AICam is a web application that offers an interface for using different types of Noctuai video analytics. Currently available analytics are Fire Smoke, Fire Smoke for tunnels, Fall, Intruder Detection, and PPE Inspection. The core of the application is video preview along with settings. They allow manipulation of analytics that analyze videofeed. Feed for analysis in the preview is added via RTSP streams. The output of the analysis can be seen on the user preview. Another way to be informed about analysis output is via MQTT messaging protocol. During the installation process, users can configure the names of channels that allow integration later on.

Installation process of free version

download the free version of AI-CAM from here.
unpack using the command tar -zxf ./noctuai-free.tar.gz
go into the unpacked directory and run docker-compose up -d
application will download all its components and will be ready in a couple of minutes.
go to the webpage: http://localhost
Fill out the form at https://noctuai.com/pricing/?want=free#quote_form to receive your login and password.

The installation process of the paid version

To perform the installation, you must have the proper hardware to allow the application to run. Below, you will find the minimum recommended hardware setups. For more information or recommendations regarding a larger amount of streams, please contact us at contact@noctuai.com

no. of streams	up to 10	up to 20	20 - 100
CPU	8 Cores / 16Threads - i5 10 gen	8 Cores / 16Threads - i5 10 gen	16 Cores / 24Threads (14thGEN Intel / AMD ThreadRipper)
RAM	32 GB	32 GB	64 GB
GPU	cards with compute capability >=6.5 https://developer.nvidia.com/cuda-gpus
GPU	NVIDIA® RTX™ 4060GPU / A2000 8GB	NVIDIA® RTX™4070TI SUPER lub RTX A4000 16GB	NVIDIA® RTX 5000 ADA lub RTX 4090, 32GB VRAM
Disk	1TB SSD	1TB SSD	1TB SSD
Operating System	Linux Ubuntu 24.X	Linux Ubuntu 24.X	Linux Ubuntu 24.X

The installation package (that can be downloaded via the appropriate link or received via email from our representative) contains the following:

Tar GZ archive file (app-installer.tar.gz),
LICENSE_KEY and LICENSING_PUBLIC_KEY,
Noctuai’s Docker registry credentials.

The archive fileapp-installer.tar.gz has to be unpacked to an application destination folder (the folder name can be anything), and then the nct-installer file can be run.

After accepting our EULA available at https://noctuai.com/eula there are phases to be run during the installation process:

(Re)Install Docker,
(Re)Install GPU drivers,
(Re)Install NVidia Container Toolkit,
Set up license keys,
Login to Noctuai Docker Registry,
Download and install the latest version.

The first three steps are currently experimental and might not work in Your environment!

Of course, the first three phases can be skipped if the computer already has proper software installed. It is essential to run the phases in the same order as listed. The duration of the process depends on the internet connection parameters as the process assumes downloading files from the Internet repository.

After completing the phases choosing option “Start Application” will initiate the AICam. The first start after installation can take a longer time. During this period, you won't be able to connect the streams to the AICam.

To access AICam after the installation, you have to open the website (http://localhost:8081) on the machine where it is installed or reach it via the network from another computer using the IP address of the computer where it is installed (http://xxx.xxx.xx.xxx:8081/)

Basic configuration

To enter an application, the user should visit a URL that holds a hosted application.

An account provided during the installation should be used to enter the application for the first time (administrator rights).

The administrator account has the right to create additional user accounts with different permissions if necessary.

Entering the correct credentials as administrator user opens the following screen.

Selecting Users in the left menu bar opens a page with a list of all the users created in the app.

Creating new USERS

To create a new user, click the New User button located in the top right corner of the screen. A pop-up modal window should appear with the user creation form. Note that while creating a new account,, it is best to select the the “User must change password” checkbox for basic credentials security.

After typing the necessary information, a new user account is created new user can enter the application with provided credentials.

Adding new stream

Note: To obtain the best analytics results, use streams with 16:9 Aspect Ratio.

Accessing the newly created application allows the user to see only the page with an empty list of streams. A new stream should be added to fully use the application. Every account type can create new streams. To create a new stream, click the Add Stream button in the top right corner.

After providing the necessary information, click the Submit button to add a new stream.

