Partial inference path technology in general object detection networks for efficient video processing
AssigneeIntel Corporation
InventorsByungseok Roh, Hyunjoon Lee, Seok-Yong Byun, Minje Park
Life of this patent
Estimating the term…
Abstract
Systems, apparatuses and methods may provide for technology that generates, by a full inference path of a neural network, a first detection result associated with one or more objects in a first video frame. The technology may also generate, by a partial inference path of the neural network, a second detection result based on the first detection result, wherein the second detection result corresponds to a second video frame that is subsequent to the first video frame.
Claims25
A computing system comprising:
- a network controller to receive a first video frame and a second video frame that is subsequent to the first video frame;
- a processor coupled to the network controller;
- and
- a memory coupled to the processor, wherein the memory includes a set of instructions, which when executed by the processor, cause the computing system to:
- generate, by a full inference path of a neural network, a first detection result associated with one or more objects in the first video frame, wherein the full inference path includes a plurality of layers that generates the first detection result,
- detect the second video frame, and
- generate, by a partial inference path of the neural network that is to bypass one or more layers of the full inference path, a second detection result based on the first detection result, wherein the second detection result corresponds to the second video frame.
A semiconductor apparatus comprising:
- one or more substrates;
- and
- logic coupled to the one or more substrates, wherein the logic is implemented at least partly in one or more of configurable logic or fixed-functionality hardware logic, the logic coupled to the one or more substrates to:
- generate, by a full inference path of a neural network, a first detection result associated with one or more objects in a first video frame, wherein the full inference path includes a plurality of layers that generates the first detection result;
- detect a second video frame that is subsequent to the first video frame;
- and
- generate, by a partial inference path of the neural network that is to bypass one or more layers of the full inference path, a second detection result based on the first detection result, wherein the second detection result corresponds to the second video frame.
At least one non-transitory computer readable storage medium comprising a set of instructions, which when executed by a computing system, cause the computing system to:
- generate, by a full inference path of a neural network, a first detection result associated with one or more objects in a first video frame, wherein the full inference path includes a plurality of layers that generates the first detection result;
- detect a second video frame that is subsequent to the first video frame;
- and
- generate, by a partial inference path of the neural network that is to bypass one or more layers of the full inference path, a second detection result based on the first detection result, wherein the second detection result corresponds to the second video frame.
A method comprising:
- generating, by a full inference path of a neural network, a first detection result associated with one or more objects in a first video frame, wherein the full inference path includes a plurality of layers that generates the first detection result;
- detecting a second video frame that is subsequent to the first video frame;
- and
- generating, by a partial inference path of the neural network that bypasses one or more layers of the full inference path, a second detection result based on the first detection result, wherein the second detection result corresponds to the second video frame.
Dependencies are inferred from the claim text.