TQVCD

KO | EN

gitlite — search

#hacktoberfest #react #ai #typescript #llm #go #golang #android #docker #nodejs #windows #awesome #kubernetes #ios

(★ 11)

Automated vehicle damage classification using the three-quarter view car damage dataset and deep learning approaches

TQVCD 최신버젼 다운로드

최종 버전 다운로드 (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# Automated vehicle damage classification using the three-quarter view car damage dataset and deep learning approaches

This repository is to supplement the paper "Automated vehicle damage classification using the three-quarter view car damage dataset (TQVCD) and deep learning approaches".

# Contribution
1. We introduce the Three-Quarter View Car Damage dataset (TQVCD dataset). The main objective of the dataset is to efficiently acquire comprehensive information on damage without explicitly labeling every damaged part. Unlike other datasets, the TQVCD dataset is constructed as a reference by mixing vehicle orientation and damage type.
2. Our research is public data and is expected to be the basis for future research related to the automobile industry. Most studies have different standards for each study using their own datasets, and few are car-related datasets.
3. To validate our dataset, we leverage transfer learning with diverse pre-trained deep learning-based models to build a binary classifier for each type of damage. To improve the robustness of classification results, we implemented a model ensemble method to reduce the deviation of individual model dependencies effectively. Our experiments included various weight values for each single model to optimize classifier performance.
4. In recognition of the reduced sensitivity to lighting and background noise, we incorporated grayscale datasets into our analysis. The final results were derived by combining prediction values obtained from RGB datasets and computing the elementwise average.    
5. Since it is an industrial issue, the effectiveness of this study was verified through interviews with industry experts, and the effectiveness of the paper was reviewed.

# Dataset
We collected the car damage images corresponding to 3/4 views from the "vehicle damage data" of AI HUB and constructed the data by dividing the damage type into breakage and crushed.

- Dataset Configuration

| Class                | Quantity |
|----------------------|----------|
| Front Normal (FN)    | 500      |
| Front Breakage (FB)  | 500      |
| Front Crushed (FC)   | 400      |
| Rear Normal (RN)     | 300      |
| Rear Breakage (RB)   | 300      |
| Rear Crushed (RC)    | 300      |
|       Total          | 2,300    |

** The data that support the findings of this study are available on request from the corresponding author.

TQVCD

TQVCD 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

TQVCD

TQVCD 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

Download README