Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/ayberkgezer/car-damage-detection
https://github.com/ayberkgezer/car-damage-detection
cnn-classification deep-neural-networks image-classification image-processing python vgg16
Last synced: about 15 hours ago
JSON representation
- Host: GitHub
- URL: https://github.com/ayberkgezer/car-damage-detection
- Owner: ayberkgezer
- License: mit
- Created: 2022-12-18T16:14:45.000Z (about 2 years ago)
- Default Branch: main
- Last Pushed: 2024-08-10T09:27:26.000Z (4 months ago)
- Last Synced: 2024-08-10T10:36:00.156Z (4 months ago)
- Topics: cnn-classification, deep-neural-networks, image-classification, image-processing, python, vgg16
- Language: Jupyter Notebook
- Homepage:
- Size: 354 KB
- Stars: 1
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: license
Awesome Lists containing this project
README
# Car Damage Detection on VGG-16
[](https://dergipark.org.tr/en/pub/estudambilisim/issue/86200/1421332)
For this we used a CNN model, VGG16[[1]](https://arxiv.org/abs/1409.1556). We used our own dataset that we had prepared and trained the VGG16 model ourselves. Our study consisted of 4 main questions:
- Is it a car?
- Is there any damage on the car?
- In which part of the car is the damage?
- What is the level of damage?
We prepared and trained our own datasets within these 4 questions.
## Data Sets
| Data Sets | Training | Validation |
| :-------- | :------- | :--------- |
| Is it a car? | `920` | `230` |
| Is there any damage on the car? | `1840` | `460` |
| In which any part of the car is the damage? | `976` | `171` |
| What is the level of damage? | `979` | `171` |
We used our original dataset, which consists of a total of 5,757 photographs.
## Is it a car?
| Data Sets | Training | Validation |
| :-------- | :------- | :--------- |
| Car | `920` | `230` |
The result we will get here is only querying whether there is a car or not.
## Is there any damage on the car?
| Data Sets | Training | Validation |
| :-------- | :------- | :--------- |
| Damaged | `920` | `230` |
| Undamaged | `920` | `230` |
We have done this training in order to determine whether there is a similar damage at this stage and to continue with the other stages according to the result.
## In which any part of the car is the damage?
| Data Sets | Training | Validation |
| :-------- | :------- | :--------- |
| Front | `418` | `73` |
| Rear | `287` | `50` |
| Side | `271` | `48` |
our aim here is to determine which part of the car the damaged area in the photo belongs to. We trained our parameters as front, back and side.
## What is the level of damage?
| Data Sets | Training | Validation |
| :-------- | :------- | :--------- |
| Minor | `278` | `48` |
| Moderate | `315` | `55` |
| Severe | `386` | `68` |
Now that we know the damage is and where it is, it's up to us to separate the level of damage we set ourselves.
## Accuracy Rating
| Training | Accuracy Rating |
| :-------- | :------- |
| Is it a car? | %98 |
| Is there any damage on the car? | %90 |
| In which any part of the car is the damage? | %70 |
| What is the level of damage? | %66 |
## Result Exp
## License
[](https://choosealicense.com/licenses/mit/)
## Authors
- [@ayberkgezer](https://www.github.com/octokatherine)
- [@TediTae](https://github.com/TediTae/)