pathbreak/deepvisualminer-haswell repository overview

This is an optimized version of https://hub.docker.com/r/pathbreak/deepvisualminer/⁠ . The main optimization is that it uses a custom optimized build of TensorFlow.

Image based on Ubuntu 16.04

• python3.5
• Darkflow (https://github.com/thtrieu/darkflow⁠)
• TensorFlow 1.0 custom build optimized for Haswell architecture instruction sets (SSEx, AVX, AVX2, FMA)
• FFMpeg
• OpenCV 3.2 (opencv-python and opencv-contrib-python from PyPI. These don't support videos because they aren't linked with ffmpeg)
• imageio (python module for video IO using ffmpeg)
• scikit-learn (for machine learning helpers)
• numpy, scipy (required by all the other frameworks above)

See my Github repo⁠ for usage.