iamaaditya / VQA_Demo

Visual Question Answering Demo on pretrained model
http://iamaaditya.github.io/2016/04/visual_question_answering_demo_notebook
MIT License
242 stars 133 forks source link

include mean pixel normalization #5

Closed mynameisvinn closed 8 years ago

mynameisvinn commented 8 years ago

ive include image normalization in get_image_features(). mean pixel values of [103.939, 116.779, 123.68] are taken from the VGG authors, and represent values computed from the training dataset.

image normalization improves feature extraction and accuracy!