SYSTEMS AND METHODS FOR DETERMINING MOTION VECTORS

1 . A computer-implemented method comprising:

training, by a computing system, a model to predict motion vectors for entities in video frames;

obtaining, by the computing system, a set of frames that correspond to a first video;

providing, by the computing system, the set of frames as input to the model; and

obtaining, by the computing system, a set of motion vectors for the set of frames from the model, wherein each motion vector describes a trajectory of at least one entity in the set of frames.

2 . The computer-implemented method of claim 1 , wherein an entity is one of a pixel, a block of pixels, an object, or a frame.

3 . The computer-implemented method of claim 1 , wherein training the model further comprises:

generating, by the computing system, training data to be used for training the model, the training data describing a plurality of entities and their respective pre-computed motion vectors; and

training, by the computing system, the model using the generated training data.

4 . The computer-implemented method of claim 3 , wherein generating the training data further comprises:

obtaining, by the computing system, a set of videos for training the model, each video having a set of frames;

determining, by the computing system, a set of respective motion vectors for one or more entities in the set of frames for each video; and

causing, by the computing system, data describing the one or more entities in the set of frames to be included in the training data as an example inputs and the corresponding motion vectors for the entities to be included in the training data as example outputs.

5 . The computer-implemented method of claim 4 , wherein the set of motion vectors are optimally determined using an exhaustive motion estimation algorithm.

6 . The computer-implemented method of claim 3 , wherein the entities correspond to objects, and wherein generating the training data further comprises:

obtaining, by the computing system, a set of videos for training the model, each video having a set of frames;

identifying, by the computing system, one or more objects in the set of frames for each video;

determining, by the computing system, a set of respective motion vectors for the one or more objects; and

causing, by the computing system, data describing the one or more objects in the set of frames to be included in the training data as an example inputs and the corresponding motion vectors for the objects to be included in the training data as example outputs.

7 . The computer-implemented method of claim 3 , the method further comprising:

providing, by the computing system, data describing one or more of the entities included in the training data as input to the model;

obtaining, by the computing system, one or more respective motion vectors for the entities from the model; and

determining, by the computing system, an inaccuracy in a motion vector determined by the model for at least one entity based at least in part on the respective pre-computed motion vector of the entity.

8 . The computer-implemented method of claim 7 , the method further comprising:

causing, by the computing system, the model to be retrained based at least in part on the inaccuracy.

9 . The computer-implemented method of claim 1 , wherein the entities correspond to frames, and wherein the motion vectors provide a general motion estimation of one or more frames.

10 . The computer-implemented method of claim 9 , the method further comprising:

providing, by the computing system, the general motion estimation to at least one motion estimation algorithm, wherein the motion estimation algorithm is configured to determine one or more motion vectors for the entities based at least in part on the general motion estimation.

11 . A system comprising:

at least one processor; and

a memory storing instructions that, when executed by the at least one processor, cause the system to perform:

training a model to predict motion vectors for entities in video frames;

obtaining a set of frames that correspond to a first video;

providing the set of frames as input to the model; and

obtaining a set of motion vectors for the set of frames from the model, wherein each motion vector describes a trajectory of at least one entity in the set of frames.

12 . The system of claim 11 , wherein an entity is one of a pixel, a block of pixels, an object, or a frame.

13 . The system of claim 11 , wherein training the model further causes the system to perform:

generating training data to be used for training the model, the training data describing a plurality of entities and their respective pre-computed motion vectors; and

training the model using the generated training data.

14 . The system of claim 13 , wherein generating the training data further causes the system to perform:

obtaining a set of videos for training the model, each video having a set of frames;

determining a set of respective motion vectors for one or more entities in the set of frames for each video; and

causing data describing the one or more entities in the set of frames to be included in the training data as an example inputs and the corresponding motion vectors for the entities to be included in the training data as example outputs.

15 . The system of claim 14 , wherein the set of motion vectors are optimally determined using an exhaustive motion estimation algorithm.

16 . A non-transitory computer-readable storage medium including instructions that, when executed by at least one processor of a computing system, cause the computing system to perform a method comprising:

training a model to predict motion vectors for entities in video frames;

obtaining a set of frames that correspond to a first video;

providing the set of frames as input to the model; and

obtaining a set of motion vectors for the set of frames from the model, wherein each motion vector describes a trajectory of at least one entity in the set of frames.

17 . The non-transitory computer-readable storage medium of claim 16 , wherein an entity is one of a pixel, a block of pixels, an object, or a frame.

18 . The non-transitory computer-readable storage medium of claim 16 , wherein training the model further causes the computing system to perform:

generating training data to be used for training the model, the training data describing a plurality of entities and their respective pre-computed motion vectors; and

training the model using the generated training data.

19 . The non-transitory computer-readable storage medium of claim 18 , wherein generating the training data further causes the computing system to perform:

obtaining a set of videos for training the model, each video having a set of frames;

determining a set of respective motion vectors for one or more entities in the set of frames for each video; and

causing data describing the one or more entities in the set of frames to be included in the training data as an example inputs and the corresponding motion vectors for the entities to be included in the training data as example outputs.

20 . The non-transitory computer-readable storage medium of claim 19 , wherein the set of motion vectors are optimally determined using an exhaustive motion estimation algorithm.