Crowd density estimation

This project utilises a VGG model with Bayesian loss to estimate the desity of a crowd.

Model Settings

device: 'cpu' # Device string used for pytorch (options: 'cpu'| 'gpu')

Model config

model_path: 'weight.pth'
mean: [0.485, 0.456, 0.406]
std: [0.229, 0.224, 0.225]
cfg:
  E: [ 64, 64, 'M', 128, 128,
      'M', 256, 256, 256, 256,
      'M', 512, 512, 512, 512,
      'M', 512, 512, 512, 512 ]

Input/ Output

• Input: It receives the RGB image as a FrameObject:

  def inference(self, frame_object: FrameObject) -> Inference

• Output: Updates the Inference.extra object with the crowd_number

  return Inference(extra={"crowd_number": crowd_no})