Fig. 7
From: Enhancing object pose estimation for RGB images in cluttered scenes
The ablation study results with different configurations. The accuracy of ADD(s), (a) the accuracy of the baseline model with/without self-attention mechanisms, (b) accuracy when removing FPN, (c) excluding the iterative refinement from the model, (d) accuracy after increasing the depth of iterative refinement, (e),(f) accuracies after using different backbones Resnet and Efficientnet respectively.
