Module 02 Summary

What this module established

Detect format from the bytes and fail at load time - the alternative is an error hours into training on one unnamed file. Resizing is permanently lossy, so measure your target feature's size in pixels after the resize before assuming the model can see it.

Carry forward

• Choose the colour space from the question: HSV for colour under varying light, grayscale when colour is noise, RGB when a pretrained model expects it.
• JPEG artefacts are structured and learnable. Train on images processed the way production images will be.
• Return a reason for every rejection and count them by kind. Silent skipping hides both dataset shrinkage and upstream changes.

Before moving on

Move on when your loader rejects with reasons, and you have measured what your resize and your compression each destroy.

Mark complete