Selective Text Removal for Videos using Computer Vision

Selectively censors all text appearing in a video that matches a vocabulary of filter texts.
Uses FAST text segmenter and PARSeq text recognition model for text detection in frames.
Users can set up sensitive words, such as names, (using the sensitive.csv) to censor in addition to Surge AI profanity text dataset.
OpenCV is used to capture frames and apply a blur filter to flagged text regions.

Example:

Filter words:

Visualization:

Output:

Name		Name	Last commit message	Last commit date
Latest commit History 8 Commits
.vs		.vs
.vscode		.vscode
FAST		FAST
__pycache__		__pycache__
configs		configs
output		output
requirements		requirements
strhub		strhub
tools		tools
README.md		README.md
TEST_OUTPUT.log.txt		TEST_OUTPUT.log.txt
environment.yaml		environment.yaml
eval_fast.py		eval_fast.py
main.py		main.py
output.json		output.json
profanity.csv		profanity.csv
reparameter.py		reparameter.py
requirements.txt		requirements.txt
requirements.yaml		requirements.yaml
sensitive.csv		sensitive.csv

Provide feedback