dnn_text_detection

Detect and recognize texts with DB (cv::dnn::TextDetectionModel_DB) and CRNN (cv::dnn::TextRecognitionModel).

Text detection and recognition (OCR) with DB (cv::dnn::TextDetectionModel_DB) and CRNN (cv::dnn::TextRecognitionModel), by default DB (ResNet-18, trained on IC15) and CRNN (English) of opencv_zoo.

opencv_apps	3.0.0 and later
ROS	ROS One, jazzy, kilted, lyrical and later
OpenCV	OpenCV 4.5.4 or later
Group	Text detection and recognition, the nodes of which publish `~/detections` (DetectionArrayStamped), with the recognized text

Usage

# ROS 2
ros2 launch opencv_apps dnn_text_detection.launch.xml image:=/camera/image_raw debug_view:=false
# ROS 1
roslaunch opencv_apps dnn_text_detection.launch image:=/camera/image_raw debug_view:=false

Try it with the demo video

Download the input video of the demo above, publish it as image with image_publisher, and run the node on it, in the terminals:

wget https://raw.githubusercontent.com/ros-perception/opencv_apps/indigo/test/text.mp4
# terminal 1: publish the video as image in a loop
ros2 run image_publisher image_publisher_node text.mp4 --ros-args -p publish_rate:=10.0 -r image_raw:=image
# terminal 2: run the node
ros2 launch opencv_apps dnn_text_detection.launch.xml image:=image
# terminal 3: view the result
ros2 run rqt_image_view rqt_image_view /dnn_text_detection/image

In ROS 1:

# terminal 1
rosrun image_publisher image_publisher text.mp4 _publish_rate:=10 __name:=camera
# terminal 2
roslaunch opencv_apps dnn_text_detection.launch image:=camera/image_raw

See Credits of the demo videos for the license of the video.

Subscribed topics

Topic	Type
`image`	`sensor_msgs/msg/Image`

Published topics

Topic	Type
`~/detections`	opencv_apps/msg/DetectionArrayStamped
`~/image`	`sensor_msgs/msg/Image`

Parameters

See also the common parameters.

Name	Type	Default (range)	Changed at run time	Description
`backend`	string	“auto”	yes	Backend of cv::dnn: auto, cpu, opencl, opencl_fp16, cuda or cuda_fp16. auto uses CUDA when OpenCV is built with CUDA and a CUDA device is found, and the CPU otherwise.
`binary_threshold`	double	0.3 (0 - 1)	yes	Threshold of the text probability of the pixels
`detection_model`	string	“”	yes	Path or URL of the text detection model of DB (onnx), empty for DB (ResNet-18, IC15) of opencv_zoo. The URL is downloaded to ~/.ros/opencv_apps/models/ at the first time
`input_height`	int	736 (32 - 4096)	yes	Height of the input of the detection model, a multiple of 32
`input_width`	int	736 (32 - 4096)	yes	Width of the input of the detection model, a multiple of 32
`max_candidates`	int	200 (1 - 10000)	yes	Maximum number of the text regions
`polygon_threshold`	double	0.5 (0 - 1)	yes	Threshold of the score of the text regions
`queue_size`	int	3	no, restart the node	Specify queue_size of input image subscribers
`recognition_model`	string	“”	yes	Path or URL of the text recognition model of CRNN (onnx), empty for CRNN (English) of opencv_zoo, or none to detect the text only
`unclip_ratio`	double	2.0 (0 - 10)	yes	Ratio to expand the detected text regions
`vocabulary`	string	“”	yes	Path or URL of the vocabulary of the recognition model, one character per line, or empty for 0-9 and a-z of CRNN (English)

Source

src/nodelet/dnn_text_detection_nodelet.cpp, launch/dnn_text_detection.launch.xml