Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..
On-Off Center-Surround Receptive Fields for Accurate and Robust Image Classification
Authors: Zahra Babaiee, Ramin Hasani, Mathias Lechner, Daniela Rus, Radu Grosu
ICML 2021 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We provide extensive empirical evidence showing that networks supplied with the OOCS edge representation gain accuracy and illumination-robustness compared to standard deep models. |
| Researcher Affiliation | Academia | 1CPS, TU Wien 2CSAIL, MIT 3IST Austria. |
| Pseudocode | Yes | Algorithm 1 Building and Training OOCS networks |
| Open Source Code | No | The paper does not provide a direct link to open-source code or explicitly state that the code for the described methodology is available. |
| Open Datasets | Yes | We used a subset of the Imagenet (Deng et al., 2009). We used the Norb dataset (Le Cun et al., 2004) to assess the robustness of our proposed OOCS-CNN architecture. We used the MNIST dataset (Le Cun & Cortes, 2010) to train our networks. |
| Dataset Splits | Yes | From these samples, we used 500 of each class for the training set, and 50 for each of the validation and test sets. |
| Hardware Specification | No | The paper does not explicitly describe the specific hardware used (e.g., GPU models, CPU types) for running its experiments, only mentioning implementation in TensorFlow. |
| Software Dependencies | Yes | We implemented all models in Tensor Flow 2.3 (Abadi et al., 2016), and used Adam for optimization (Diederik & Ba, 2015) |
| Experiment Setup | Yes | We used Adam for optimization (Diederik & Ba, 2015) with a learning rate of 10^-4. All convolutional layers have 3x3 kernels and strides of 1. The max-pooling layers have 2x2 kernels and strides of 2. Dropouts (Srivastava et al., 2014) follow the fully connected layers, and the last layer predicts the category with a softmax function. |