test_visage.py

import streamlit as st
from PIL import Image
import numpy as np
import cv2
from ultralytics import YOLO
import functions
import time
import requests

import time

from io import BytesIO

import base64
import cv2

import cv2
import numpy as np

def pil_to_base64(img):
    buffer = BytesIO()
    img.save(buffer, format="PNG")
    return base64.b64encode(buffer.getvalue()).decode()




st.set_page_config(page_title="Trouve la Poké-pétite", layout="wide")

# Masquer toolbar et footer
hide_streamlit_style = """
<style>
div[data-testid="stToolbar"], div[data-testid="stDecoration"], div[data-testid="stStatusWidget"], #MainMenu, header, footer {
    visibility: hidden;
    height: 0%;
    position: fixed;
}
</style>
"""
st.markdown(hide_streamlit_style, unsafe_allow_html=True)



st.markdown("## Pokemon scanner")

# Charger modèles
@st.cache_resource(show_spinner=False)
def load_models():
    collection, meta = functions.load_faiss_index()
    embedding_model, preprocess, device = functions.load_custom_model("models/dino_small_lora_merged")
    yolo_model = 'models/my-model12.pt'
    model = YOLO(yolo_model)
    return model, collection, meta, embedding_model, preprocess, device

model, collection, meta, embedding_model, preprocess, device = load_models()

print(collection)

image = None

# Inputs
img_file_buffer = st.camera_input("📸 Prends une photo ou sélectionne-en une", key="camera_input")
uploaded_file = None #st.file_uploader("Ou charge une image existante", type=["jpg", "jpeg", "png"])

# Nouvelle image sélectionnée
if img_file_buffer:
    image = Image.open(img_file_buffer).convert("RGB")
    # Réinitialiser les états
    st.session_state.detected_cards = []
    st.session_state.detected_collections = []

elif uploaded_file:
    image = Image.open(uploaded_file).convert("RGB")
    # Réinitialiser les états
    st.session_state.detected_cards = []
    st.session_state.detected_collections = []

# Container pour affichage progressif

# --- Initialisation session_state pour collections ---
if "detected_collections" not in st.session_state:
    st.session_state.detected_collections = []


# Session state pour collection
if "detected_cards" not in st.session_state:
    st.session_state.detected_cards = []

# --- Affichage de l'image principale ---
#image_placeholder = st.empty()  # Placeholder pour l'image source annotée


# Placeholder unique pour toutes les collections
collections_placeholder = st.empty()

# --- Initialisation session_state pour collections ---
if "detected_collections" not in st.session_state:
    st.session_state.detected_collections = []

# --- Traitement de l'image ---
if image is not None:
    img = np.array(image, dtype=np.uint8)
    
    crop_pil = Image.fromarray(img)

    candidate_order_ids = [int(card["order"]) for card in meta]
    distances, indices = functions.search_card_correspondance(
        collection, crop_pil, embedding_model, preprocess , 10, candidate_order_ids
    )

    search_results = []
    for i, idx in enumerate(indices):
        search_results.append(meta[idx])
        search_results[i]['distance_faiss'] = float(distances[i])

    #Un gap est anormal si il est au moins N fois plus grand que la moyenne des 2–3 gaps suivants.
    search_results = sorted(search_results, key=lambda x: x["distance_faiss"], reverse=False)
    gaps = [search_results[i+1]['distance_faiss'] - search_results[i]['distance_faiss'] for i in range(len(search_results)-1)]
    #Pour chaque on regarde les deux suivants
    for j in range(len(gaps)-1):
        # On regarde les 2 gaps suivants (si disponibles)
        next_gaps = gaps[j+1:j+3]
        if not next_gaps:
            continue

        mean_next = sum(next_gaps) / len(next_gaps)

        # Gap anormal si il est >> des gaps suivants
        if gaps[j] > 3 * mean_next:
            search_results = search_results[0:j+1]
            break


                
    # --- Créer la collection ---
    card_data = {
        "crop": img,
        "reference": search_results[0]['img'],
        "set" : search_results[0]['set'],
        "number" : search_results[0]['number'],
        "name": search_results[0]['name'],
        "price": 0 if search_results[0]['price_eur'] ==None else search_results[0]['price_eur'],
        "tcgplayer": search_results[0]['tcgplayer_link'],
        "cardmarket": search_results[0]['cardmarket_link'],
        "history": search_results[0]['price_evolution_url'],
        "distance_faiss": search_results[0]['distance_faiss'],
        "img": search_results[0]['img'],
        "distance_phash": 0
    }

    st.session_state.detected_collections.append({
        "main_card": card_data,
        "search_results": search_results
    })

    # --- Trier toutes les collections par prix décroissant ---
    st.session_state.detected_collections.sort(
        key=lambda c: c["main_card"]["price"], reverse=True
    )


    #image_placeholder.image(img_annotated, use_container_width=True)

    # Mettre à jour l'affichage dans le placeholder
    with collections_placeholder.container():
        for c in st.session_state.detected_collections:
            card_data = c["main_card"]
            search_results = c["search_results"]


            #img_b64 = pil_to_base64(card_data["crop"])
            modal_id = f"modal_img_{i}"  # ID unique pour chaque carte

            md = f"""
            <table style="margin-bottom:12px; border-radius:8px; padding:6px; width:100%;">
                <tr>
                    <td style="width:400px;">
                        <img src="{card_data['img']}" width=150% style="border-radius:4px;"/>
                    </td>
                    <td style="vertical-align:top; padding-left:10px;">
                        <strong style="font-size:30px;">{card_data['name']}</strong><br/>
                        <span style="font-size:16px;">{card_data['set']}/{card_data['number']} </span><br/>
                        <span style="color:#e63946;font-size:20px; font-weight:bold;">💰 {card_data['price']} EUR</span><br/>
                        <span style="font-size:14px;">
                            <a href="{card_data['tcgplayer']}" target="_blank">TCGPlayer</a> | 
                            <a href="{card_data['cardmarket']}" target="_blank">CardMarket</a> | 
                            <a href="{card_data['history']}" target="_blank">Historique</a>
                        </span>
                    </td>
                </tr>
            </table>
            """

            st.markdown(md, unsafe_allow_html=True)





st.markdown("## Analyse terminée")

# --- Valeur totale de la collection ---
total_value = sum([c["main_card"]['price'] for c in st.session_state.detected_collections])
st.markdown(f"##### 💰 Valeur totale : {total_value:.2f} €")