tum-ei-eda / mlonmcu-models Goto Github PK

I am trying to figure out which models here belong to the official MLPerf™ Tiny Deep Learning Benchmark Models.

I am looking for the following:

• Keyword Spotting -> ??
• Visual Wake Words -> here vww/
• Image Classification -> here resnet/
• Anomaly Detection -> here toycar/

I used the md5sum to check for equivalency.

Currently this repository is just a copy of the old ml_on_mcu/data directory. It definitely needs a lot of cleanup work (e.g. remove broken or irrelevant models)

I was wondering if we get into copyright problems when we are distributing third part models as soon as MLonMCU goes open source?

Should it be enough to just mention the original source of a .tflite file and some additional metadata such as the authors (if available) or do we need to decide this on a per-model basis?

@rafzi @DanMueGri