Diningtable-unspecified

The following model shows the root filter (column 1), part filters (column 2) and the deformation model (column 3).

Model

Localization Results

The final output of our localization scheme is one bounding box per-image. We show our estimated bounding boxes in the following images. The ground truth bounding boxes are shown in Green. Our bounding box estimate is shown in Blue if it satisfies the PASCAL criterion for correct localization, and in Red if it fails the correct localization criterion.