Depending on your stream availability, adding might take some time. Once done, a new stream with connectivity status: Connecting should appear on the list.

Now, you can start using the analytics on a given stream. To enter the stream preview and settings, click on the stream name or URL. You will see a stream preview screen with two tabs on the top: Preview and Settings.

Note: Stream preview will differ depending on the analytics used.

Stream statuses

Possible stream statuses:

🟠Connecting - trying to connect to stream
🟢Connected – stream is connected and analytics is running
🔴Disconnected - lost connection to stream. Reconnecting will be run periodically. Please wait for the connection to be established. If waiting does not solve the problem, please verify the provided stream quality.

Test events

Test events are an option standard for every analytics application. It requires a stream to be added to the list but does not require it to be connected. The stream can be in any state to send an event. After clicking the button, one event is sent to all open channels (e.g., reporting and alarming).

MQTT Configuration for integration with external systems

Our Application comes with built-in mqtt server with default mqtt topics configured.

Here are default parameters for built-in mqtt server:

MQTT port: 1883
no user / password required
default topics for analytics:

event FIRE-SMOKE - alarming.fire-smoke
reporting FIRE-SMOKE - reporting.fire-smoke
event FIRE-SMOKE-TUNNEL - alarming.fire-smoke-tunnel
reporting FIRE-SMOKE-TUNNEL - reporting.fire-smoke-tunnel
event FLARE-INSPECTION - alarming.flare-inspection
reporting FLARE-INSPECTION - reporting.flare-inspection
event PPE - alarming.ppe
reporting PPE - reporting.ppe
event FALL-DETECTION - alarming.fall-detection
event INTRUDER-DETECTION - alarming.intruder-detection
reporting INTRUDER-DETECTION - reporting.intruder-detection

It is possible to set up the application to use other MQTT servers. Please contact our representative for help.

Applications for different analytics functionalities

Detecting Fire Smoke

Preview

The Fire Smoke application allows detection of fire and smoke on video streams.

Settings

Settings page offers detection thresholds for fire and smoke - values that are responsible for triggering alarms. Any value detected in the video stream that surpasses the set value will be treated as an alarm.

The settings also allow to switch on or off and adjust alarming and reporting options. In order to adjust alarming, 2 parameters need to be set. Delay and cooldown time. Both in seconds. Delay time is time of minimal event duration that triggers an alarm. When event is spotted e.g. fire, it needs to be present on frames for the delay time duration in order to generate an event. Cooldown time is time that needs to pass before sending another event. The unit of measurement in alarming is seconds.

In order to adjust reporting mode, there are three settings that can be changed:

Analysed timeframe (minutes) is the duration of stream that is subjected to analysis.
Reporting interval (minutes) is the amount of time that has to pass once the event was sent in order t generate another one.
Cumulative event occurrence time (%) is the percentage of time measured (analysed timeframe) that has to show the event occurrence in order to generate event. e.g. when 10 minutes of video is analysed and event occurrence is set to 60%, then in order to trigger the report event, smoke or fire should be visible on at least 6 minutes of that feed.

Events

Name of default channels for the event along with environment variables names for custom adjustment.

Events format examples:

Alarming:

fire:

{

"streamId": "UUID string",

"streamName": "string",

"settings": {

"fireThreshold": 0.1,

"smokeThreshold": 0.1

"event": {

"fireConfidence" : 0.1,

"fireStatus" : "DETECTED / NOT_DETECTED"

"timestamp": "ISO Timestamp"

}

smoke:

{

"streamId": "UUID string",

"streamName": "string",

"settings": {

"fireThreshold": 0.1,

"smokeThreshold": 0.1

"event": {

"smokeConfidence" : 0.1,

"smokeStatus" : "DETECTED / NOT_DETECTED"

"timestamp": "ISO Timestamp"

}

Reporting:

{

"streamId": "3185c18a-a55e-4a72-989c-4186faa57143",

"streamName": "FS_01",

"eventType": "FIRE",

"cumulatedTimeEventOccurredWithinFrameSeconds": 24,

"currentFireThreshold": 0.5,

"currentSmokeThreshold": 0.5,

"analysisIntervalSeconds": 60,

"id": "a09a41c8-a8d6-4cd0-a946-e53f608dad0e"

}