lp_recognition's Introduction

RILP

Robust Iranian License Plate Recognition Designed for Complex Conditions

This Repo is not maintaned anymore, there are issues with the code and the dataset is not available anymore for download. However, please adequately cite the paper if you compare to our method or your method is built using our code/method backbone.

This is a Modular Framework designed for License Plate Number Recognition in Complex Conditions. The distinct Design ables us to reconfigure the framework for other regions/Conditions in no time!

Prerequisites

The Following Setup is tested and working:

Python>=3.5
Pytorch>=0.4.1
Tensorflow>=1.12.2
Cuda>=9.0
opencv>=3.4.2

Testing

Place the images inside test_set/images directory
Delete all other images inside folders (don't delete the folders, just files inside them)
In main directory run: python3 runner.py

Training

The pre-trained model provided, we will not publish the training code
In order to train use pre-trained model or try another model

datasets

Glyphs

Properties:
- Volume: 5000 images
- Labeled
- Size: 100 x K (20 < K < 100)
Link:
- Will be avaiable soon, stay tuned.

Plates

Properties:
- Volume: 350 images
- Various conditions
- Various sizes
- License plate number (only one of them) is labeled
Link:
- It is only available for research purposes. Ask the first author a_samad[at]aut[dot]ac[dot]ir

Citing

Please adequately refer to the papers any time this Method/Code is being used. If you do publish a paper where this Work helped your research, Please cite the following papers in your publications.

@inproceedings{Samadzadeh2020RILP,
  title={RILP: Robust Iranian License Plate Recognition Designed for Complex Conditions},
  author={Ali Samadzadeh, Amir Mehdi Shayan, Bahman Rouhani, Ahmad Nickabadi, Mohammad Rahmati},
  year={2020},
  organization={IEEE}}

lp_recognition's People

Contributors

Stargazers

Watchers

lp_recognition's Issues

zero divisoin error

from below code:

how resolve it?

`def find_corners(poly):
if len(poly) <= 4:
return poly
box = []
box.append(poly[0])

if float(poly[0][1]) - float(poly[1][1]) == 0:
    prev_slope = (float(poly[1][1]) - float(poly[0][1])) * 1000
else:
    prev_slope = (float(poly[1][1]) - float(poly[0][1])) / (float(poly[1][0]) - float(poly[0][0]))

for i in range(1, len(poly)):
    if float(poly[i - 1][1]) - float(poly[i][1]) == 0:
        slope = (float(poly[i][1]) - float(poly[i - 1][1])) * 1000
    else:
        slope = (float(poly[i][1]) - float(poly[i - 1][1])) / (float(poly[i][0]) - float(poly[i - 1][0]))
    if abs(prev_slope - slope) > 2:
        box.append(poly[i - 1])
    prev_slope = slope

box.append(poly[-1])
return np.array(box)`

segmentation and ocr dosn't work

Hi,
i got error like below after process 5 images in module 2
RuntimeError: CUDA out of memory. Tried to allocate 12.50 MiB (GPU 0; 10.92 GiB total capacity; 8.57 MiB already allocated; 9.28 GiB free; 4.68 MiB cached) i know this may relay to my gpu version but on nvidia geforce gtx 1560 corei7 16 gig RAM
it just process 22 images and then error runtime show up again.
also that 5 image that have been processed doesn't have outputs for seg or char , etc ...

aa-samad / lp_recognition Goto Github PK