Betect Train

A tool for training neural network patches (i.e. the lower layers of a VGG16) for your own targeted object classification. These neural patches can be applied as plugins to other tools, such as iOS and Android mobile apps (Betect) for live image detection of targeted categories.

Simply place training images in a folder, with images of each category in a subfolder with a representative name. Then set batch size and error rate and hit run. Alternatively, use "map" mode where each training image file's name is prepended with its category tag, and a map file is proved (similar to imagenet data). See here for more detailed usage guide: http://www.advancedkernels.com/betect/theory/