Work with VOC files to support bounding boxes for tensorflow -> annotation concatenation and json<->XML transcoding
This is the support page for "MetaMerge- supporting the transmogrification of Labelled Images for use in TensorBox and other Entity Detection Systems.
I recognized
Found some problems?
Post the problem to our Github issues page https://github.com/dhushon/MetaMerge-support/issues
Thank you.
Currently we support: RectLabel versions of XML & JSON annotation files [inout] - (https://rectlabel.com) TensorBox JSON annotation files [inout] - (https://github.com/Russell91/TensorBox)
MetaMerge has been built in Swift 4.0 using the new Encoder/Decoder -> Codable styles
- customization has been provided to take XML -> a Dictionary that can then be used by the Decoders, however there is not yet a custom encoder [TODO]
- ability to concatenate / single-ify annotations depending on the target engine
- deal with ClassID's and UID's associated with TensorBox models... current handling is unique, but not sophisticated nor correct
- buildout graphical controller
- test image loading w/ box overlays
- provide image scaling services alongside re-sizing the declared boxes
- concentrate coders/decoders under a more consistent file structure/strategy -> right now the A->B, A->C, B->C structures/strategies are messy (multi-file changes)