Adding DEEP_CNN for SDNET

deep_CNN/data/README.md

@@ -0,0 +1,7 @@
+## Data set Description:
+
+SDNET2018 is an annotated image dataset for training, validation, and benchmarking of artificial intelligence based crack detection algorithms for concrete.
+SDNET2018 contains over 56,000 images of cracked and non-cracked concrete bridge decks, walls, and pavements. 
+
+Link to the Dataset: 
+[SDNET](https://digitalcommons.usu.edu/all_datasets/48/)

deep_CNN/exercise/README.md

@@ -0,0 +1,24 @@
+## Description
+This problem deals with detecting the faults (cracks) in the SDNET dataset having concrete bridge decks, walls, and pavements images. In the proposed approach, a deep convolution neural newtork (D-CNN) is developed
+for detect the faults from the Pavement images of the dataset. To deal with imablanced classes and reduce overfitting, data augmentation have also been applied before feature extraction. 
+
+
+## Prerequisite
+
+	- Computer vision concepts and knowledge of making data pipelines in Python
+
+	- Knowledge of Convolutional Neural Network (CNNs) and Hyper-parameter Optimization for better performance.
+
+	- Basic Knowledge of evaluating Deep learning models
+
+## Evaluation Metrics
+
+The proposed model is evaluated on the basis: Accuracy, Confusion Matrix, Sensitivity, Specifcity, Precision, F-Measure.
+
+## Future Work
+
+1.To solve the issue of imabalanced classes, advanced resampling technqiues (random or rule-based) could be used before feature extraction. 
+
+2.Transfer Learning methodologies can also be implemented for obtaining superior metrics of evaluation.
+
+3.For hyper-parametr optimization, meta-heuristic technqiues (for example: Genetic algorithms) can be implemented.