An image annotation.

Represents a bounding box on an image, with some additional functionality.

A collection of videos.

Generates additional training examples by taking random crops of the target object, causing apparent translation and scale changes.

An image frame and corresponding annotation.

Easy way to print the time used for various functions. For more advanced timing analysis, we recommend use of a profiler. CLOCK_MONOTONIC_RAW will not

Loads annotations from the ALOV tracking dataset.

Loads images from the ImageNet object detection challenge.

Loads videos from the VOT tracking dataset.

We subclass the Caffe solver object so that we can set protected variables like net_ and test_nets_. (Admittedly, this is probably not the intended us

A neural network for the tracker must inherit from this class.

The class used to train the tracker should inherit from this class.

Manage the iteration over all videos and tracking the objects inside.

Save tracking output and video; record timing.

Track objects and visualize the tracker output.

Container for video data and the corresponding frame annotations.