Find research datasets worth reusing

opencc-by-4.0Feb 2024View details →

Dataset for training Dev-ResNet for detection of developmental events in Lymnaea stagnalis

Dataset for training <a href="https://github.com/EmbryoPhenomics/dev-resnet">Dev-ResNet</a> for detection of developmental events in Lymnaea stagnalis.    The following naming scheme is used for the individual GIF files: Example: "./video/20C_A_B1_315hr.gif" <ul> <li>"./video/"  -->  Source folder.</li> <li>"20C"        -->  Treatment, here incubation temperature.</li> <li>"A"            -->  Egg mass label.</li> <li>"B1"          -->  Individual embryo label.</li> <li>"315hr"     -->  Hours since first cell division.</li> </ul>

opencc-by-4.0Aug 2022View details →

Model Zoo: A Dataset of Diverse Populations of Resnet-18 Models - CIFAR-10

Abstract In the last years, neural networks have evolved from laboratory environments to the state-of-the-art for many real-world problems. Our hypothesis is that neural network models (i.e., their weights and biases) evolve on unique, smooth trajectories in weight space during training. Following, a population of such neural network models (refereed to as “model zoo”) would form topological structures in weight space. We think that the geometry, curvature and smoothness of these structures contain information about the state of training and can be reveal latent properties of individual models. With such zoos, one could investigate novel approaches for (i) model analysis, (ii) discover unknown learning dynamics, (iii) learn rich representations of such populations, or (iv) exploit the model zoos for generative modelling of neural network weights and biases. Unfortunately, the lack of standardized model zoos and available benchmarks significantly increases the friction for further research about populations of neural networks. With this work, we publish a novel dataset of model zoos containing systematically generated and diverse populations of neural network models for further research. In total the proposed model zoo dataset is based on six image datasets, consist of 24 model zoos with varying hyperparameter combinations are generated and includes 47’360 unique neural network models resulting in over 2’415’360 collected model states. Additionally, to the model zoo data we provide an in-depth analysis of the zoos and provide benchmarks for multiple downstream tasks as mentioned before. Dataset This dataset is part of a larger collection of model zoos and contains the zoo of 1000 ResNet18 models trained on CIFAR10. All zoos with extensive information and code can be found at www.modelzoos.cc. The complete zoo is 2.3TB large. Due to the size, this repository contains the checkpoints of epochs 1, 10 and 50. For a link to the full dataset as well as more information on the zoos and code to access and use the zoos, please see www.modelzoos.cc.

opencc-by-4.0Aug 2022View details →

Model Zoo: A Dataset of Diverse Populations of Resnet-18 Models - CIFAR-100

Abstract In the last years, neural networks have evolved from laboratory environments to the state-of-the-art for many real-world problems. Our hypothesis is that neural network models (i.e., their weights and biases) evolve on unique, smooth trajectories in weight space during training. Following, a population of such neural network models (refereed to as “model zoo”) would form topological structures in weight space. We think that the geometry, curvature and smoothness of these structures contain information about the state of training and can be reveal latent properties of individual models. With such zoos, one could investigate novel approaches for (i) model analysis, (ii) discover unknown learning dynamics, (iii) learn rich representations of such populations, or (iv) exploit the model zoos for generative modelling of neural network weights and biases. Unfortunately, the lack of standardized model zoos and available benchmarks significantly increases the friction for further research about populations of neural networks. With this work, we publish a novel dataset of model zoos containing systematically generated and diverse populations of neural network models for further research. In total the proposed model zoo dataset is based on six image datasets, consist of 27 model zoos with varying hyperparameter combinations are generated and includes 50’360 unique neural network models resulting in over 2’585’360 collected model states. Additionally, to the model zoo data we provide an in-depth analysis of the zoos and provide benchmarks for multiple downstream tasks as mentioned before. Dataset This dataset is part of a larger collection of model zoos and contains the zoo of 1000 ResNet18 models trained on CIFAR100. All zoos with extensive information and code can be found at www.modelzoos.cc. The complete zoo is 2.6TB large. Due to the size, this repository contains the checkpoints of the last epoch 60. For a link to the full dataset as well as more information on the zoos and code to access and use the zoos, please see www.modelzoos.cc.

