from huggingface_hub import hf_hub_download import tensorflow as tf import numpy as np from PIL import Image, ImageOps import gradio as gr # ----------------------- # CONFIG # ----------------------- MODEL_REPO_ID = "BunkerMinecraft/neurecaiv2" # <-- change this MODEL_FILENAME = "keras_model.h5" LABELS_FILENAME = "labels.txt" IMAGE_SIZE = (224, 224) # TM fixed size # ----------------------- # DOWNLOAD MODEL + LABELS # ----------------------- model_path = hf_hub_download( repo_id=MODEL_REPO_ID, filename=MODEL_FILENAME, repo_type="model" ) labels_path = hf_hub_download( repo_id=MODEL_REPO_ID, filename=LABELS_FILENAME, repo_type="model" ) # Load the Teachable Machine Keras .h5 model model = tf.keras.models.load_model(model_path, compile=False) print(f"Loaded model: {model_path}") # Load labels with open(labels_path, "r") as f: class_names = [line.strip() for line in f.readlines()] print("Loaded labels:", class_names) # ----------------------- # PREPROCESS FUNCTION (matches TM) # ----------------------- def preprocess(img: Image.Image): img = img.convert("RGB") img = ImageOps.fit(img, IMAGE_SIZE, Image.Resampling.LANCZOS) image_array = np.asarray(img).astype(np.float32) # TM normalization: normalized = (image_array / 127.5) - 1.0 return np.expand_dims(normalized, axis=0) # ----------------------- # PREDICT FUNCTION # ----------------------- def predict(image: Image.Image): x = preprocess(image) preds = model.predict(x, verbose=0)[0] idx = int(np.argmax(preds)) conf = float(preds[idx]) return class_names[idx], conf # ----------------------- # UI (Gradio) # ----------------------- interface = gr.Interface( fn=predict, inputs=gr.Image(type="pil", label="Upload an image"), outputs=[gr.Label(label="Predicted class"), gr.Number(label="Confidence")], title="Teachable Machine Classifier", description="Upload an image to classify using your keras_Model.h5 Teachable Machine model." ) if __name__ == "__main__": interface.launch()