Latent Diffusion: The Compression Ratio
Intermediate
The problem, from first principles
The point of latent diffusion is compression: the model works on far fewer numbers than the pixels. This ratio measures how much smaller the latent is than the image it encodes.
From theory to code
Implement compression_ratio(h, w, c_img, f, c_lat), the number of image values divided by the number of latent values.
Constraints
- Returns a Python float.
Hints
Hint 1
Count the image values, count the latent values with the latent grid size, and divide.
Keywords
research-paperscomputer-visiondiffusionlatent
Python
Run your code to see results
No results yet
Run your code to check it against the examples, or submit to run every test.