Topic: ImageNet

We compare how well human observers and neural networks recognize ImageNet images in filtered noise using the critical-band masking paradigm (Fletcher, 1940). We assess the spatial-frequency tuning of object recognition by measuring its noise tolerance at various frequencies. 16 observers and 4 neural networks performed 16-way categorization of 1050 ImageNet images perturbed by band-limited Gaussian noise of five strengths centered at seven spatial frequencies. We find that the noise sensitivity of object recognition is an inverted, U-shaped function of spatial frequency. Human observer perfor

This study investigates whether retinotopic mapping can be integrated into classical computational architectures that model biological visual pathways, specifically deep convolutional neural networks (dCNNs), to enhance image categorization and localization performance. The study retrains a retinotopic dCNN on the ImageNet task, which involves classifying images into 1000 labels.The study's primary finding is that standard deep CNNs are highly effective in processing retinotopic inputs, despite the significant transformation of visual inputs that this mapping entails (A). The ResNet101 network