Lars Nieradzik

kmeans-anchor-boxes. k-means clustering with the Intersection over Union (IoU) metric as described in the YOLO9000 paper

swift-f0. Fast and accurate fundamental frequency (F0) detector using convolutional neural networks

object-localization. Object localization in images using simple CNNs and Keras

chinese-subtitle-ocr. Optical character recognition for Chinese subtitles using SSD and CNN

pitch-benchmark. Comprehensive benchmark suite comparing pitch detection algorithms across multiple datasets.

forced-alignment-chinese. Mandarin Chinese audio datasets aligned with Montreal Forced Aligner

fastspeech2-clean. Clean and modernized implementation of FastSpeech2/LightSpeech using IPA

bigvgan-mirror. A mirror of BigVGAN and HiFi-GAN for access via PyTorch Hub.

story-evaluation-llm. LLM-generated story dataset with quality evaluations across 15 models for training and benchmarking creative writing capabilities.

segmentation_activations. Code for the paper "Effect of the output activation function on the probabilities and errors in medical image segmentation"

pysais-utf8. Python C module for creating suffix, LCP and BWT arrays with UTF-8 text.