Do Deep Neural Networks Suffer from Crowding?

Authors: Anna Volokitin, Gemma Roig, Tomaso A. Poggio

NeurIPS 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We analyze both deep convolutional neural networks (DCNNs) as well as an extension of DCNNs that are multi-scale and that change the receptive field size of the convolution filters with their position in the image. Our results reveal that the eccentricity-dependent model, trained on target objects in isolation, can recognize such targets in the presence of flankers, if the targets are near the center of the image, whereas DCNNs cannot.
Researcher Affiliation Academia Anna Volokitin Gemma Roig ι Tomaso Poggio voanna@vision.ee.ethz.ch gemmar@mit.edu tp@csail.mit.edu Center for Brains, Minds and Machines, Massachusetts Institute of Technology, Cambridge, MA Istituto Italiano di Tecnologia at Massachusetts Institute of Technology, Cambridge, MA Computer Vision Laboratory, ETH Zurich, Switzerland ιSingapore University of Technology and Design, Singapore
Pseudocode No No structured pseudocode or algorithm blocks were found in the paper.
Open Source Code Yes Code to reproduce experiments is available at https://github.com/CBMM/eccentricity
Open Datasets Yes Examples of the generated images using MNIST [20], not MNIST [21], and Omniglot [22] datasets are depicted in Fig 1, in which even MNIST digits are the target objects.
Dataset Splits No No explicit mention of a validation dataset split or its specifics (e.g., percentages, counts) was found. The paper mentions 'We keep the training and testing splits provided by the MNIST dataset, and use it respectively for training and testing.'
Hardware Specification No No specific hardware details (e.g., exact GPU/CPU models, processor types, memory amounts, or detailed computer specifications) used for running experiments were found.
Software Dependencies No No specific software dependencies with version numbers (e.g., library names like PyTorch, TensorFlow, or specific Python/CUDA versions) were found.
Experiment Setup Yes The input image to the model is resized to 60x60 pixels. In our training, we used minibatches of 128 images, 32 feature channels for all convolutional layers, and convolutional filters of size 5x5 and stride 1.