SqueezeNet_Dataset_TransferLearning
- 1. Istituto Nazionale di Geofisica e Vulcanologia, Sezione di Catania, Osservatorio Etneo & Department of Mathematics and Computer Science, University of Palermo, Italy
- 2. Istituto Nazionale di Geofisica e Vulcanologia, Sezione di Catania, Osservatorio Etneo, Italy
- 3. Istituto Nazionale di Geofisica e Vulcanologia, Sezione di Catania, Osservatorio Etneo & Department of Electrical, Electronic and Computer Engineering, University of Catania, Italy
Description
We have collected and used this dataset for the detection of volcanic thermal anomalies from satellite images. We developed a Deep Learning approach based on a Convolutional Neural Network (CNN) model to detect the presence or not of volcanic thermal activity, using satellite remote sensing data. In particular, we selected 200 images from ESA Sentinel-2 MSI and NASA&USGS Landsat 8 OLI satellite sensors, using the NIR and SWIR bands, and we used the Deep Squeezenet CNN model. This dataset was fed to the CNN model to retrain it, applying the Transfer Learning approach.
We divided the dataset into two classes, "Yes Activity" (Class1, i.e., presence of lava flows and active vents) and "No Activity" (Class0, i.e., volcanoes at rest, in scenes with clouds, snows, etc...). To have a balanced dataset, each class contains the same number of images (100 images).
Files
SqueezeNet_Dataset_TransferLearning.zip
Files
(16.2 MB)
Name | Size | Download all |
---|---|---|
md5:12c9a69cb45e126c2d4b81435020ad5c
|
16.2 MB | Preview Download |