Dense CNN Learning with Equivalent Mappings
This addresses a key bottleneck in computer vision for tasks requiring dense pixel-level predictions, offering an incremental improvement over existing methods.
The paper tackled the contradiction between large receptive fields and dense predictions in pixel labeling tasks by proposing equivalent convolution (eConv) and pooling (ePool) layers, which achieved higher accuracy than baseline CNNs across multiple tasks including semantic segmentation and object localization.
Large receptive field and dense prediction are both important for achieving high accuracy in pixel labeling tasks such as semantic segmentation. These two properties, however, contradict with each other. A pooling layer (with stride 2) quadruples the receptive field size but reduces the number of predictions to 25\%. Some existing methods lead to dense predictions using computations that are not equivalent to the original model. In this paper, we propose the equivalent convolution (eConv) and equivalent pooling (ePool) layers, leading to predictions that are both dense and equivalent to the baseline CNN model. Dense prediction models learned using eConv and ePool can transfer the baseline CNN's parameters as a starting point, and can inverse transfer the learned parameters in a dense model back to the original one, which has both fast testing speed and high accuracy. The proposed eConv and ePool layers have achieved higher accuracy than baseline CNN in various tasks, including semantic segmentation, object localization, image categorization and apparent age estimation, not only in those tasks requiring dense pixel labeling.