This repository contains the classification model for the tasks required to be accomplised as a part of the Forest Department Project done by Aerodynamics Club, BITS Goa.
In this project we aim to 3D Map Mangrove Forests around Goa aerially. We use a DJI Phantom Pro v2. The collected images are then post processed and stiched together to obtain a 3D Map of the area. As a part of the project, we built a Machine Learning Model to classify the species observed in the forests. We built our dataset from freely available data online, augmented it an used the same for training.
We performed Transfer Learning on pretrained Resnet18 and VGG models from Pytorch. We use the ensemble of the two for inference.
| Model | Epochs trained | Accuracy | F1 Score |
|---|---|---|---|
Resnet18 |
30 | 81.67 | 0.81 |
VGG |
10 | 83.165 | 0.81 |
Ensemble |
- | 86.334 | 0.86 |
Confusion matrix generated using the ensemble model.
The model was tested on real world data collected from our visits to Mangrove forests. More sample results can be found in the fig directory.
We made an app to run predictions and visualise results from the model. The app is built using tkinter in python. Here are a few screenshots of the app
