Open LemurPwned opened 1 year ago
Probably some visualisation of sampled frames would be useful.
Perhaps for some better reduction post-sampling, we can do some clustering and sample once per cluster? Probably also doable on-the-go with smaller network such as likes of the resnet50.
Probably some visualisation of sampled frames would be useful.
Perhaps for some better reduction post-sampling, we can do some clustering and sample once per cluster? Probably also doable on-the-go with smaller network such as likes of the resnet50.