This article provides the download to the best working version of the bot and the 374 MB datasets for both 28x28 and 96x192 sample sizes. This version run the feed-forward process of the CNN in Python using TensorFlow Keras which means it can take full advantage of GPU acceleration. The C program uses a python daemon to communicate with the TensorFlow Keras code.

This version trains a CNN using TensorFlow Keras and then exports the weights out to a C program which runs the CNN model on the CPU.

This version of the bot trains a FNN using TensorFlow Keras and then exports the weights out to a C program which runs the FNN model on the CPU.

This version of the bot trains a CNN network made in C based on a cut-down version of the VGG network. The feed-forward process and training is all done in C. In this part the network is trained to detect player heads using offline training.

This version of the bot trains a CNN network made in C based on a cut-down version of the VGG network. The feed-forward process and training is all done in C. In this part the network is trained to detect a football using online training.

This part describes my brief history in making colour aimbots for CSS and CS 1.6 as an adolescent and attempts to use TFCNNv2 to make an autoshoot bot for CS:GO but with dismal results.

This part describes making a very simple neural network that outperforms regular colour trigger bots in the game of Quake3.