3 ms·
One interesting feature of ML-based image encoders is that it might be hard to evaluate them with standard benchmarks, because those are likely to be part of th
by codeflo 4y ago
One interesting feature of ML-based image encoders is that it might be hard to evaluate them with standard benchmarks, because those are likely to be part of the training set, simply by virtue of being scraped from the web. How many copies of Lenna has Stable Diffusion been trained with? It’s on so many websites.
- zxexz 4y agoWe might enter a time when every time a new model/compression algo is introduced, a new series of benchmark images may need to be introduced/taken and ALL historical benchmarks of major compression algos redone on the new images.