This is a modified version of the streaming demo that shows how to detect bounding boxes in a neural network model.
Please use eBusPlayer first to configure the image quality, gain, and exposure settings.
This demo uses the YoloV3 model to detect bounding boxes in the image, please
download the precompiled model for Bottlenose from here, a good basic example
is yolov3_1_416_416_3.tar.
You can use the Bottlenose utility to download the model to the camera, or provide the path to the model as commandline parameter to the demo.
The Python script shows how to programmatically enable chunk data transmission for Bounding Box data, and optionally how to programmatically upload a model to Bottlenose. Note the model files are packaged with meta information, please provide the Labforge model (.tar) as is for the upload. Bottlenose will validate the file after the upload.
Change the following demo parameters to your desired settings in the demo.py file.
| Parameter | Description |
|---|---|
confidence |
Confidence to accept a valid detection. |
python demo.py ?ModelFile ?MAC
# ModelFile - (optional) path to model file to upload to Bottlenose
# MAC - (optional) mac address of Bottlenose to connect to or first one
Please contact us for training services and custom model porting.
Back to Samples