opencc-by-4.0Aug 2022View details →

Model Zoo: A Dataset of Diverse Populations of Resnet-18 Models - Tiny ImageNet

Abstract In the last years, neural networks have evolved from laboratory environments to the state-of-the-art for many real-world problems. Our hypothesis is that neural network models (i.e., their weights and biases) evolve on unique, smooth trajectories in weight space during training. Following, a population of such neural network models (refereed to as “model zoo”) would form topological structures in weight space. We think that the geometry, curvature and smoothness of these structures contain information about the state of training and can be reveal latent properties of individual models. With such zoos, one could investigate novel approaches for (i) model analysis, (ii) discover unknown learning dynamics, (iii) learn rich representations of such populations, or (iv) exploit the model zoos for generative modelling of neural network weights and biases. Unfortunately, the lack of standardized model zoos and available benchmarks significantly increases the friction for further research about populations of neural networks. With this work, we publish a novel dataset of model zoos containing systematically generated and diverse populations of neural network models for further research. In total the proposed model zoo dataset is based on six image datasets, consist of 27 model zoos with varying hyperparameter combinations are generated and includes 50’360 unique neural network models resulting in over 2’585’360 collected model states. Additionally, to the model zoo data we provide an in-depth analysis of the zoos and provide benchmarks for multiple downstream tasks as mentioned before. Dataset This dataset is part of a larger collection of model zoos and contains the zoo of 1000 ResNet18 models trained on Tiny Imagenet. All zoos with extensive information and code can be found at www.modelzoos.cc. The complete zoo is 2.6TB large. Due to the size, this repository contains the checkpoints of the first 115 models at their last epoch 60. For a link to the full dataset as well as more information on the zoos and code to access and use the zoos, please see www.modelzoos.cc.

opencc-by-4.0Nov 2024View details →

ResNet

This includes ResNet, a dataset of 57,452 United States dams on National Hydrography Dataset (NHD) flowlines, incorporating input from the National Inventory of Dams (NID) and the Global Reservoir and Dam Database (GRanD). We also cross-reference our network with prominent global dam databases (resnet_datacrossreference), which were previously not co-located. We include the resnet_attributes file, where users can input additional dam data such as location, storage, or year completed. This file also allows a user to implement the site tagging function of ResNet. We include our manually placed GRanD dams input file, the NHD flowlines with the CountryOut field appended, and the removed dams database, which are all inputs to re-run the code on GitHub. To re-run the code, a user needs to download the NHDFlowline_Network_NHDPlus_Countries.gdb.zip and NHD_Flowline_Network_countriesOut.csv files and unzip them into the Inputs folder in the Github file directory.

opencc-by-4.0Oct 2023View details →

GTDA-based ReebNet for full ImageNet data with ResNet, VOLO, and AlexNet

This folder is used to produce the experiments of ImageNet reported in paper Topological structure of complex predictions. A preprint version of the paper is available on arXiv: https://arxiv.org/abs/2207.14358Unzip the three .tar.gz files. At the end, you should have a directory that looks like:<ul><li>exp_comparison_plotting_resnet_alexnet.py</li><li>exp_comparison_plotting_volo_resnet.py</li><li>GTDA</li><li>GTDA.tar.gz</li><li>GTDA.yml</li><li>ImageNet</li><li>ImageNet-details.txt</li><li>meta.mat</li><li>README.txt</li><li>resnet_alexnet</li><li>resnet_alexnet.tar.gz</li><li>volo_resnet</li><li>volo_resnet.tar.gz</li></ul>The ImageNet files are too large to include and are available from many other sources. You can verify you have the right setup by checking the ImageNet-details.txt file, which has the sha sum for each file we used along with the directory structure. There are additional instructions inside the REAMDE.txt file on how to setup the python environment to reproduce the figures.

opencc-by-4.0Oct 2024View details →

ResNet and Asymmetric ResNet Weights, for "The Empirical Impact of Neural Parameter Symmetries, or Lack Thereof"

Trained checkpoints of tens of thousands of neural networks, for use in our paper "The Empirical Impact of Neural Parameter Symmetries, or Lack Thereof" (https://arxiv.org/abs/2405.20231).

opencc-by-4.0Oct 2024View details →

DiCARN-DNase: Enhancing Cell-to-Cell Hi-C Resolution Using Dilated Cascading ResNet with Self-Attention and DNase-seq Chromatin Accessibility Data

