GreenCrossingAI — Notebook quick start

This README covers the notebooks in this repo. There are two distinct workflows:

• Tarazed notebooks — intended for Windows machines with NVIDIA GPUs.
• Video_Processing_Current.ipynb — intended for use on NAIRR Jetstream2 JupyterHub instances.

Tarazed notebooks — Windows GPU machines

Platform: Windows with an NVIDIA GPU. Match CUDA/cuDNN to the ML package you install.

1. Python: 3.8+ recommended.
2. Create and activate a venv (PowerShell):
   • python -m venv .venv
   • .\.venv\Scripts\Activate.ps1
3. Install packages (adjust TensorFlow/PyTorch to match your CUDA):
   • pip install jupyterlab jupyter numpy pandas opencv-python pillow tqdm matplotlib
   • Install a GPU ML package per official instructions (e.g. tensorflow or torch built for your CUDA version).
4. Models & paths:
   • Download MegaDetector/model weights and set notebook variables (MEGADETECTOR_PATH, MODEL_FILE, DATA_DIR, OUTPUT_DIR) in the top cells.
5. Run:
   • jupyter lab
   • Open Notebooks/Tarazed_*.ipynb and run top-to-bottom.

Video_Processing_Current.ipynb — NAIRR (Jetstream2) JupyterHub

Platform: NAIRR Jetstream2 JupyterHub instance (use instance compute/storage).

1. Prefer conda if available; venv works too. Example (bash on the instance):
   • python3 -m venv .venv
   • source .venv/bin/activate
2. Install packages:
   • pip install jupyterlab jupyter numpy pandas opencv-python pillow tqdm matplotlib
3. Data & storage:
   • Use $SCRATCH or project persistent storage for large video/frame sets. Update DATA_DIR/OUTPUT_DIR in the notebook accordingly.
4. MegaDetector / models:
   • Ensure model files and any MegaDetector code are available on the instance and point notebook variables to those paths.
5. Run:
   • Open the JupyterHub web UI, start Video_Processing_Current.ipynb, edit the top cells to set paths, then run cells top-to-bottom.

Common notes

• If a requirements.txt exists, prefer: pip install -r requirements.txt
• Always edit the top cells in each notebook to set DATA_DIR, OUTPUT_DIR, MEGADETECTOR_PATH, and MODEL_FILE before running.
• For GPU usage, install GPU‑compatible TensorFlow or PyTorch per your CUDA/cuDNN and OS setup.