Class Based Thresholding in Early Exit Semantic Segmentation Networks
This addresses computational bottlenecks for real-time semantic segmentation applications, though it is an incremental improvement over existing early exit methods.
The paper tackles computational efficiency in early exit semantic segmentation by proposing Class Based Thresholding (CBT), which reduces computational cost by 23% compared to previous state-of-the-art models while preserving mean intersection over union performance on Cityscapes and ADE20K datasets.
We propose Class Based Thresholding (CBT) to reduce the computational cost of early exit semantic segmentation models while preserving the mean intersection over union (mIoU) performance. A key idea of CBT is to exploit the naturally-occurring neural collapse phenomenon. Specifically, by calculating the mean prediction probabilities of each class in the training set, CBT assigns different masking threshold values to each class, so that the computation can be terminated sooner for pixels belonging to easy-to-predict classes. We show the effectiveness of CBT on Cityscapes and ADE20K datasets. CBT can reduce the computational cost by $23\%$ compared to the previous state-of-the-art early exit models.