Seeking Assistance with Eliminating False Positives in Motion Detection

[TimG]

Ebay it was for me too ! As I said earlier, the GTX1650 was recommended by MikeLud for being good enough while not using too much power. Looking at his posts on the CPAI forum, I think he uses a much faster GPU now. It's up to you.

You would then need to re-install CPAI - after choosing a video driver that works for CUDA 11.8 - you have to delve into the specification files to verify this. I did try the latest driver, but ALPR wouldn't use the GPU, so I reverted the pc to its previous working state with Macrium.

I will be trying a newer driver this week with CPAI 2.5.1. Back-up first

[TimG]

At present, I have specified push and sound alerts in the Alerts tab, under 'Actions' on Alert.

However, I have recently seen the option under the Trigger tab, then 'When Triggered', followed by 'Immediate Actions'.

Apples and Oranges. The Alert actions only work if the Alert is verified. The Immediate action works after a trigger, and before the Alert has been verified.

[TimG]

However, it's crucial that I do reduce the latency; after inputting "unicorn" in the cancel field and reviewing the .dat file, it's evident that the system is now comprehensively scanning the total amount of images I specify in the real-time field, but this results in considerable delays in sending alerts to my phone.

Note: Previously it found something and instantly alerted you. It may even have alerted after the first image and the analysis stops. Now you have to wait for ALL images to be checked, and then for BI5 to decide which one to use.

[Sardon]

Thanks for all the above info

[Sardon]

I had assumed, rather optimistically, that the absence of sound alerts on my BI PC indicated no detection of a person. However, this assumption is entirely incorrect. There appears to be an issue with my configured Sound Alert.

sound.jpg (34.75 KiB) Viewed 1141 times

In the excerpt above, I've removed 'person' from the Required AI Objects section. If I include 'person' in the field, no sound is triggered, which is quite baffling.

In the excerpt below, I've added "person" to the Required AI Objects field for Push Notifications, and it functions as expected. I only receive a push notification on my phone if a person is detected.

push.jpg (17.88 KiB) Viewed 1141 times

Regarding the sound configuration, I experimented by removing 'person', as shown above, and the sound alert resumed functioning. However, it now activates audio for all alerts, even for the alerts that will be cancelled. I am at a loss as to why it won't alert using 'person' only, as specified in the Requires AI Object field.

Consequently, I've had to disable it for now, but I felt it important to post here as this issue has left me quite perplexed. My initial thought was that any sound alert would definitively indicate the approach of a person, but this is not the case. In fact I haven't heard anything as of late which prompted me to double check to see what was going on. I'm not getting anything with person specified in that box, am I overlooking something?

Additionally, while I'm here, I'd like to enquire about the meaning of the icon below extracted from the .dat . I understand that the orange person symbolises 'person detected' and the red circle with a white X indicates 'nothing found', but what does the icon resembling dual monitors signify?

dat-icons.jpg (11.93 KiB) Viewed 1141 times

I would greatly appreciate any insights into what might be causing the issue with the sound notification as it's now doing my head in.

[Sardon]

This morning, upon my return home, I noticed that I was not detected at all. It doesn't seem to be a problem with the CPAI itself, but rather with the motion detector. No data was sent to CP for processing. No alert registered , nor a cancelled alert.

I am curious, is it necessary to have these options enabled, or can I disable them? I've never fully grasped the optimal settings for these features, so I've kept them at their default settings. My primary objective is to ensure that everything is captured and forwarded to CPAI for appropriate processing.

If someone could shed some light please as it is a major concern that I was not detected...

object-detection.jpg (21.89 KiB) Viewed 1064 times

[TimG]

These are the nuisance trigger preventing devices I mentioned earlier that BI5 used to need to stop the nuisance triggers, but with CPAI you just throw everything at it "set sensitivity to 11" and let CPAI decide if it is a required object or not. They are de-sensitising your triggers and stopping motion detection:

1. Only trigger when object travels 16 pixels on sub stream is to prevent a trigger unless it has moved that number of pixels. Untick it.
2. Reset detector when object size exceeds 67% is for those UK days when the clouds keep passing and triggering the camera. You could try it off to see if it was that preventing your trigger, and then maybe add it back with a much higher value.

Ken has spent years adding in little tweaks to stop trees moving in the wind, or shadows going across the image triggering BI5. Now we can turn them all OFF and let CPAI make the decision. Or somewhere between

[Sardon]

After navigating through the complexities of optimising motion detection to reduce false positives with my setup, I've come to appreciate the intricate balance between technology, environment, and practical usability. This journey has been both challenging and enlightening, bringing to light several key insights that I believe are invaluable for anyone embarking on a similar quest.

My Learnings and Reflections:

Embracing Technological Experimentation: I've learnt the importance of being open to adjusting settings, experimenting with AI integration, and considering hardware upgrades. These steps are crucial in pushing the boundaries of what my current system can achieve.

The Power of Community: Engaging with this community has been instrumental in my journey. The insights, suggestions, and support from members like Pogo, TimG, and Louyo have not only broadened my understanding but also provided me with a range of strategies to tackle the challenges I faced.

Iterative Process: This process has taught me that optimising is an ongoing journey of trial, error, and learning. Each adjustment and test has brought me closer to a system that meets my specific needs.

In concluding, this thread has not only been a testament to the challenges of motion detection optimisation but also a rich resource for shared knowledge and experiences. My journey, enriched by the contributions from the community, has highlighted the importance of persistence, openness to new technologies, and the invaluable support of a community in navigating the complexities of blue iris. My efforts to refine motion detection to eliminate false positives, while sometimes frustrating, have ultimately been rewarding, offering practical insights and solutions in this thread that I hope will benefit others.

I extend my heartfelt thanks to everyone who has contributed to this discussion. Your advice and support have been incredibly appreciated, paving the way for significant improvements in my system’s effectiveness. Let’s continue to explore, share, and support each other in this ever-evolving journey.

Sardo N