0/1 Knapsack problem

[Petersteve]

0/1 Knapsack problem

[KaiyangZhou]

yes, after the score prediction, we average scores within the same shot to get shot level scores, then select a subset of them to construct a summary

[KaiyangZhou]

The knapsack algorithm is used to select a subset of shots with the maximum scores while ensuring the length does not exceed a threshold. An alternative way is the greedy selection, but it turned out to be inferior.

[Petersteve]

In this case we need for each frame his score ,isn't in unsupervised learning we don't need gtscore?

[KaiyangZhou]

knapsack is used in inference so the scores are predicted by the network

[Petersteve]

P t-1 , P , P t+1... is this the scores predicted by the network?