mSAIL: Milligram-Scale Multi-Modal Sensor & Analytics Monitoring Platform for Monarch Butterfly Migration Tracking

Training Data:

The voluneteer light and temeperature data for 2018, 2019, 2020 are stored on ./Light_Data and ./Temp_Data. Volunteer data coordinates are intentionally made inaccurate in the resolution of 10km and we will gradually populate more data as we get permission from volunteers.

Generate the training and testing data for neural network

• In order to generate the training data, you need to simply run the Generate_trainset_light.m and Generate_trainset_tmp.m.

• In order to generate the testing data, you need to simply run the Generate_testset_light.m and Generate_testset.m.

Neural network models:

Neural network models are located in ./Model_codes. It includes both Python and Jupyter notebook versions. You can simply run the Mobicom_Temp.ipynb and Mobicom_light.ipynb to generate the heatmap results. The pretrained neural network models are located on ./Trained_models/ .

Heatmap and Visualization:

Heatmap_Mobicom.m and Visualization_Sampling.m contain the codes to generate the localization results from the heatmaps generated from the NN models. They also include the codes to generate the following pictures.

• Heatmap results:

• Absolute Mean Error:

Contact Information:

The platform as a whole as well as individual chips are available at cost to academic researchers. If you have any questions about the chips, please contact the following emails: hunseok@umich.edu , inhee.lee@pitt.edu , and blaauw@umich.edu .