/ssd_keras

Port of Single Shot MultiBox Detector to Keras

Primary LanguageJupyter NotebookMIT LicenseMIT

A port of SSD: Single Shot MultiBox Detector to Keras framework.

For more details, please refer to arXiv paper. For forward pass for 300x300 model, please, follow SSD.ipynb for examples. For training procedure for 300x300 model, please, follow SSD_training.ipynb for examples. Moreover, in testing_utils folder there is a useful script to test SSD on video or on camera input.

Weights are ported from the original models and are available here. You need weights_SSD300.hdf5, weights_300x300_old.hdf5 is for the old version of architecture with 3x3 convolution for pool6.

Memo

I didn't try below codes yet.

  • Howt to convert caffe weights to keras weights?

https://gist.github.com/rykov8/5fa4d2faff2e5e4704df54729bf57159

  • How to create own training data?

Create gt_pascal.pkl. Don't need to create prior_boxes_ssd300.pkl because it depends on ssd300 architecture.

To create gt_pascal.pkl

# data type: dict
# key: filename
# value: [xmin, ymin, xmax, ymax, label (one-hot vector)]
#        maybe, xmin, ymin, xmax, ymax are 0. to 1. for regularized input image

# sample (detect 4 classes)
{
  'frame00600.png' : [
    [xmin, ymin, xmax, ymax, 1, 0, 0, 0],
    [xmin, ymin, xmax, ymax, 0, 1, 0, 0]
  ],
  'frame00601.png' : [
    [xmin, ymin, xmax, ymax, 0, 1, 0, 0],
    [xmin, ymin, xmax, ymax, 0, 1, 0, 0]
  ],
  ...
}