Halliday & Matthiessen (2014: 505):
Here what is being perceived is again some action or event; the clause is typically imperfective, but sometimes perfective (without to) to highlight the end state as distinct from the process (cf. Kirsner & Thompson, 1976):
[imperfective:]
I saw the boats turning/(passive) being turned
[perfective:]
I saw the boats turn/(passive) turned
If the embedded clause is used as Postmodifier the Head noun is usually one of sight or sound: I heard the noise of ... , I had a view of ... , etc. (cf. the smell of something burning); …
In this case the clause is always imperfective.