AlphaVideo

Video understanding toolbox developed by MVIG in SJTU, for the moment, providing models for action recognition and multi-object tracking.

Quick Start

Run this command:

pip install alphavideo

Pre-requisites

Installation

pip

Run this command:

pip install alphavideo

from source

Clone repository from github:

git clone https://github.com/ alphavideo
cd alphavideo

Setup and install MVIdeoGallery:

pip install .

Verification

Firstly, check the version of MVIdeoGallery.

import alphavideo
print(alphavideo.__version__)

To ensure that MVIdeoGallery was installed correctly, we can verify the installation by running sample MVIdeoGallery code. Here we will construct a MOT model.

model = alphavideo.model.tubeTK(pretrain=False)
print(model)

The output should be something similar to:

...(other structure)...
(TubeTK_cls): Conv3d(256, 1, kernel_size=(3, 3, 3), stride=(1, 1, 1), padding=(1, 1, 1))
    (TubeTK_reg): Conv3d(256, 14, kernel_size=(3, 3, 3), stride=(1, 1, 1), padding=(1, 1, 1))
    (TubeTK_centerness): Conv3d(256, 1, kernel_size=(3, 3, 3), stride=(1, 1, 1), padding=(1, 1, 1))
    (scales): ModuleList(
      (0): Scale()
      (1): Scale()
      (2): Scale()
      (3): Scale()
      (4): Scale()
    )
  )
)