InsectAI-WG3-STSM – Educational Materials & Tools

This repository was developed during the InsectAI Short-Term Scientific Mission (STSM), hosted by Dr. Paul Bodesheim at the Computer Vision Group, Friedrich Schiller University Jena, Germany, between 18 August and 12 September 2025.

It provides hands-on tutorials and demo applications to train, test, and deploy insect image classifiers without requiring advanced coding knowledge.

---

Repository Structure

• Google Colab Notebooks/
  Train and test deep learning models (TensorFlow/Keras or PyTorch) entirely in Google Colab.

  • train_model_with_keras.ipynb
  • train_model_with_pytorch.ipynb
  • test_model_with_gradio.ipynb

• Hugging Face Spaces Demo/
  Minimal working examples showing how to upload and interact with models via Hugging Face Spaces.

• Google Cloud Demo App/
  A Flask-based deployment example for serving models on Google Cloud Run, including a ready-to-use demo web interface.

---

Quick Start

1. Model Training & Testing (Google Colab)

• Open a notebook from the Google_Colab_Notebooks folder.
• Follow the step-by-step instructions to:
  • Prepare your dataset
  • Train a model (Keras or PyTorch)
  • Export trained models (.keras, .tflite, .pt, .pth)
  • Test predictions interactively with Gradio

No local installation required — everything runs in Colab.

---

2. Hugging Face Demo

The Hugging_Face_Spaces_Demo folder shows how to host your model on Hugging Face Spaces.

You have two options:

• Deploying a Trained Model with Gradio

  • Upload your model files and Python script.
  • Build an interactive demo where users can test your classifier directly in the browser.

• Deploying AutoTrain Models

  • Use Hugging Face’s AutoTrain platform to fine-tune and publish models.
  • Connect your AutoTrain model directly to a Space with minimal effort.

👉 This approach is ideal for quickly sharing results with collaborators without requiring any infrastructure setup.

---

3. Google Cloud Deployment with Flask App

The Google_Cloud_Demo_App folder provides a step-by-step guide to deploy your model as a web application using Flask and Google Cloud Run.

Features:

• Ready-to-use app.py (Flask backend) and index.html (frontend).
• Example trained model included for quick testing.
• Deployment instructions with Dockerfile and requirements.txt.
• Scalable and secure — users can access your app via a public URL.

Example workflow:

1. Train a model in Colab and export it.
2. Place the model file inside the Google_Cloud_Demo_App folder.
3. Build and push a Docker image to Google Cloud.
4. Deploy with Cloud Run and share the URL.

---

License

This repository is released under the MIT License. You are free to use, modify, and share it for educational and research purposes.