It is true (by definition) that IP cameras have always been “IoT capable” devices! And thus things like Gartner’s Hype Cycle might have less relevance to this community than might otherwise be the case; indeed it may be that the collective experience of this community has lessons to impart to IoT…
1. Typical IoT devices are $10s or even $singles, whereas cameras are $100s and occasionally $1000s. So it would follow that cameras need to do more than the typical IoT thing; obviously this includes sophisticated sensing but might also include a mid-span / infrastructure role, where an IoT camera (powered and networked over Ethernet) acts as a local gateway for nearby smaller things (battery powered and with short range wireless comms). Thus one might see short range wireless added to cameras to support IoT.
2. Cameras are IP (obviously) but typically they’re not accessible on the public Internet, and often completely isolated within a given facility; not even on an enterprise intranet. The intention with IoT is at least the latter, and this is quite problematic (for example if an industrial robot and a camera viewing it were compromised via IoT then people could easily be killed).
3. Surveillance networks resemble broadcast television networks (but with the peculiarity of video running in the opposite direction) in there is very little communication between edge nodes. There are occasional exceptions: an IP access control might trigger an IP camera to move to a preset; a fixed camera (or a 3D field device like a Bosch/DENSO area sensor) might similarly trigger a PTZ camera to move to point of interest, but these are rare. Thus current surveillance networks might be expected to (like broadcast networks) follow Sarnoff’s “Law”; where the economic value of the network is directly proportional to the number of cameras. I would say right now surveillance networks are worse than Sarnoff's "Law": value is less than linear in the number of cameras!
With IoT in contrast it is expected (and assumed) that there will be “network effects” (Metcalfe’s “Law”) arising from edge to edge communication, and that these will lead to economic value that is better than linear in the number of IoT things (including cameras). The hope is to follow the Internet: Google makes $$$$ from finding and exploring links between edges of the WWW graph (i.e. web pages) so IoT pioneers would hope to do something similarly lucrative.
Back in the surveillance world, the increasing prevalence of "edge" ("edginess"??) in cameras might be an enabler giving rise to network effects in the future.
4. Most IoT devices produce relatively simple data. For example a temperature sensors produces (time, temperature) pairs or (time, temparature, location) triples. Such data is easily ingested in IT systems: it can be stored and queried in a database in an obvious way, it can be related to other data (join by time and/or location) and can also be input into more sophisticated systems (such as machine learning systems) in order to extract value from the data.
In contrast (and despite much effort) there are still huge challenges in going from video to data. Simply storing video in a database (of whatever type) and allowing it to be queried by time and camera doesn’t do much since the video needs to be interpreted (in contrast the interpretation of a temperature reading is straightforward and obvious). Again, there is progress: LPR cameras can provide (time,location,vehicle) triples; people counters can (arguably) provide (time range, number of people, location) triples; POS systems linked to camera with face detection can provide (transaction, facial image) pairs; but this has been slow and ad hoc.
5. Organisation of the Internet (and IoT?) is substantially decentralised and automated. For example, IP routing is maintained by decentralised algorithms communicating by standard protocols (BGP etc.). This is key to scale (both technically and in terms of participation) and scale in turn is key to the economic value (back in 1987 we had a few 1000 nodes and no-one knew or cared).