File size: 2,787 Bytes
43abac3 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 | # EcoPulse Usage Manual
This guide details how to operate the various components of the EcoPulse platform, from the CLI scripts to the interactive dashboard.
## 1. Interactive Dashboard
The dashboard is the primary interface for visual analysis and presentations.
```bash
streamlit run app.py
```
### Features:
- **Single Image Analysis:** Upload a satellite image to view greenery coverage and Grad-CAM heatmaps.
- **Region Comparison:** Upload two images to contrast their environmental metrics side-by-side.
- **System Controls:** Monitor GPU VRAM usage and clear model cache in real-time.
## 2. CLI Evaluation
To run a batch evaluation of the pipeline against a labeled dataset (e.g., DeepGlobe):
```bash
python scripts/evaluate_segmentation.py
```
This will output Mean IoU and Dice Score metrics to the console and save result visualizations to `outputs/figures/`.
## 3. Region Comparison (CLI)
For rapid comparison of two geographic areas without the GUI:
```bash
python scripts/compare_regions.py --image_a data/area1.jpg --image_b data/area2.jpg
```
**Outputs:**
- `outputs/figures/region_comparison.png`: A side-by-side visual report.
- `outputs/figures/region_comparison_report.csv`: A data report for further analysis.
## 4. CNN Training
To retrain or fine-tune the land-cover classifier on the EuroSAT dataset:
```bash
python scripts/train_classifier.py --config config/config.yaml
```
- Logs are saved to `outputs/logs/` (compatible with TensorBoard).
- The best performing model weights are saved to `outputs/models/resnet50_eurosat.pth`.
## 5. Interpretability Analysis
To run the automated interpretability analysis which generates Grad-CAM heatmaps and Pearson correlation statistics:
```bash
python scripts/evaluate_interpretability.py
```
This generates scatter plots and correlation coefficients in `outputs/figures/`.
## 6. Training on Google Colab
EcoPulse is fully compatible with Google Colab for cloud training.
### Zero-Code-Change Workflow:
1. **Clone the Repo:**
```python
!git clone https://github.com/your-username/ecopulse-satellite-analysis.git
%cd ecopulse-satellite-analysis
```
2. **Install Dependencies:**
```python
!pip install -r requirements.txt
```
3. **Data Mounting:**
If your datasets are on Google Drive, mount them and create a symlink so the internal paths remain valid:
```python
from google.colab import drive
drive.mount('/content/drive')
!ln -s /content/drive/MyDrive/EcoPulseData/data data
```
4. **Execute Training:**
```python
!python scripts/train_classifier.py --config config/config.yaml
```
Because all paths in `config/config.yaml` are relative to the project root, the code will execute perfectly in the Colab environment without modification.
|