Adds SSD: Single Shot MultiBox Detector model

tryolabs / luminoth

Deep Learning toolkit for Computer Vision.

https://tryolabs.com

BSD 3-Clause "New" or "Revised" License

2.4k stars 399 forks source link

Closed gastonrod07 closed 6 years ago

gastonrod07 commented 6 years ago

There is some work to do:

Do the surgery with the VGG16 FC6 and FC7 layer, and the Fully Convolutional Reduced Network
Analyze "big anchors" problem (currently the big objects aren't well delimited)
Improve the documentation
Add more configurable parameters in the config file
Add compatibility with 512x512 images
Add compatibility with other base networks (beyond VGG16)
Use data augmentation