A novel convolution neural network model, abbreviated NL-CNN is proposed,\nwhere nonlinear convolution is emulated in a cascade of convolution +\nnonlinearity layers. The code for its implementation and some trained models\nare made publicly available. Performance evaluation for several widely known\ndatasets is provided, showing several relevant features: i) for small / medium\ninput image sizes the proposed network gives very good testing accuracy, given\na low implementation complexity and model size; ii) compares favorably with\nother widely known resources-constrained models, for instance in comparison to\nMobileNetv2 provides better accuracy with several times less training times and\nup to ten times less parameters (memory occupied by the model); iii) has a\nrelevant set of hyper-parameters which can be easily and rapidly tuned due to\nthe fast training specific to it. All these features make NL-CNN suitable for\nIoT, smart sensing, bio-medical portable instrumentation and other applications\nwhere artificial intelligence must be deployed in energy-constrained\nenvironments.\n