app.py

import gradio as gr
import pandas as pd
import numpy as np
import cv2
from sklearn.neighbors import KNeighborsClassifier
from sklearn.preprocessing import LabelEncoder

# Load CSVs
df_train = pd.read_csv("train_dataset.csv")

# Split features and labels
X_train = df_train.drop(columns=["label"]).values
y_train = df_train["label"].values

# Encode labels (if they're strings)
le = LabelEncoder()
y_train_encoded = le.fit_transform(y_train)

# Train KNN
knn = KNeighborsClassifier(n_neighbors=3)
knn.fit(X_train, y_train_encoded)

# Constants
IMAGE_SIZE = 64  # should match training

# Prediction function
def predict_animal(image):
    # Convert image to grayscale and resize
    image = cv2.resize(image, (IMAGE_SIZE, IMAGE_SIZE))
    gray = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY)
    flat = gray.flatten().reshape(1, -1)

    # Predict
    pred_encoded = knn.predict(flat)[0]
    pred_label = le.inverse_transform([pred_encoded])[0]
    return pred_label

# Gradio UI
gr.Interface(
    fn=predict_animal,
    inputs=gr.Image(type="numpy", label="Upload Animal Image"),
    outputs=gr.Label(label="Predicted Animal"),
    title="Animal Classifier (KNN)",
    description="Upload an animal image to classify it using a KNN model trained on CSV data."
).launch()