afcarl / dimensionalityestimation

Devoted to experiments on using the Kmeans++ to estimate dimension

DimensionalityEstimation

Devoted to experiments on using the Kmeans++ to estimate dimension

src: Source Directory

• get_read_videos Code to download videos from a youtube list and to then load each frame into a numpy array using open-cv
• FaceDetection Code to transform a video of an announcer (here colbert) into a sequence of np.arrays containing 300x300 BW/windows
  • detect face locations in using boosted haar (standard in open-cv)
  • Form tracks here the face is moving smoothly from frame to frame
  • estimate the skin color and trim the non-skin parts, translate the image into BW
• kmeans++
• Spark-KMeans++