This repository contains the training, validation, and testing data for the DiCARN-DNase project. The GitHub repository is available via https://github.com/OluwadareLab/DiCARN_DNase

opencc-by-4.0Aug 2021View details →

StarDist model for Arabidopsis using Resnet backbone

Stardist model for Arabidopsis using Resnet backbone to be used in conjugation with U-Net model in vollseg setting

opencc-by-4.0Jun 2020View details →

Training and Testing Data, Associated Code, and SCAM Validations Code and Data for ResNet in moist physics (ResCu)

Data and codes for a deep convolutional residual neural network moist physics parameterization (ResCu). In this new version, the randomly selected training data samples and part of testing data samples (June, July and August) are provided. They are processed into a new data structure, which can be directly utilized in training and testing. For the entire second year training samples and the entire third year testing samples, we provide them in a repository at Dryad (<a href="https://doi.org/10.6075/J0CZ35PP">https://doi.org/10.6075/J0CZ35PP</a> and https://doi.org/10.6075/J03J3BGF). Please download and decompress ResCu_Han_et_al_JAMES.tar.gz. Follow the instructions in README.txt and download the training and testing data (The Dryad depositary is provided in the description).  Here we provide 3 parts of data and codes: 1, Training and testing data from SPCAM; 2, Training and testing codes for ResCu and many other NN architectures; 3, SCAM validations.

opencc-by-4.0Nov 2024View details →

Dynamic Visualization of ResNet Layer Activations for Brain Health Classification

This GIF file provides a dynamic visualization of the internal representations (activations) from the ResNet 18 model layers (2 to 69) during a brain health classification task. The sequence begins by showing the original input image, followed by successive activation maps visualized using the "jet" colormap. Each frame corresponds to the activations extracted from a specific layer in the ResNet, resized to match the input image dimensions for better interpretability.  The dataset used for this visualization is from S. Bhuvaji, "Brain Tumor Classification MRI," published on Kaggle in 2023. This dataset contains MRI images of brain tumors and has been utilized to train and evaluate the ResNet model for the classification of brain health states. The input sample displayed in the GIF is one such MRI image from the dataset, highlighting the model's ability to extract and analyze features relevant to brain tumor diagnosis. The activations reveal how the ResNet processes the input image hierarchically. In the early layers (e.g., Layers 2-10), the network preserves much of the spatial structure of the original image, focusing on edges and low-level features. Moving to the intermediate layers (e.g., Layers 11-40), the network begins to emphasize localized patterns while filtering out irrelevant structures such as the skull, concentrating instead on regions associated with tumors or health-related features. Finally, the deep layers (e.g., Layers 41-69) extract highly abstract and classification-relevant patterns, concentrating on tumor-related features while discarding most of the background. The progression of the activations in the GIF demonstrates how the network transitions from general image features to highly specialized, diagnostic features that are critical for the classification task. This visualization helps provide an intuitive understanding of the hierarchical processing capabilities of convolutional neural networks (CNNs) in medical image analysis. <h3>Key Features:</h3> The visualization includes an input brain image and its corresponding activation maps, extracted from each ResNet layer. The activation maps are resized to match the original image for consistency, and the "jet" colormap is applied to enhance visual interpretation of activation intensities. Each frame in the GIF dynamically updates to show the activations of the next layer, offering an engaging representation of the network's internal behavior. <h3>Use Cases:</h3> This GIF is a valuable resource for education, research, and presentations. It can be used to illustrate how deep learning models process medical images, providing insights into the hierarchical feature extraction process. Researchers and educators can leverage this visualization to explain the concept of feature abstraction in CNNs. It is also ideal for inclusion in talks, posters, and papers to showcase the dynamic analysis of neural network activations.

opencc-by-4.0Jun 2023View details →

ResNet 34 Ensemble Predictions on TinyImageNet

Predictions generated by an ensemble of 4 ResNet 34 Deep Neural Networks Trained on TinyImagenet, as used in repository https://anonymous.4open.science/r/ensemble_attention-7616/README.md. Ensembles are trained to encourage/discourage predictive diversity. Each timestamped folder contains individual training runs, with the labels and probabilistic predictions of the ensemble on 1) the training set (train_labels.npy, train_preds.npy) and 2) the test set (ind_labels.npy, ind_preds.npy) for tinyimagenet. The file (tinyimagenet/resnet34/version_0/hparams.yaml) contains specific hyperparameters used on a particular training run. Figures visualizing training results can be generated by: 1.unzipping the four folders in to the directory `ensemble_attention/scripts/outputs/`  2. running the script `ensemble_attention/scripts/vis_scripts/all_weights_resnet34_tinyimagenet.py`.