fastlabel-mxnet-ssd

Setup

Install libraries

$ pip install -r requirements.txt

Clone original Face Recognition repository

$ git submodule init
$ git submodule update

How to execute

Prepare inputs data

Add downloaded model file(name must be model.tar.gz) to ./data/inputs/model/. Add images to ./data/inputs/images/.

Prediction

$ python main.py

Tips

• COLOR_PALETTE is defined in color_palette.py. You can change it if you want.
• CONFIDENCE_THRESHOLD is defined in main.py. You can change it if you want.

Output

JSON

In ./data/outputs/.

[
  {
    "image": "data/inputs/images/pedestrian.png",
    "prediction": [ // [annotation index, confidence score, top-left x, top-left y, bottom-right x, bottom-right y point]
      [
        0.0, 0.6833887696266174, 237.33348083496094, 53.119441986083984,
        349.9132080078125, 350.6388244628906
      ],
      [
        0.0, 0.665876030921936, 18.011009216308594, 81.62196350097656,
        133.0487518310547, 311.8656005859375
      ]
    ]
  },
  ...
]

Images

In ./data/outputs/images. The annotated images will be output.

Model

In ./data/outputs/model. The unzipped files and the deployable model files will be output.