This attributes uses YOLOv7 models for object detection. Libtorch is used for detecting and have to be installed.

Returns areas with objects recognized by YOLOv7 and having a probability not lower than Probability.

Not implemented

Not implemented

• Name (string, must be set) attribute unique name
• Type (string, must be "TorchYOLOv7") attribute type
• Weight (double, default: 1) attribute weight
• Probability (double, default: 0.75) acceptable detection accuracy, used in Detect mode.
• Contour (bool, default: true) Returns the contour of the attribute if true.

• model_path (string, must be set) The path to the weight .pt file, which must be exported from pytorch (see below).
• input_size (int, default: 640) Width and height for CNN input, for most YOLOv7 models is 640.
• labels (string, default: none) if provided adds labels to info and visualization. Labels format is .txt file where each label on each line.
• force_cuda (int: default: 0) if not 0, tries to run on CUDA (needs proper installation of it)

<?xml version="1.0" ?>

<AttributeLib>
    
    
    <Attribute Name="torch" Type="torchyolov7" model_path="/home/anton/Projects/yolov7/yolov7-tiny.torchscript.pt" input_size="640" Probability="0.2">
        <Filter Type="NMS" threshold="0.5"/>
    </Attribute>

</AttributeLib>

<SimpleObjectBase>  
    
     <SimpleObject Name="torch" ID="0"  Probability="0.2">
        <Attribute Type="Detect">torch</Attribute>                            
    </SimpleObject>
    
</SimpleObjectBase>

Use export.py script from original repository like:

python export.py --weights yolov7-tiny.pt --grid

This will create file yolov7-tiny.torchscript.pt that can be used.