MediaPipe Hand Landmarker

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

0 Likes

Model Details

Model Description

The MediaPipe Hand landmark model performs precise keypoint localization of 21 3D hand-knuckle coordinates inside the detected hand regions via regression, that is direct coordinate prediction. The model learns a consistent internal hand pose representation and is robust even to partially visible hands and self-occlusions.

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

Training Details

Training Data

For training the hand landmarker 3 different datasets were used:

In-the-wild dataset contains 6K images of a large variety, e.g. geographical diversity, various lighting conditions, and hand appearance. The limitation of this dataset is that it doesn’t contain complex articulation of hands.
In-house collected gesture dataset: This dataset contains 10K images that cover various angles of all physically possible hand gestures. The limitation of this dataset is that it’s collected from only 30 people with limited variation in background. The in-the-wild and in-house datasets are great complements to each other to improve robustness.
Synthetic dataset: To even better cover the possible hand poses and provide additional supervision for depth, we render a high-quality synthetic hand model over various backgrounds and map it to the corresponding 3D coordinates.

Testing Details

Metrics

Evaluation results are taken from .

Metric	Value
Mean Squared Error	11.83

Technical Specifications

Input/Output Details

Input:
- Name: input_1
  - Info: NCHW BGR un-normalized image
Output:
- Name: multiple (see NN archive)
  - Info: Hand keypoints, handedness (left or right hand), confidence score

Model Architecture

The model has a shared feature extractor and 3 separate heads for 3 outputs (hand landmarks, hand presence, and handedness). Each head is trained by correspondent datasets.

Please consult for more information on model architecture.

Throughput

Model variant: mediapipe-hand-landmarker:224x224

• Input shape: [1, 3, 224, 224] • Output shapes: [[1, 63], [1, 1], [1, 1], [1, 63]]

Platform	Precision	Throughput (infs/sec)	Power Consumption (W)
RVC2	FP16	60.44	N/A
RVC4	INT8	720.88	2.60

* 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 50-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/mediapipe-hand-landmarker:224x224"
)

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

Inspect model head(s):

KeypointParser that outputs message (detected hand keypoints).
RegressionParser that outputs message (score).
RegressionParser that outputs message (handedness).

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

keypoints_parser_output_queue = nn.getOutput(0).createOutputQueue()
score_parser_output_queue = nn.getOutput(1).createOutputQueue()
handedness_parser_output_queue = nn.getOutput(2).createOutputQueue()

Get parsed output(s):

while pipeline.isRuning():
    keypoints_parser_output: Keypoints = keypoints_parser_output_queue.get()
    score_parser_output: Predictions = score_parser_output_queue.get()
    handedness_parser_output: Predictions = handedness_parser_output_queue.get()

Example

You can quickly run the model using our example.

The example demonstrates how to build a 2-stage DepthAI pipeline consisting of a palm detection model and a hand landmark model. It automatically downloads the models, creates a DepthAI pipeline, runs the inference, and displays the results using our DepthAI visualizer tool.

To try it out, run:

python3 main.py

Hand Landmark model.
License	Apache 2.0 Commercial use
Downloads	1299
Tasks	Keypoint Detection
Model Types	ONNX

Model Variants

Name	Version	Available For	Created At	Deploy
		RVC2, RVC3, RVC4	9 months ago