- Faulty Image Dataset

The faulty image dataset is an image library that contains samples of faulty images that a robot camera has recorded as a result of various faults it may encounter. Possible camera malfunctions, such as the camera recording the faulty images at that time and the system continuing to work on these faulty images, can cause serious problems in the system. CamFITool is designed to simulate such failures, collect faulty picture outputs of the systems as a result of these failures, and create faulty output source for artificial intelligence systems to be developed to detect these problems. CamFITool allows us to create a dataset of 5000 normal and 5000 fault injected images (see Table 1) from the SRVT system. In addition, users can create such datasets in desired configurations and camera-based perception-based environments by using CamFITool.

RGB Camera Fault Types: 

Dilation: There are 476 of these faults injected pictures in the dataset.

Erosion: There are 637 of these faults injected images in the dataset

Open: There are 640 of these faults injected images in the dataset.

Close: There are 841 of these faults injected images in the dataset.

Gradient: There are 632 of these faults injected images in the dataset.

Motion-blur: There are 687 of these faults injected images in the dataset.

Partialloss: There are 1087 of these faults injected pictures in the dataset. 

https://arxiv.org/abs/2108.13803

https://drive.google.com/drive/folders/1pIqEnIIKFN13Z4m38zEB-XYQp_qb0T6S?usp=sharing.