Techniques for obfuscating video generated by a camera device
This disclosure describes, in part, techniques for implementing customized motion zones for security monitoring using a privacy screen. In embodiments, such techniques may comprise receiving first data defining a first area associated with a motion zone, receiving image data generated by a camera, the image data encompassing at least a portion of the first area, determining a position of an object detected within the image data, and determining, based on the first data and the position of the object, that the object is inside of the first area. The techniques may further comprise defining a portion of image data that corresponds to the object detected within the image data and the first area, applying at least one obfuscation technique to the image data less that portion of the image data, and sending the image data having the applied obfuscation technique to at least one second electronic device.
1 . An electronic device comprising:
a camera;
a wireless transceiver;
one or more processors;
one or more computer readable media storing computer executable instructions which, when executed using the one or more processors, cause the electronic device to perform operations comprising
receiving first image data generated by the camera representing a first frame of a video,
determining, based on the first image data, first object data indicating a first set of pixel locations corresponding to a detected object,
accessing stored detection zone data indicating a detection zone, wherein the detection zone includes a first portion of the first frame, and a remaining portion of the first frame is outside the detection zone,
determining, based on the first object data and the stored detection zone data, that the detected object is located partially within the detection zone and partially within the remaining portion of the first frame,
generating, based on the first image data, the first object data, and the detection zone data, second image data, wherein in the second image data:
the entire detection zone is unblurred;
a second portion of the remaining portion of the first frame is unblurred,
the second portion corresponding to the detected object; and
a third portion of the remaining portion of the first frame is blurred; and
causing transmission of the second image data to a remote system using the wireless transceiver.
2 . The electronic device of claim 1 , wherein the electronic device comprises a passive infrared sensor.
3 . The electronic device of claim 1 , wherein the first object data is determined based on using one or more machine learning models, and wherein the first object data comprises bounding box data that indicates an x position of a first corner, a y position of a first corner, a width, and a height.
4 . A method comprising:
receiving first image data generated by a camera of an electronic device, the first image data representing a first frame of a video;
determining, based on the first image data, first object data indicating a first set of pixel locations corresponding to a detected object;
accessing stored detection zone data indicating a detection zone, wherein the detection zone includes a first portion of the first frame, and a remaining portion of the first frame is outside the detection zone;
determining, based on the first object data and the stored detection zone data, that the detected object is located partially within the detection zone and partially within the remaining portion;
generating, based on the first image data, the first object data, and the detection zone data, second image data, wherein in the second image data:
the entire detection zone is unblurred;
a second portion of the remaining portion is unblurred, the second portion corresponding to the detected object; and
a third portion of the remaining portion is blurred.
5 . The method of claim 4 , wherein the first object data is determined based on using one or more machine learning models.
6 . The method of claim 5 , wherein the one or more machine learning models comprise a convolutional neural network.
7 . The method of claim 5 , wherein the one or more machine learning models comprise a visual transformer.
8 . The method of claim 5 , wherein the first object data comprises bounding box data.
9 . The method of claim 8 , wherein the bounding box data indicates an x position of a first corner, a y position of a first corner, a width, and a height.
10 . The method of claim 8 , wherein the first set of pixel locations are pixel locations within a bounding box defined by the bounding box data.
11 . The method of claim 4 , wherein the method comprises determining that the first set of pixel locations comprises one or more pixel locations outside of the detection zone.
12 . The method of claim 4 , wherein the method comprises determining an intersection over union value based on a bounding box for the detected object and a bounding box for the detection zone, and wherein the determining that the detected object is located within the detection zone is based on the determining of the intersection over union value.
13 . The method of claim 4 , wherein the method comprises
determining an intersection over union value based on a bounding box for the detected object and a bounding box for the detection zone; and
comparing the intersection over union value to a threshold;
wherein the determining that the detected object is located within the detection zone is based on the determining of the intersection over union value.
14 . The method of claim 4 , wherein the method comprises receiving third image data generated by the camera of the electronic device, the third image data representing a second frame of a video;
determining, based on the first object data, predicted object data indicating a predicted position of the detected object;
determining, based on the third image data, second object data indicating a third set of pixel locations corresponding to a second detected object;
determining, based on the predicted object data and the second object data, that the detected object and the second detected object correspond to the same object;
accessing stored detection zone data indicating a second detection zone;
determining, based on the second object data and the stored detection zone data, that the second detected object is not located within the second detection zone;
based on the determining that the detected object and the second detected object correspond to the same object,
generating, based on the third image data, the second object data, and the detection zone data, third image data representing a blurred version of the second frame, wherein the blurred version of the second frame does not include blurring for the third set of pixel locations.
15 . The method of claim 14 , wherein the method comprises determining that a second time associated with the second frame is within a configured threshold amount of time to a first time associated with the first frame, and wherein the third image data does not include blurring for the third set of pixel locations based on the determining that the second time is within the configured threshold amount of time to the first time.