Deep Dive into Deep Learning with KotlinDL

This talk introduces KotlinDL, a Keras-like machine learning framework for Kotlin. It covers constructing and training complex neural networks on the JVM, including support for transfer learning with popular Keras models and image preprocessing. The framework's capabilities extend to ONNX module support, multiplatform compatibility, and running on Android devices. The presentation includes demonstrations of building, training, and using convolutional neural networks (CNNs) with KotlinDL, as well as deploying KotlinDL models in production. It also touches on concepts like transfer learning and custom model training.