EfficientNet-Lite

Our new model ZOO works with DepthAI V3. Find out more in our documentation.

0 Likes

Model Details

Model Description

EfficientNet-Lite is a mobile-friendly pure convolutional model (ConvNet) that proposes a new scaling method that uniformly scales all dimensions of depth/width/resolution using a simple yet highly effective compound coefficient. It can classify the image into one of 1000 classes.

Developed by: Google
Shared by:
Model type: Computer vision
License:
Resources for more information:

Training Details

Training Data

EfficientNet-Lite model is trained on at resolution 300x300.

For more information about training data check the .

Testing Details

Metrics

Metric	Value
Top-1 Acc.	81.50%

For more information please check the .

Technical Specifications

Input/Output Details

Input:
- Name: images
  - Info: NCHW BGR 0-255 image.
Output:
- Name: Sofmax
  - Info: Softmaxed scores for 1000 classes.

Model Architecture

The EfficientNet-Lite network consists of:

Stem: Initial layer with a standard convolution followed by a batch normalization and a ReLU6 activation.
Body: Consists of a series of MBConv blocks with different configurations. Each block includes depthwise separable convolutions and squeeze-and-excitation layers.
Head: Includes a final convolutional block, followed by a global average pooling layer.

Please consult the and for more information on model architecture.

Throughput

Model variant: efficientnet-lite:lite0-224x224

• Input shape: [1, 3, 224, 224] • Output shape: [1, 1000]

• Params (M): 4.631 • GFLOPs: 0.405

Platform	Precision	Throughput (infs/sec)	Power Consumption (W)
RVC2	FP16	55.23	N/A
RVC4	INT8	601.79	3.16

Model variant: efficientnet-lite:lite4-300x300

• Input shape: [1, 3, 300, 300] • Output shape: [1, 1000]

• Params (M): 12.950 • GFLOPs: 2.636

Platform	Precision	Throughput (infs/sec)	Power Consumption (W)
RVC2	FP16	8.58	N/A
RVC4	INT8	601.88	4.74

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

* Parameters and FLOPs are obtained from the package.

Quantization

RVC4 version of the model was quantized using a custom dataset. This was created by taking 1000-image subset of dataset.

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/efficientnet-lite:lite0-224x224"
)

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

Inspect model head(s):

ClassificationParser that outputs message (detected classes and scores).

Get parsed output(s):

while pipeline.isRuning():
    parser_output: Classifications = parser_output_queue.get()

Example

You can quickly run the model using our script. It automatically downloads the model, creates a DepthAI pipeline, runs the inference, and displays the results using our DepthAI visualizer tool. To try it out, run:

python3 main.py \
    --model luxonis/efficientnet-lite:lite0-224x224

1000-class image classification.
License	Apache 2.0 Commercial use
Downloads	504
Tasks	Classification
Model Types	ONNX

Model Variants

Name	Version	Available For	Created At	Deploy
		RVC2, RVC4	10 months ago
		RVC2, RVC4	10 months ago