Paddle Text Recognition

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

1+ Likes

Model Details

Model Description

PaddleOCR is a highly efficient and flexible optical character recognition (OCR) system designed for real-world applications. It supports over 80 languages and is optimized for deployment in resource-constrained environments. It achieves high recognition performance on multi-language text, including complex scripts, while remaining efficient for real-time use. It also offers robust capabilities for detecting and recognizing text in challenging conditions such as low-quality images or varying text orientations.

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

Training Details

The model was trained using model distilation from to PPLCNetV3. Training was performed on the dataset.

Testing Details

Metrics

The model was tested on 300 images of different real application scenarios to evaluate the overall OCR system, including contract samples, license plates, nameplates, train tickets, test sheets, forms, certificates, street view images, business cards, digital meter, etc.

Model Name	Number of images	F-score
PP-OCRv4	300	0.5224

Technical Specifications

Input/Output Details

Input:
- Name: x
  - Info: 0-255 BGR un-normalized image.
Output:
- Name: output
  - Info: A sequence of 40 characters, each described as classifiction probability over 97 possible symbols. Blank symbol is at index 0 and is removed by default.

Throughput

Model variant: paddle-text-recognition:320x48

• Input shape: [1, 3, 48, 320] • Output shape: [1, 40, 97]

• GFLOPs: 0.733

Platform	Precision	Throughput (infs/sec)	Power Consumption (W)
RVC2	FP16	32.34	N/A
RVC4	FP16	117.31	2.94
RVC4	INT8	664.11	2.47

* 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/paddle-text-recognition:320x48"
)

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

Inspect model head(s):

ClassificationSequenceParser that outputs message (classes (letters) and their respective scores).

Get parsed output(s):

while pipeline.isRuning():
    parser_output: ClassificationSequenceParser = 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/paddle-text-recognition:320x48

If you have OAK 4 camera you can try running the example.

A lightweight OCR model developed by PaddlePaddle.
License	Apache 2.0 Commercial use
Downloads	1189
Tasks	Classification
Model Types	ONNX

Model Variants

Name	Version	Available For	Created At	Deploy
		RVC2, RVC4	2 days ago