ivanaslamov / hog_sse

Implementation of HoG feature extractor that uses SSE instructions.
MIT License
45 stars 27 forks source link

Difference Between Other HOGs #3

Open ghost opened 7 years ago

ghost commented 7 years ago

Hi Ivan, Your implementation outputs 16 dimensions at total, where as https://people.eecs.berkeley.edu/~rbg/latent/ outputs 31 dimensions and first 18 is contrast sensitive features, 9 contrast insensitive and 4 texture features. Can you categorize the 16 channels like this. Thanks in advance

ivanaslamov commented 7 years ago

Hi @sinanariyurek, I will try to look into it.