This CLAMS app is a text localization app based on OpenCV and EAST text detection model.
General user instruction for CLAMS apps is available at CLAMS Apps documentation.
This is an OpenCV-based text localization app that used EAST text detection model. Text localization is a technique to detect the location of "scene text" in an image or video. "
The app implementation is based on EAST algorithm, and the frozen EAST model is downloaded from this tutorial: https://learnopencv.com/deep-learning-based-text-detection-using-opencv-c-python/ .
This app can process both images and videos and when analyzing videos, it can run in two modes: segment-based mode and resampling mode (these modes are irrelevant when analyzing images).
When the app finds existing TimeFrame annotation (Please visit https://mmif.clams.ai/ for details about CLAMS annotation types) in the input MMIF, it takes them as segmentation and runs in segment-based mode. In segment-based mode the app extracts 2 frames (the first and the last) from one segment, then perform text localization on these two frames.
frameType: the type of segment to use (default is empty value ("") and it means using all types of segments)
When there's no TimeFrame annotation to use, the app automatically runs in resampling mode. In resampling mode, the app extracts frames from the input video at a fixed interval and perform text localization on these frames.
sampleRatio: the sampling interval in frames numbers (default is 30, which means extracting one frame every 30 frames)
For the full list of parameters, please refer to the app metadata from CLAMS App Directory or metadata.py file in this repository.