🌳 Tree Intelligence Assistant

This AI-powered app helps students and nature enthusiasts identify and explore tree species based on image, location, and tree attributes.

---

🧠 Features

🌲 Recommend Trees by Location

• Input GPS coordinates (latitude, longitude)
• Specify tree diameter, native status, city, and state
• Returns the top 5 tree species likely found in that area

📍 Find Locations for a Tree

• Choose a tree species from a dropdown
• Displays cities and states where the species is most commonly found

📷 Identify Tree from Image

• Upload an image of a tree
• CNN model predicts the species
• If found in the dataset, shows common locations for that species

---

📊 Dataset Description

🗂️ Tree Metadata

• Source: Open tree surveys from multiple cities (e.g., Louisville, Chicago)
• Total records: ~1.38 million
• Key columns:
  • common_name: Tree species (e.g., Bur Oak)
  • scientific_name: Botanical name (e.g., Quercus macrocarpa)
  • latitude_coordinate, longitude_coordinate
  • city, state, address
  • native: Whether the tree is native to the area
  • diameter_breast_height_CM: Tree height/width measure

🖼️ Tree Image Dataset

• Structure: Folder-based, each folder named after a tree species
• Use: Used to train the CNN for species recognition
• Preprocessing:
  • Images resized to 224x224
  • Normalized pixel values
  • Augmented with flips, zoom, and rotation

---

🧪 Algorithms Used

🔍 Recommender System

• Algorithm: K-Nearest Neighbors (KNN)
• Library: sklearn.neighbors.NearestNeighbors
• Inputs: location, diameter, native status, city/state
• Output: Most common tree species nearby

🧠 CNN Classifier

• Model: Sequential CNN (Conv2D + MaxPooling + Dense layers)
• Library: tensorflow.keras
• Input: 224x224 image
• Output: Predicted tree species with probability
• Loss: Categorical Crossentropy
• Optimizer: Adam

📊 Preprocessing & Encoding

• Categorical Encoding: LabelEncoder
• Scaling: StandardScaler for lat/lon/diameter
• Data Splits: 80% training, 20% validation

---

✅ How to Run

1. Run 5M_trees.ipynb to train the recommender and save:

   • tree_data.pkl
   • scaler.joblib
   • nn_model.joblib

2. Run tree_CNN.ipynb to train the image classifier and save:

   • basic_cnn_tree_species.h5

3. Launch the app:

streamlit run streamlit_integrated.py