X

Name	Version	Available For	Created At	Deploy
		RVC2, RVC4	About 1 year ago

Model Details

Model Description

The Openvino vehicle-attributes-recognition-barrier-0039 model is a fast vehicle attributes classification algorithm built to run efficiently on edge devices. For each input image the model returns two classification heads:

color: a softmax output across seven color classes [white, gray, yellow, red, green, blue, black]
type: a softmax output across four type classes [car, bus, truck, van]

More information:

Original model:
Re-implemented by:
Model type: Computer Vision
License:
Resources for more information:

Training Details

The model was trained on cropped images of cars, trucks, vans and buses all of seven different colors, making it suitable for a traffic analysis scenario.

Testing Details

The below metrics are based on the underlying source model before it was converted to RVC2 and RVC4 compatible format. Results of RVC2/4 compiled model may be different.

Metrics

Color accuracy (%)

	blue	gray	yellow	green	black	white	red
blue	79.53	4.32	0.62	6.41	6.54	2.47	0.12
gray	2.53	78.01	0	1.36	1.18	16.74	0.18
yellow	0	13.9	54.01	11.21	0	10.7	10.16
green	3.79	1.52	1.52	83.33	6.06	3.03	0.76
black	0.85	1.92	0	0.32	96.1	0.74	0.07
white	1.45	10.86	0.17	2.53	0.08	84.83	0.08
red	0.89	0.3	2.18	2.18	0.3	1.88	92.27

Type accuracy (%)

	car	van	truck	bus
car	98.26	0.56	0.98	0.2
van	3.72	89.16	6.15	0.97
truck	1.71	2.46	94.27	1.56
bus	7.94	3.8	19.69	68.57

Technical Specifications

Input/Output Details

Input:
- Name: image
  - Info: NCHW BGR non-normalized image (cropped image of a vehicle)
Output:
- Name: tf.identity_1
  - Info: Softmax scores across seven color classes: [white, gray, yellow, red, green, blue, black]
- Name: output2_yolov6r2
  - Info: Softmax scores across four type classes : [car, bus, truck, van]

Model Architecture

Please consult the for more information on model architecture.

Throughput

Model variant: vehicle-attributes-classification:72x72

• Input shape: [1, 3, 72, 72] • Output shapes: [[1, 1, 1, 4], [1, 1, 1, 7]]

• Params (M): 0.626 • GFLOPs: 0.063

Platform	Precision	Throughput (infs/sec)	Power Consumption (W)
RVC2	FP16	918.01	N/A
RVC4	FP16	634.45	2.55

* Benchmarked with , using 2 threads (and the DSP runtime in balanced mode for RVC4).

* Parameters and FLOPs are obtained from the package.

Utilization

Models converted for RVC Platforms can be used for inference on OAK devices. DepthAI pipelines are used to define the information flow linking the device, inference model, and the output parser (as defined in model head(s)). Below, we present the most crucial utilization steps for the particular model. Please consult the docs for more information.

Install DAIv3 and depthai-nodes libraries:

pip install depthai
pip install depthai-nodes

Define model:

model_description = dai.NNModelDescription(
    "luxonis/vehicle-attributes-classification:72x72"
)

nn = pipeline.create(ParsingNeuralNetwork).build(
    <CameraNode>, model_description
)

Inspect model head(s):

ClassificationParser that outputs message (vehicle type).
ClassificationParser that outputs message (vehicle color).

The model is multi-headed. You can set up the queues as follows:

queue_vehicle_type = nn.getOutput(0).createOutputQueue()
queue_vehicle_color = nn.getOutput(1).createOutputQueue()

Get parsed output(s):

while pipeline.isRuning():
    vehicle_type: Classifications = queue_vehicle_type.get()
    vehicle_color: Classifications = queue_vehicle_color.get()

Example

TBA

A fast Openvino model built for classifying vehicle type and color.
License	MIT Commercial use
Downloads	2893
Tasks	Classification
Model Types	ONNX