Deploy first ML model to predict viral moments in live streams

pillargg / pillar_algos

Finds best timestamps to cut at

https://docs.pillar.gg/pillar_algos/

GNU General Public License v3.0

1 stars 0 forks source link

Open gatesyp opened 2 years ago

gatesyp commented 2 years ago

We want to deploy a better version of ClipFinder that is tuned to 'predicting viral moments in livestreams'.

Here is a very rough roadmap:

[X] Create initial dataset
[x] Iterate current statistical pillar_algos to experiment + try to fit model to dataset
[ ] Create ML model that outperforms previous statistical model
[x] Integrate Metaflow with our AWS environment
[ ] Deploy model to AWS

pomkos commented 2 years ago

re ML model: 75% used to train, 25% used to test. We will use the CCC dataset.

pomkos commented 2 years ago

Each CCC will be labeled as viral or not. Virality defined by @gatesyp.

Binary categorization ML model will be used to predict whether a clip will go viral or not.