search

pesser / edflow

Reduce boilerplate code for your ML projects. TensorFlow and PyTorch.
https://edflow.readthedocs.io/
MIT License
24 stars 13 forks source link

Sequence base step #233

Closed hperrot closed 4 years ago

hperrot commented 4 years ago

Add a base_step argument to the SequenceDataset. The base fids have to be divisible by the base_step. This allows a smaller sequence dataset because the sequences don't necessarily start at every frame in the dataset. For example, with base_step of 10, the sequences start at frame 0, 10, 20, ... instead of 0, 1, 2, 3, ... This makes the dataset effectively smaller and samples the examples equally within the possible sequences.

pesser commented 4 years ago

looks good to me, but since I'm not currently working with sequence datasets let's have @jhaux take a look