Segmentation of Bovid Dentition Under Imperfect Annotations: A Comparative Study of Convolutional and Attention Models
2026-08-31 • Computer Vision and Pattern Recognition
Computer Vision and Pattern RecognitionMachine Learning
AI summaryⓘ
The authors studied how different image segmentation models perform on a dataset of bovid teeth photos, where the masks used for training were handmade and not originally meant for machine learning. They tested various preprocessing methods to fix problems with these imperfect masks. While these fixes didn't greatly change the numbers measuring model accuracy, they did noticeably improve how the final predicted segmentations looked. This suggests that even if the performance scores stay similar, better preparation of training data can help models produce clearer results.
semantic segmentationdeep neural networksmask alignmentB.O.V.I.D. datasetpreprocessingDice scoremIoUconvolutional backbonesvision transformersmachine learning
Authors
Keith G. Mills, Evan B. Sanders, Gregory J. Matthews, Juliet K. Brophy
Abstract
Semantic segmentation decomposes an image into distinct mask regions corresponding to different object categories, such as people, cars, signs or buildings. Advances in machine learning (ML) have shifted this task away from traditional rule-based heuristics such as edge detection, towards deep neural networks (DNN) that learn to classify pixels directly. However, semantic segmentation DNNs crucially depend on expertly designed mask targets to learn from, and imperfect or misaligned masks can interfere with a model's ability to learn effectively. This paper presents a comparative study of segmentation architectures, ranging from convolutional backbones to vision transformers, applied to the B.O.V.I.D. dataset, a corpus of high-resolution bovid dental photographs paired with hand-made segmentation masks not originally designed for ML-based training. We evaluate a range of preprocessing and alignment techniques to mitigate the resulting label imperfections. We find that while these preprocessing choices have limited effect on quantitative metrics such as Dice score and mIoU, their qualitative impact on predicted masks is substantial.