A noticeable number of pages are not correctly loaded

[stefanmagureanu]

There are a couple of thousand pages (less than 10k - but still many) that do not load properly. It is hard to tell whether it is because the MHTML doesn't render correctly in the WTL browser or because the MHTML itself was not saved properly.

[stefanmagureanu]

One way of identifying pages that are guaranteed to load properly is to use only pages that have corresponding screenshots in the screenshot dataset. This will exclude more pages than just the ones that do not load properly as WTL snapshots. Around 10k pages do not render appropriately for computer vision applications.