-adult video, or a suspected adult video.

29. An apparatus comprising at least a processor and a memory, wherein the processor and/or memory are configured to perform operations, comprising:

sending a request for a plurality of videos, wherein the request is associated with a parameter that indicates that pornographic or adult videos are to be filtered from such videos; and

receiving a plurality of references to a plurality of videos from which a plurality of adult videos have been filtered using an adult detection model that was generated by a learning process based on a training set of videos that each include an adult indicator that specifies that the known video is an adult video when the adult indicator is in a first state, that the known video is a non-adult video when the adult indicator is in a second state, or that the known video is a suspected adult video when the adult indicator is in a third state.

30. At least one non-transitory computer readable storage medium having computer program instructions stored thereon that are arranged to perform operations, comprising:

sending a request for a plurality of videos, wherein the request is associated with a parameter that indicates that pornographic or adult videos are to be filtered from such videos; and

receiving a plurality of references to a plurality of videos from which a plurality of adult videos have been filtered using an adult detection model that was generated by a learning process based on a training set of videos that each include an adult indicator that specifies that the known video is an adult video when the adult indicator is in a first state, that the known video is a non-adult video when the adult indicator is in a second state, or that the known video is a suspected adult video when the adult indicator is in a third state.