https://github.com/blurred-machine/computer-vision-problems

In this repository, I have solved two problems with computer vision. The first Implementation is building a CNN in which the input is a noisy number and the output is a denoised number based on AutoEncoders. The second implementations is sequencing the characters read from number plates done using object detection.
https://github.com/blurred-machine/computer-vision-problems

cnn-classification cnn-keras computer-vision deeplearning detection image-processing neural-network

Last synced: over 1 year ago
JSON representation

In this repository, I have solved two problems with computer vision. The first Implementation is building a CNN in which the input is a noisy number and the output is a denoised number based on AutoEncoders. The second implementations is sequencing the characters read from number plates done using object detection.

• Host: GitHub
• URL: https://github.com/blurred-machine/computer-vision-problems
• Owner: blurred-machine
• Created: 2020-05-28T16:57:28.000Z (about 6 years ago)
• Default Branch: master
• Last Pushed: 2020-07-02T15:02:36.000Z (almost 6 years ago)
• Last Synced: 2025-01-09T05:18:19.286Z (over 1 year ago)
• Topics: cnn-classification, cnn-keras, computer-vision, deeplearning, detection, image-processing, neural-network
• Language: Jupyter Notebook
• Homepage:
• Size: 683 KB
• Stars: 2
• Watchers: 1
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
# Computer-Vision-Problems

In this repository, I have solved two problems with computer vision. The first Implementation is building a CNN in which the input is a noisy number and the output is a denoised number based on AutoEncoders. The second implementations is sequencing the characters read from number plates done using object detection.

* Please use the jupyter notebook to run the above files in iPython environment.

* or directly run the `Problem_x.py` file in the python shell.

## Implementation P1

- Loss Calculation during training:

- ![Loss Calculation](https://github.com/blurred-machine/Computer-Vision-Problems/blob/master/screenshots/training%20loss.PNG)

- Loss Calculation during training:

- ![Accuracy calculation](https://github.com/blurred-machine/Computer-Vision-Problems/blob/master/screenshots/training%20accuracy.PNG)

- Final Reult with accuracy of **81.27%** on Training and **81.35%** on Validation Set:

- ![Result](https://github.com/blurred-machine/Computer-Vision-Problems/blob/master/screenshots/final%20result.PNG)

## Implementation P2

Steps followed:

- Extracted the characters from the XML result of object detection using `xml.etree.ElementTree`.

- Wrote functions to extract **character** and **boundung boxes** data from XML root.

- Built a sorting sequence algorithm for characters based on bounding box coordinates.

- Tested and verified the algorithm on single line and double line number plates.

Thank You!