Event Center
The Event Center is where you go back over what happened. By the end of this page you can filter down to the handful of events you want, find "a person in white" by describing it in plain language, watch the few seconds of video around an event, flag what the system got wrong so it does better later, and safely clear out the records you do not want to keep.
Before you start
1. Reading a single event
The event list is on the left; click one and the details open on the right. What the details give you:
- Snapshot / Video tabs — the snapshot is that one frame, the video is the clip a few seconds either side of the event
- Show detection boxes: draws the boxes that were around the targets at the time; events with no box, such as plain motion, do not show this switch
- Show trajectory: the route the target took across the frame, with dwell times and a "Left the frame" marker
- Event info: type, confidence, status, the camera it came from, event time, related recording
- Detection details: the fields specific to this kind of event — identity and similarity for a face, plate and color for a vehicle, descent velocity and time on the ground for a fall, zone name and direction for a zone event, and so on
- AI tags: structured tags pulled out of the AI description (people count, clothing color, what they carried, the scene…) — only present when the AI description is on
Two handy little moves
2. Filtering down to the ones you want
- 1
Filter by type
Where: the type chips above the list. What to do: click a type. A broad type expands into its narrow ones — Pet covers cat / dog / bird, Vehicle covers car / truck / bus / motorcycle / bicycle. How you know it worked: only that type is left in the list.
- 2
Filter by camera
What to do: use the "Filter by camera" dropdown. Note: disabled and archived cameras are in the dropdown too — their past events are still in the database, and leaving them out would make those events unreachable.
- 3
Clear the unread dots
What to do: the header shows "N · UNREAD M"; click "Mark all read". How you know it worked: the unread count drops to zero.
Search and the type chips are one or the other
3. Searching events in plain language
- 1
Just describe what you are looking for
What to do: type something like "person in white", "courier" or "red bag" into the search box. Note: a single character does not start a search — you need at least two, and it begins half a second after you stop typing.
- 2
Notice when it has fallen back
How you know it worked: normally the results just appear. If a line shows up at the top saying "The semantic search model hasn't been downloaded yet", "Semantic search is unavailable right now" or "Cloud semantic search couldn't be used this time", then you are only getting keyword matches, not a real search by meaning — for the first one, go to Settings → AI Summary and download the model.
- 3
Back to the default list
What to do: empty the search box, or click "Clear search".
Only events that already have a description can be found
4. Watching the event clip
The snapshot is a single frame; the clip is the few seconds of video around the event, and you do not have to go over to the playback page for it.
- 1
Play
What to do: switch to the "Video" tab, or click "Play clip". How you know it worked: "Preparing clip…" appears and then playback starts.
- 2
When it cannot play, it tells you why
"No clip available for this event" = that camera was not recording when the event happened, so there is only a snapshot. "Event spans recording segments; see Playback for the full sequence" = the event straddles two recording segments, the clip only caught one of them, so watch the whole thing on the playback page.
For a clip to exist, it had to be recording at the time
5. Flagging what it got wrong
A cat taken for a dog, a tree shadow taken for a person — click "Mark as false positive". It does not delete the event; it sends it to the front of the Training page, and once enough have piled up you can train a model that understands your own scenes better.
- 1
Flag it
What to do: click "Mark as false positive" in the event details. How you know it worked: it says "Marked as false positive — it will be prioritized in the training queue".
- 2
Undo a wrong flag
What to do: click "Undo false positive" in the same place.
Flagging does not take effect immediately
6. Deleting events
- 1
Delete one
What to do: click "Delete" in the event details and confirm.
- 2
Clear everything
What to do: click "Clear" above the list. Note: this deletes every event record, cannot be undone, and there is no "only the old ones" option.
- 3
Bulk delete by criteria (recommended)
Where: File manager → Snapshots. What to do: filter by From / To date and type, tick what you want ("Select all" is there), then click "Delete selected". How you know it worked: it says "Deleted N snapshot(s)" — the event records and the images go together.
- 4
Export a copy before you delete
What to do: on the same page, tick your selection and click "Export selected" to download them as a zip. How you know it worked: the file names look like
20260617-120000_LivingRoom_person_1234.jpg, with the time first, so sorting by name after unzipping sorts by time.
7. Common questions
- It is in the events but the phone stayed silent — see the checklist in the last section of Notifications & Push.
- Related recording says "None" — for a fresh event, the recording segment it belongs to has not been closed off yet; it fills in by itself once that is done. If it stays "None", that camera really was not recording at the time.
- I want to see the whole thing — click "Jump to playback"; it starts 5 seconds before the event.
- Types I do not care about keep showing up in the list — instead of filtering them out over and over, demote that event type under Settings → Notifications, or cut it off at the source with an ignore zone.
Related guides
- Notifications & Push — getting these events onto your phone and into a DingTalk / Feishu group
- Zones & Tripwires — fewer unwanted events, cut off at the source
- Retention & Disk Quota — event snapshots take space too, and this is where you manage it
- AI Detection — how events are produced in the first place
Was this page helpful?