Custom Trained AI Models for object detection in images or video.
Detects logos of various delivery services
labels: amazon, fedex, ups, usps, uhaul, dhl
| Model | Module | CodeProject.AI Location |
|---|---|---|
| delivery.pt | YOLOv5 6.2 | ObjectDetectionsYOLOv5-6.2\custom-models |
| delivery.onnx | YOLOv5 NET | ObjectDetectionYOLOv5Net\custom-models |
| delivery.pt | YOLOv8 | ObjectDetectionYOLOv8\custom-models |
| delivery.onnx | YOLOv8 NET | not implemented in CPAI |
| delivery.pt | YOLO11 | not yet implemented in CPAI |
Go to the CodeProject.AI Server Dashboard (localhost:32168) and click on InstallModules. Make sure the YOLO module that you want to use is installed. Download the corresponding delivery module from this GitHub repository and place it in the custom-models folder for that module.
Note
The default install location for CodeProject.AI modules is C:\Program Files\CodeProject\AI\modules
After copying in any new model, be sure to restart the CodeProject.AI server. You can have multiple modules installed, but generally you only want one YOLO module running at one time.
From the CodeProject.AI Server Dashboard (localhost:32168), click on the blue button for the CodeProject.AI Explorer. At the bottom of the module you are running, pick delivery from the drop-down. Select an image that contains a delivery logo and press the orange button labeled Custom Detect.
Once you have confirmed that the model is correctly running and detecting delivery logos, you can then add it to a camera in Blue Iris.
Note
It's highly recommended that you create a clone of an existing camera for detecting delivery logos so that you don't have multiple models trying to detect objects. It will also help with isolating alerts.
In Blue Iris, after creating a clone of an existing camera, right-click and choose Camera settings.... Go to the Alert tab and check the box to Confirm Alert with AI then click the AI Configuration button.
| Field | Value | Notes |
|---|---|---|
| To confirm | amazon,dhl,fedex,ups,usps,uhaul | You can choose a subset of these if you don't want all |
| To cancel | DoNotCancel | This lets BI keep checking images on a moving vehicle until it sees a logo |
| Custom models | delivery | For best results, make sure you have turned off the default object detection, or use delivery,objects:0 |
| Mark as vehicle* | amazon,dhl,fedex,ups,usps,uhaul | All the labels should be considered vehicles. *This is a global setting (not per camera) now. Add these to the list of vehicle objects in the global AI settings. |
| Use mainstream if available | Checked | Recommended for better detections and notifications |
This work is licensed under CC BY-NC-SA 4.0
