v0.1.0

First release of aistron

Major features:

• Data pipeline: We aim to standardize the annotations of existing AIS datasets so that the methods can be easily implemented and generalized. This diagram illustrates the data pipeline of aistron

• Amodal Instance Segmentation Evaluator: AIStron helps compute the performance of both visible and amodal masks. If a method provides both pred_visible_masks and pred_amodal_masks in its predictions, both performances will be computed. If only the conventional pred_masks are provided, only the amodal performance is computed.

• Utilities: We offer an amodal visualizer utility that allows you to visualize the ground truth or predictions with option to choose between visible masks, occluder masks and amodal masks.

• Builtin-Methods: Amodal Mask R-CNN