Co-segmentation Ablations on PASCAL- Co
We present ablations of the saliency baselines on PASCAL-Co for four randomly chosen sets on several scenarios (see Tab. 3): DINO ResNet, Supervised Resnet, DINO ViT and Supervised ViT. Our method best manages to localize the common objects in the images.
set_164
| Image |  |
 |
 |
 |
 |
 |
|---|---|---|---|---|---|---|
| DINO Resnet |  |
 |
 |
 |
 |
 |
| Sup. Resnet |  |
 |
 |
 |
 |
 |
| Sup. ViT |  |
 |
 |
 |
 |
 |
| DINO ViT |  |
 |
 |
 |
 |
 |
| Ours |  |
 |
 |
 |
 |
 |
set_138
| Image |  |
 |
 |
 |
 |
 |
|---|---|---|---|---|---|---|
| DINO Resnet |  |
 |
 |
 |
 |
 |
| Sup. Resnet |  |
 |
 |
 |
 |
 |
| Sup. ViT |  |
 |
 |
 |
 |
 |
| DINO ViT |  |
 |
 |
 |
 |
 |
| Ours |  |
 |
 |
 |
 |
 |
set_20
| Image |  |
 |
 |
 |
 |
 |
|---|---|---|---|---|---|---|
| DINO Resnet |  |
 |
 |
 |
 |
 |
| Sup. Resnet |  |
 |
 |
 |
 |
 |
| Sup. ViT |  |
 |
 |
 |
 |
 |
| DINO ViT |  |
 |
 |
 |
 |
 |
| Ours |  |
 |
 |
 |
 |
 |
set_27
| Image |  |
 |
 |
 |
 |
 |
|---|---|---|---|---|---|---|
| DINO Resnet |  |
 |
 |
 |
 |
 |
| Sup. Resnet |  |
 |
 |
 |
 |
 |
| Sup. ViT |  |
 |
 |
 |
 |
 |
| DINO ViT |  |
 |
 |
 |
 |
 |
| Ours |  |
 |
 |
 |
 |
 |