eto-ai / rikai

Parquet-based ML data format optimized for working with unstructured data
https://rikai.readthedocs.io/en/latest/
Apache License 2.0
137 stars 19 forks source link

Fix pandas writer issue #687

Closed changhiskhan closed 2 years ago

changhiskhan commented 2 years ago

The pandas writer uses ToDict but Image.to_dict converted the bytes to string. When this string is then read back, the Image constructor thinks that it is actually the uri.

It looks like I originally made it convert to string, but I couldn't remember why. i searched through usages of to_dict and I didn't find any instance that really required the result to be non-binary.