- Powerful tool to recognise license plates of vehicles from a video and convert them to text data for storage/data analysis
- Uses AWS Services such as Textract and Rekognition to efficiently extract text from the video
Keywords: AWS API, EC2 instance, Cloud, Textract, Rekognition, python, OpenCV, Ubuntu, boto3, numpy, License Plate Recognition, Alpr, Image analysis, Real life image text recognition
- Extract frames from the video as images (OpenCV)
- Pass the frames to AWS Textract or AWS Rekognition service to analyse and get detected text in the frames
- Output detected text data to file/command prompt with timestamp
- OpenCV with Python3
- AWS Ubuntu Linux VM EC2 instance with SSH Access (Free Tier works for testing/limited use)
| Sl NO | FileName | Functionality | Requirements |
|---|---|---|---|
| 1 | VideoParserFPS.py |
Parses a local video file into frames at a specified frame rate and displays the parsed frames | 1. OpenCV with Python (Python3 for uniformity across the project) 2. numpy python module |
| 2 | TextractLocal.py |
Processes a local image file (frame) using AWS Textract API and returns the result | 1. AWS Ubuntu Linux EC2 Instance over SSH with Textract Security policies applied to IAM role of the instance 2. boto3 python module |
| 3 | RekognitionLocal.py |
Processes a local image file (frame) using AWS Rekognition API and returns the result. Claimed to produce better results than Textract on real life images. | 1. AWS Ubuntu Linux EC2 Instance over SSH with Rekognition Security policies applied to IAM role of the instance 2. boto3 python module |
| 4 | TextractParserCumDetector.py |
Processes a local video file at a specified frame rate using Textract API and returns the extracted results with timestamp | 1. OpenCV with Python (Python3 for uniformity across the project), 2. AWS Ubuntu Linux EC2 Instance over SSH with Textract Security policies applied to IAM role of the instance 3. boto3 and numpy python modules |
| 5 | RekognitionParserCumDetector.py |
Processes a local video file at a specified frame rate using Rekognition API and returns the extracted results with timestamp. Claimed to produce better results than Textract on real life images. | 1. OpenCV with Python (Python3 for uniformity across the project), 2. AWS Ubuntu Linux EC2 Instance over SSH with Rekognition Security policies applied to IAM role of the instance 3. boto3 and numpy python modules |
Files above are not dependent on each other.
- Use
VideoParserFPS.pyto test parsing - Use
TextractLocal.pyandRekognitionLocal.pyto test Text Detection on a frame - Use
TextractParserCumDetector.pyandRekognitionParserCumDetector.pyfor combined full functionality
Run the files in Ubuntu EC2 instance. Make sure to assign an IAM Role to the instance with Textract/Rekognition Policies attached. API calls to Textract/Rekognition will not work without these.
python3 VideoParserFPS.pypython3 TextractLocal.py
and so on...
- Run the desired file from the same directory where
frame.jpgorvideo.mp4is present. Frame name/Video Name hard coded - Set aws-region in files 2, 3, 4, 5
sudo apt update
sudo apt install python3-opencv
You may want to compile OpenCV from source if above method doesn't work.
sudo apt update
sudo apt install python3-pip
sudo pip3 install numpy
sudo pip3 install boto3
- Use
python3 TextractParserCumDetector.py >> DataDump.csvto redirect output to csv file for further analysis - Post processing of data and cleaning to make it more practically useful
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent