app.py

import numpy as np
from PIL import Image
from fastapi import FastAPI, File, UploadFile, HTTPException
from fastapi.responses import JSONResponse
import io
import traceback
import logging
import torch
from transformers import DepthProImageProcessorFast, DepthProForDepthEstimation

logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)

# ──────────────────────────────────────────────
app = FastAPI(title="Depth Pro (Apple) — Metric Depth API")

DEVICE = "cpu"
MODEL_ID = "apple/DepthPro-hf"

logger.info(f"Chargement {MODEL_ID} sur {DEVICE} ...")
processor = DepthProImageProcessorFast.from_pretrained(MODEL_ID)
model = DepthProForDepthEstimation.from_pretrained(MODEL_ID)
model = model.to(DEVICE).eval()
logger.info("Modèle Depth Pro prêt.")


# ──────────────────────────────────────────────
@app.post("/predict")
async def predict(file: UploadFile = File(...)):

    # 1) Lecture image
    logger.info(f"Fichier reçu : {file.filename} | type : {file.content_type}")
    if not file.content_type.startswith("image/"):
        raise HTTPException(status_code=400, detail=f"Content-type invalide : {file.content_type}")

    contents = await file.read()
    logger.info(f"Taille : {len(contents)} octets")

    try:
        pil_img = Image.open(io.BytesIO(contents)).convert("RGB")
        logger.info(f"Image : {pil_img.size[0]}x{pil_img.size[1]}")
    except Exception as e:
        logger.error(traceback.format_exc())
        raise HTTPException(status_code=400, detail=f"Impossible de lire l'image : {e}")

    # 2) Prétraitement
    try:
        inputs = processor(images=pil_img, return_tensors="pt").to(DEVICE)
        logger.info(f"Inputs prêts : {list(inputs.keys())}")
    except Exception as e:
        logger.error(traceback.format_exc())
        raise HTTPException(status_code=500, detail=f"Erreur prétraitement : {e}")

    # 3) Inférence
    try:
        logger.info("Inférence en cours ...")
        with torch.no_grad():
            outputs = model(**inputs)
        logger.info("Inférence terminée.")
    except Exception as e:
        logger.error(traceback.format_exc())
        raise HTTPException(status_code=500, detail=f"Erreur inférence : {e}")

    # 4) Post-traitement → profondeur en mètres à la résolution originale
    try:
        post = processor.post_process_depth_estimation(
            outputs,
            target_sizes=[(pil_img.height, pil_img.width)],
        )
        depth_map = post[0]["predicted_depth"].squeeze().cpu().numpy()  # [H, W] float32, mètres
        logger.info(f"depth_map shape={depth_map.shape} min={depth_map.min():.3f} max={depth_map.max():.3f}")
    except Exception as e:
        logger.error(traceback.format_exc())
        raise HTTPException(status_code=500, detail=f"Erreur post-traitement : {e}")

    # 5) Résultats
    H, W = depth_map.shape
    closest_distance = float(np.min(depth_map))
    cy, cx = H // 2, W // 2
    center_distance = float(depth_map[cy, cx])

    logger.info(f"closest={closest_distance:.3f}m | center={center_distance:.3f}m")

    return JSONResponse(content={
        "closest_distance": closest_distance,
        "center_distance":  center_distance,
    })


# ──────────────────────────────────────────────
@app.get("/")
def root():
    return {"status": "ok", "model": MODEL_ID, "device": DEVICE}