import json import numpy as np import streamlit as st from PIL import Image, ImageOps import tensorflow as tf from tensorflow.keras.models import load_model st.set_page_config(page_title="Sign Language Classification", page_icon="🤟", layout="centered") MODEL_PATH = "src/signlanguage_cnn.keras" LABEL_MAP_PATH = "src/label_map.json" @st.cache_resource def load_artifacts(): model = load_model(MODEL_PATH) with open(LABEL_MAP_PATH, "r") as f: label_map = json.load(f) # JSON keys are strings -> convert to int keys label_map = {int(k): v for k, v in label_map.items()} return model, label_map def preprocess_image(img: Image.Image) -> np.ndarray: # Convert to grayscale img = img.convert("L") # Optional: normalize orientation and remove EXIF rotation img = ImageOps.exif_transpose(img) # Resize to 28x28 (Sign-MNIST style) img = img.resize((28, 28)) arr = np.array(img).astype("float32") / 255.0 # Shape: (1, 28, 28, 1) arr = arr.reshape(1, 28, 28, 1) return arr st.title("🤟 Sign Language Classification") st.write("Upload an image of a hand sign and the model will predict the class.") with st.sidebar: st.header("Files required") st.write("- `signlanguage_cnn.keras`") st.write("- `label_map.json`") try: model, label_map = load_artifacts() except Exception as e: st.error("Model could not be loaded. Make sure the files are in the repo root:") st.code("signlanguage_cnn.keras\nlabel_map.json\napp.py") st.exception(e) st.stop() uploaded = st.file_uploader("Upload an image (JPG/PNG)", type=["jpg", "jpeg", "png"]) col1, col2 = st.columns(2) if uploaded is not None: img = Image.open(uploaded) with col1: st.subheader("Input") st.image(img, use_container_width=True) x = preprocess_image(img) probs = model.predict(x, verbose=0)[0] pred_idx = int(np.argmax(probs)) pred_conf = float(np.max(probs)) pred_label = label_map.get(pred_idx, f"Class {pred_idx}") # Top-5 top_k = 5 top_idxs = np.argsort(probs)[::-1][:top_k] top_rows = [] for i in top_idxs: top_rows.append({ "class_id": int(i), "label": label_map.get(int(i), f"Class {int(i)}"), "probability": float(probs[i]) }) with col2: st.subheader("Prediction") st.metric("Predicted class", pred_label) st.metric("Confidence", f"{pred_conf:.2%}") st.write("Top predictions:") st.dataframe(top_rows, use_container_width=True) st.divider() st.caption("Note: The dataset/model may include a placeholder label like 'Unknown' to handle missing/unused indices.") else: st.info("Please upload an image to start.")