CNN v1.2

The full release of the CNN

Notes

• The convolution layer is now taking place on the GPU, that is normalization, convolution and max_pooling
• The forward propagation is now running on the GPU as well, this however is only implemented for the validation function
• The backward propagation is still running sequentially on the CPU.

How to run the program

• The CNN code is in the "Image_classifer" folder
• In this folder please run the executables: "install_dependencies.sh" and "download_dataset.sh"
• After successfully running the scripts mentioned above please proceed to make a directory named "build" (or another name to your preference) in the root of the "Image_classifier" directory.
• Navigate into the newly created directory
• Run "cmake ..", verify successful completion
• Run "make -j8", verify successful completion
• If the above steps have been successfully executed you can now run the newly generated executable named "image_classifier"
• The GUI and terminal output should be quite intuitive

CNN v1.1

The full release of the CNN

Notes

• The convolution layer is now taking place on the GPU, that is normalization, convolution and max_pooling
• The forward propagation is now running on the GPU as well, this however is only implemented for the validation function
• The backward propagation is still running sequentially on the CPU.

How to run the program

• The CNN code is in the "Image_classifer" folder
• In this folder please run the executables: "install_dependencies.sh" and "download_dataset.sh"
• After successfully running the scripts mentioned above please proceed to make a directory named "build" (or another name to your preference) in the root of the "Image_classifier" directory.
• Navigate into the newly created directory
• Run "cmake ..", verify successful completion
• Run "make -j8", verify successful completion
• If the above steps have been successfully executed you can now run the newly generated executable named "image_classifier"
• The GUI and terminal output should be quite intuitive

CNN v1.0

The first full release of the CNN

Notes

• The convolution layer is now taking place on the GPU, that is normalization, convolution and max_pooling
• The forward propagation is now running on the GPU as well, this however is only implemented for the validation function
• The backward propagation is still running sequentially on the CPU.

How to run the program

• The CNN code is in the "Image_classifer" folder
• In this folder please run the executables: "install_dependencies.sh" and "download_dataset.sh"
• After successfully running the scripts mentioned above please proceed to make a directory named "build" (or another name to your preference) in the root of the "Image_classifier" directory.
• Navigate into the newly created directory
• Run "cmake ..", verify successful completion
• Run "make -j8", verify successful completion
• If the above steps have been successfully executed you can now run the newly generated executable named "image_classifier"
• The GUI and terminal output should be quite intuitive