Reproduce “Segmentation of Bovid Dentition Under Imperfect Annotations: A Comparative Study of Convolutional and Attention Models” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Keith G. Mills, Evan B. Sanders, Gregory J. Matthews, Juliet K. Brophy
This research developed and evaluated semantic segmentation models for bovid dentition using a dataset with imperfect, hand-crafted annotations, comparing convolutional and transformer architectures. SegFormer MiT-B3 generally produced more accurate and detailed masks, including successful generalization to new data, while U-Net++ with MobileNetV2 demonstrated competent performance given its computational efficiency.
Key insights
- Robust semantic segmentation can be achieved on real-world datasets with imperfect, hand-crafted annotations through targeted preprocessing and appropriate architectural choices.
- Preprocessing techniques had a substantial qualitative impact on predicted mask quality, even when quantitative metrics showed limited differences, highlighting the importance of visual evaluation in noisy data contexts.
- A computationally lightweight CNN (MobileNetV2) demonstrated segmentation competence comparable to a larger transformer-based model (SegFormer MiT-B3) for this task, despite its lower parameter count.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.