I extended this insight to note that while tabular data is less than 20 % of enterprise data, EDA should also cover images, videos, and audio, mirroring the broader multimodal push seen in gans explained progress and limitations.
Irony is that tabular, rows and columns data mainly represent less than 20% of the overall data collected at startups and enterprises.
If EDA can be done on tabular data, it should be doable on images, videos, audio data.
I announced Labellerr’s new “Exploratory Image Analysis” (EIA) feature, which lets computer‑vision engineers explore image datasets directly from the data lake while respecting RBAC policies.
The feature reduces network load by >80 % compared with downloading full image collections.