System and process for reducing time of transmission for single-band, multiple-band or hyperspectral imagery using machine learning based compression

A machine learning-based media compression and decompression process comprising:

• pre-processing an original media file with a hyperspectral image by a pre-processing module, wherein the hyperspectral image comprises a plurality of channels of any data type and arbitrary bit depth, wherein the pre-processed media file comprises a matrix of floating point values, wherein pre-processing the original media file comprises (i) determining, for each channel in the plurality of channels, spatial dimensionality of the hyperspectral image in the original media file, (ii) adjusting dimensions of an image plane of the hyperspectral image, for each channel in the plurality of channels of the hyperspectral image, in the original media file, and (iii) normalizing the adjusted dimensions of the image plane of the hyperspectral image, for each channel in the plurality of channels of the hyperspectral image, to the matrix of floating point values, wherein determining spatial dimensionality of the hyperspectral image comprises determining dimensions of the image plane of the hyperspectral image in the original media file, wherein the dimensions of the image plane comprise a height dimension and a width dimension of the image plane for each channel in the plurality of channels of the hyperspectral image, wherein adjusting dimensions of the image plane of the hyperspectral image for each channel in the plurality of channels of the hyperspectral image in the original media file comprises (i) determining a height value measured in number of pixels of the height dimension and a width value measured in number of pixels of the width dimension of the image plane for each channel in the plurality of channels of the hyperspectral image, (ii) dividing, for each channel in the plurality of channels of the hyperspectral image, the height value and the width value by a positive integer value to obtain a resultant height value and a resultant width value for each channel in the plurality of channels of the hyperspectral image, and (iii) adjusting, for each channel in the plurality of channels of the hyperspectral image, each of the resultant height value and the resultant width value to a multiple of thirty-two, wherein each pixel comprises a measurement of spectral information across a plurality of continuous spectral bands;
• reducing spatial dimensionality of each channel in the plurality of channels of the hyperspectral image in the pre-processed media file by an encoder module passing the pre-processed media file through an encoder neural network and outputting an encoded media file with a reduced spatial dimensionality image;
• creating a bit string by passing the encoded media file with the reduced spatial dimensionality image through an entropy module, wherein the bit string is created as a compressed version of the original media file;
• using the bit string created by the entropy module as a compressed media file that is configured for transmission over a data network and reproduction as the original media file, wherein a time of transmission of the compressed media file over the data network is reduced from the time of transmission of the original media file over the data network;
• and
• transferring, over the data network, the compressed media file to a destination, wherein the destination comprises one of a destination storage location and a destination computing device connected to the data network, wherein the compressed media file is configured for reproduction of the original media file at the destination.