- Number
- 12165395
- Published
- 2024-12-10
- Filed
- 2017-12-04
- Assignee
- DISNEY ENTERPRISES, INC.
- Inventors
- Carr; George Peter Kenneth et al.
- CPC
- A63B24/0006; G06N3/0442; G06N3/09; G06N3/092; G06V10/454; G06V10/70; G06V10/82; G06V20/41; G06V40/23
- Verdict
- Low Notable software
- Source
- Google Patents · FreePatentsOnline
The keeper's note
Deep imitation-learning motion-ghosting technique.
Abstract
One embodiment provides a method, comprising: training, using deep imitation learning, a neural network associated with a predetermined ghosting model to predict player movements for at least one player during at least one sequence in a game; receiving, at an information handling device, tracking data associated with a player movement path for at least one player during the at least one sequence; analyzing, using a processor, the tracking data to determine at least one feature associated with the at least one player at a plurality of predetermined time points during the at least one sequence; and determining, using the predetermined ghosting model and the at least one feature, a ghosted movement path for the at least one player beginning from one of the plurality of predetermined time points. Other aspects are described and claimed.
Background
BACKGROUND (1) Current state-of-the-art sports statistics compare players and teams to league average performance. For example, metrics such as “Wins-above-Replacement” (WAR) in baseball, “Expected Point Value” (EPV) in basketball and “Expected Goal Value” (EGV) in soccer and hockey are now commonplace in performance analysis. Such measures provide analysts with a variety of useful statistical information such as, for example, how a player or team compares to other players or teams in their respective league or how a player's or team's current performance compares to their expected performance. BRIEF SUMMARY (2) In summary, one aspect provides a method, comprising: training, using deep imitation learning, a neural network associated with a predetermined ghosting model to predict player movements for at least one player during at least one sequence in a game; receiving, at an information handling device, tracking data associated with a player movement path for at least one player during the at least one sequence; analyzing, using a processor, the tracking data to determine at least one feature associated with the at least one player at a plurality of predetermined time points during the at least one sequence; and determining, using the predetermined ghosting model and the at least one feature, a ghosted movement path for the at least one player beginning from one of the plurality of predetermined time points. (3) Another aspect provides an information handling device, comprisi
Claims
1. A method, comprising: training, using deep imitation learning, a neural network associated with a predetermined ghosting model to predict player movements for at least one player during at least one sequence in a game by back-propagating a loss between the predicted player movements and actual player movement, based on a difference of dynamic positions of the actual player movement compared to the predicted player movement for at least one player during at least one sequence in a game; defining a role for the at least one player prior to a plurality of sequences in the game, wherein the role is an expected play position for the at least one player during the at least one sequence in the game; training the neural network to align roles of a plurality of players ordered in a strategic form to compare plays of the plurality of players that align; receiving, at an information handling device, tracking data associated with a player movement path for the at least one player during the at least one sequence; receiving, at the information handling device, a context metric for one of the at least one player or an opposing player during the at least one sequence; analyzing, using a processor, the tracking data to determine at least one feature associated with the at least one player at a plurality of predetermined time points during the at least one sequence; and determining, using the predetermined ghosting model, the context metric and the at least one feature, a ghosted movement path for the at least one player beginning from one of the plurality of predetermined time point, wherein the ghosted movement path for the at least one player. ||
14. An information handling device, comprising: a processor; a memory device that stores instructions executable by the processor to: train, using deep imitation learning, a neural network associated with a predetermined ghosting model to predict player movements for at least one player during at least one sequence in a game by back-propagating a loss between the predicted player movements and actual player movement, based on a difference of dynamic positions of the actual player movement compared to the predicted player movement for at least one player during at least one sequence in a game; define a role for the at least one player prior to a plurality of sequences in the game, wherein the role is an expected play position for the at least one player during the at least one sequence in the game; train the neural network to align roles of a plurality of players ordered in a strategic form to compare plays of the plurality of players that align; receive tracking data associated with a player movement path for the at least one player during the at least one sequence; receive a context metric for one of the at least one player or an opposing player during the at least one sequence; analyze the tracking data to determine at least one feature associated with the at least one player at a plurality of predetermined time points during the at least one sequence; and determine, using the predetermined ghosting model, the context metric and the at least one feature, a ghosted movement path for the at least one player beginning from one of the plurality of predetermined time points. ||
23. A product, comprising: a non-transitory storage device that stores code, the code being executable by a processor and comprising: code that trains, using deep imitation learning, a neural network associated with a predetermined ghosting model to predict player movements for at least one player during at least one sequence in a game by back-propagating a loss between the predicted player movements and actual player movement, based on a difference of dynamic positions of the actual player movement compared to the predicted player movement for at least one player during at least one sequence in a game; code that defines a role for the at least one player prior to a plurality of sequences in the game, wherein the role is an expected play position for the at least one player during the at least one sequence in the game; code that trains the neural network to align roles of a plurality of players ordered in a strategic form to compare plays of the plurality of players that align; code that receives tracking data associated with a player movement path for the at least one player during the at least one sequence; code that receives a context metric for one of the at least one player or an opposing player during the at least one sequence; code that analyzes the tracking data to determine at least one feature associated with the at least one player at a plurality of predetermined time points during the at least one sequence; and code that determines, using the predetermined ghosting model, the context metric and the at least one feature, a ghosted movement path for the at least one player beginning from one of the plurality of predetermined time points.