Used several semantic segmentation models (i.e., UNet++, FPN, DeepLabV3+) with different CNN encoders, pre-trained with 12M ImageNet dataset, to detect cracks in built environment images (e.g., bridges, infrastructures, pavement, etc.) with quite favorable results (See Figure [1]).
Fig [1] : A few sample inference results of the Test set images.
- Please refer to the file
requirements.txt
for a comprehensive list of packages and their corresponding version.
.
.
├── data
│ ├── testcrop
│ ├── testdata
│ └── traincrop
├── images
├── logs
├── models
├── outputs
│ ├── history
│ ├── hyperparams
│ ├── hyperparams_search
│ ├── Inferences
│ └── plots
├── reports
├── runs
├── temp
└── utils
└── models
58 directories
Should you have any questions, feel free to contact TekBoArt @tekboart.
- Refer to the file
LICENSE
for more information regarding the license of this repository.
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.