Authoring

Parametric Spectral Filters for Fast Converging, Scalable Convolutional Neural Networks

Luke Wood, Eric C. Larson

Primary author of 2021 ICASSP Publication: Parametric Spectral Filters for Fast Converging, Scalable Convolutional Neural Networks. Please see GitHub repo for full information.

Scaling Continuous Kernels with Sparse Fourier Domain Learning

Clay Harper, Luke Wood, Eric C. Larson, Peter Gerstoft

We address three key challenges in learning continuous kernel representations: computational efficiency, parameter efficiency, and spectral bias. Continuous kernels have shown significant potential, but their practical adoption is often limited by high computational and memory demands. Additionally, these methods are prone to spectral bias, which impedes their ability to capture high-frequency details. To overcome these limitations, we propose a novel approach that leverages sparse learning in the Fourier domain. Our method enables the efficient scaling of continuous kernels, drastically reduces computational and memory requirements, and mitigates spectral bias by exploiting the Gibbs phenomenon.

Efficient Graph-Friendly COCO Metric Computation for Train-Time Model Evaluation

Pre-print written alongside Francois Chollet on a novel algorithm to closely approximate Mean Average Precision within the constraints of the TensorFlow graph. The algorithm used in the publication is used in the KerasCV COCO metric implementation, and can be used to perform train time evaluation with any KerasCV object detection model.

Deep Learning Object Detection Approaches to Signal Identification

Pre-print written for ICASSP 2022. We decided not to continue this line of work, but our results and our spectrogram object detection dataset are open source and available on GitHub. If you'd like to finish this work and attempt to get it published, feel free to reach out.

The Definitive Guide to Object Detection

Luke Wood

My "definitive guide" to object detection is live on keras.io! This tutorial is a bit more like a textbook chapter than a typical keras.io tutorial, but by the end of it you will have an extremely strong sense of how to tackle object detection problems with deep learning.

The Definitive Guide to Image Classification

Luke Wood

My "definitive guide" to image classification is live on keras.io! This tutorial is a bit more like a textbook chapter than a typical keras.io tutorial, but by the end of it you will have an extremely strong sense of how to tackle classification problems with deep learning.

Teach StableDiffusion new concepts via Textual Inversion

Learning new visual concepts with KerasCV's StableDiffusion implementation.

High-performance image generation using Stable Diffusion in KerasCV

Francois Chollet, Luke Wood, Divam Gupta

Generate new images using KerasCV's StableDiffusion model.

A walk through latent space with Stable Diffusion

Ian Stenbit, Francois Chollet, Luke Wood

Explore the latent manifold of Stable Diffusion.

Custom Image Augmentations with BaseImageAugmentationLayer

Luke Wood

Use BaseImageAugmentationLayer to implement custom data augmentations.

CutMix, MixUp, and RandAugment image augmentation with KerasCV

Luke Wood

Use KerasCV to augment images with CutMix, MixUp, RandAugment, and more.

Evaluating and exporting scikit-learn metrics in a Keras callback

Luke Wood

This example shows how to use Keras callbacks to evaluate and export non-TensorFlow based metrics.

Customizing the convolution operation of a Conv2D layer

Luke Wood

This example shows how to implement custom convolution layers using the Conv.convolution_op() API.

Writing Keras Models With TensorFlow NumPy

Luke Wood

Overview of how to use the TensorFlow NumPy API to write Keras models. Published on keras.io