Neural Nearest Neighbors Networks
Authors: Tobias Plötz, Stefan Roth
NeurIPS 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We show its effectiveness for the set reasoning task of correspondence classification as well as for image restoration, including image denoising and single image super-resolution, where we outperform strong convolutional neural network (CNN) baselines and recent non-local models |
| Researcher Affiliation | Academia | Tobias Plötz Stefan Roth Department of Computer Science, TU Darmstadt |
| Pseudocode | No | The paper does not contain structured pseudocode or algorithm blocks. |
| Open Source Code | Yes | Code and pretrained models are available at https://github.com/visinf/n3net/.1 |
| Open Datasets | Yes | We follow the protocol of Zhang et al. [50] and use the 400 images in the train and test split of the BSD500 dataset for training. |
| Dataset Splits | Yes | We follow the protocol of Zhang et al. [50] and use the 400 images in the train and test split of the BSD500 dataset for training. Note that these images are strictly separate from the validation images. |
| Hardware Specification | No | The paper does not provide specific hardware details (exact GPU/CPU models, processor types with speeds, memory amounts, or detailed computer specifications) used for running its experiments. |
| Software Dependencies | No | The paper mentions the use of the Adam optimizer but does not specify software dependencies like libraries or frameworks with version numbers (e.g., Python, PyTorch, TensorFlow, CUDA versions). |
| Experiment Setup | Yes | In total, we train for 50 epochs with a batch size of 32, using the Adam optimizer [21] with default parameters β1 = 0.9, β2 = 0.999 to minimize the squared error. The learning rate is initially set to 10^-3 and exponentially decreased to 10^-8 over the course of training. Following the publicly available implementation of Dn CNN [50], we apply a weight decay with strength 10^-4 to the weights of the convolution layers and the scaling of batch normalization layers. |