In this video, you’ll learn how to leverage Python Tensorflow Object Detection to be able to detect license plates using Kaggle Data. Once those plates have been detected you’ll then be able to apply OCR to extract the text from each and every plate using PyTorch and EachOCR.
On top of it all, you’ll be able to export your results. We’ll setup a saving function to output the regions of interest as well as the detected text. This could be used as part of a broader system or standalone.
In this video you’ll learn how to:
Chapters:
0:00 - Start
0:26 - Tutorial Start
1:11 - Gameplan
3:58 - PART 1 | Setup
7:13 - Cloning Baseline Code
10:42 - Creating a Virtual Environment
12:51 - Installing Dependencies
17:55 - Installing Tensorflow Object Detection
32:47 - Cloning Pre-Trained Models
34:21 - PART 2 | Data
34:44 - Cloning Images from Kaggle
35:52 - Creating a Training and Testing Partition
41:33 - PART 3 | Training
42:59 - Updating the LabelMap
44:37 - Creating TF Records
49:30 - Updating Transfer Learning Config
51:51 - Training the Model
1:00:00 - PART 4 | Detecting Plates
1:01:00 - Detecting Plates from an Image
1:03:45 - Detecting Plates from Video
1:05:01 - PART 5 | Applying OCR
1:06:29 - Splitting GPU
1:10:46 - Setup EasyOCR
1:13:46 - Applying Detection Thresholding
1:18:42 - Extract Image Width and Height
1:20:16 - Loop Through Detections and Apply OCR
1:25:40 - Filtering Algorithm
1:36:28 - Final OCR Function
1:42:15 - Applying ANPR in Real Time
1:45:38 - PART 6 | Saving Results
1:46:10 - Importing Dependencies
1:46:29 - Building a Save Function
1:52:06 - Saving Plates from. Video
Links
Final Notebook: https://github.com/nicknochnack/RealT…
Baseline Code: https://github.com/nicknochnack/TFODC…
PyTorch: https://pytorch.org/
Full Tutorial: https://youtu.be/yqkISICHH-U
Subscribe: https://www.youtube.com/c/NicholasRenotte/featured
#python #tensorflow