Harald Scheidl

SimpleHTR. Handwritten Text Recognition (HTR) system implemented with TensorFlow.

CTCDecoder. Connectionist Temporal Classification (CTC) decoding algorithms: best path, beam search, lexicon search, prefix search, and token passing. Implemented in Python.

CTCWordBeamSearch. Connectionist Temporal Classification (CTC) decoder with dictionary and language model.

WordDetector. Detect handwritten words (classic image processing based method).

DeslantImg. The deslanting algorithm sets text upright in images. Python, C++ and OpenCL implementations provided.

HTRPipeline. Detect and read handwritten words on scanned pages.

WordDetectorNN. Detect handwritten words (neural network based).

analyze_ada_hessian. Analyze AdaHessian optimizer on 2D functions.

GPUImageProcessingArticle. Python

cpp_mnist. A neural network written in ~260 lines of plain C++ code without any dependencies.

ParticleSystem. A simple particle system implemented with Python and OpenCL.

fast_inv_sqrt. Fast approximation of the inverse square root 1/√(x).

quasi_cauchy_optimizer. Implementation of the quasi Cauchy optimizer, an optimization method from the quasi Newton family. It uses a diagonal approximation of the Hessian and therefore has a small memory footprint.

separate_kernel. Separate a 2D convolution kernel into two 1D kernels.

githubharald.github.io. HTML