crack-detector

A nice crack detector :)

How to run:

WebApp

• python predict.py --meta_file="./alg1_output/model_complete.meta" --CP_dir="./alg1_output" --start_as_server=True
• python predict.py --meta_file="./alg2_output/model_complete.meta" --CP_dir="./alg2_output" --start_as_server=True --port=8081 --model_number=2
• open WebApp/index.html in a browser disabling CORS, here is how to do it for chrome: https://alfilatov.com/posts/run-chrome-without-cors/

Project 1:

Based on https://www.researchgate.net/publication/315613676_Deep_Learning-Based_Crack_Damage_Detection_Using_Convolutional_Neural_Networks?fbclid=IwAR1NqL8c7qwKNCFxAW7E9BAW6c98DKCSEfdgoYAB0WYY5iaQVSTqxBMVqCY

Requirements:

• tensorflow: 1.15.0
• matplotlib: 3.1.2
• opencv-python: 4.1.2.30

How to run:

• Change directory to CracksDetectionApp (cd CracksDetectionApp)
• python trainAlg1.py
• python trainAlg2.py

How to test:

• Change directory to CracksDetectionApp (cd CracksDetectionApp)
• python predict.py --meta_file="./alg1_output/model_complete.meta" --CP_dir="./alg1_output"
• look into folder CracksDetectionApp/results for the resulting images

In order to test the other algorithm run:

• python predict.py --meta_file="./alg2_output/model_complete.meta" --CP_dir="./alg2_output"

To start the predictor as servers for each algorithm run the following commands:

• python predict.py --meta_file="./alg1_output/model_complete.meta" --CP_dir="./alg1_output" --start_as_server=True
• python predict.py --meta_file="./alg2_output/model_complete.meta" --CP_dir="./alg2_output" --start_as_server=True --port=8081 --model_number=2

Project 2 Unet:

Based on https://towardsdatascience.com/understanding-semantic-segmentation-with-unet-6be4f42d4b47

Requirements:

• numpy 1.17.4
• pandas 0.25.3
• matplotlib 3.1.2
• Pillow 6.2.1
• tqdm 4.40.0
• scikit-image 0.16.2
• tensorflow 2.0.0

How to run:

• cd CommandLineApp/CrackDetection/
• python ImageProcessorUsingUNET.py

Project 3 V2:

Requirements:

• tensorflow 2.0.0
• matplotlib 3.1.2
• aspectlib 1.4.2

How to run:

• python CommandLineApp/CrackDetection/ImageProcessorV2.py

How to test:

• python CommandLineApp/CrackDetection/predictionV2.py

technical report link : https://docs.google.com/document/d/1XFI4-NG3JoQ1_NO4GCPsOfUy0MokdtFchDorpWL-vLI/edit?fbclid=IwAR0nZoZapOnE3ei6PkDeUQHi26dwDNDQne-qzYXeWncD7e6qYNYw2_YZ4h8

google docs link : https://docs.google.com/document/d/1HZ3XukJdGuqtobCRUvc7KHZCh3Oun3Pu6DqvPCUoquE/edit?usp=sharing

presentation link : https://docs.google.com/presentation/d/1DjbLFpMnE6zBXJQ2NxbrZQuKOJR_Lwlu5tKPKM8E8As/edit?fbclid=IwAR3YqGYsOe-zBtoFLWWVQR8RJFR6HVrujhF05Yt-fmvpuGBRf9jjZQ8cPVc#slide=id.p2

trello link : https://trello.com/b/60KTWa0j/crack-detector

Coordinators: Associate Professor, PhD Adrian Iftene PhD Anca Ignat

Members:

• Razvan Nica (contact: razvan.nica30@gmail.com)
• Serban Mihai Botez (contact: serbanbotez2@gmail.com)
• Stefan Cosmin Romanescu (contact: stefancosmin.romanescu@gmail.com)
• Florin Cristian Finaru (contact: florincristian.finaru@gmail.com)