Data for: Detecting Explosive Trace Dataset using CNN
Description
The dataset was originally numerical dataset for Detecting Explosives by PGNAA using KNN. Due to the need to train the data for trace element detection using CNN, it was important to modify the data by converting them to 2D images. CNN performs better with images than numerical values. Out of 65, 000 data samples, 1000 points were used. Each data point had only two features, explosive and non-explosive.
Files
Steps to reproduce
1. Initial Data Transformation Data Reshaping: Two dummy features were added to the data to increase the data columns to 4 features. [■(■(■(explosive&non-explosive)&dummy_1)&dummy_2)] The new dataset was then rearranged to form a grid of 2 x 2 matrix [■(explosive&dummy_1@non-explosive&dummy_2)] 2. Normalization The dataset was scaled to range between 0 and 1 using min – max scalar: 3. Convert to 3D Grayscale Image Representation Reshape each 2×22 matrix into a shape of 2×2×1, where the final "1" represents a single grayscale channel. [■([explosive]&[dummy_1]@[non-explosive]&[dummy_2] )][2×2×1] 4. Prepare the Dataset for Batch Processing Next, the dataset was broken into batches before feeding into the CNN model. This was achieved by representing the data as: (1000,2,2,1). This ensures that each row of the data was represented as a separate grayscale image. Presented as: [■([explosive]&[dummy_1]@[non-explosive]&[dummy_2] )]