File size: 4,534 Bytes
bfae9aa | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 | """
API REST — Pipeline de Escoliose
Endpoint: POST /analyze
Uso:
uvicorn server:app --host 0.0.0.0 --port 8000
"""
import time
import sys
from pathlib import Path
import numpy as np
import torch
import cv2
import requests
import albumentations as A
from albumentations.pytorch import ToTensorV2
from fastapi import FastAPI, HTTPException
from fastapi.middleware.cors import CORSMiddleware
from pydantic import BaseModel
sys.path.insert(0, str(Path(__file__).parent))
from models.hrnet import build_model
from data.dataset import IMAGE_SIZE
# ── Configuração ──────────────────────────────────────────────────────────────
CKPT_PATH = "results/scoliosis_hrnet/best.pth"
TRANSFORM = A.Compose([
A.CLAHE(clip_limit=3.0, p=1.0),
A.Resize(IMAGE_SIZE, IMAGE_SIZE),
A.Normalize(mean=(0.485, 0.456, 0.406), std=(0.229, 0.224, 0.225)),
ToTensorV2(),
])
# ── Carregar modelo uma vez no arranque ───────────────────────────────────────
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model = build_model(arch="hrnet", num_outputs=3)
ckpt = torch.load(CKPT_PATH, map_location=device, weights_only=False)
model.load_state_dict(ckpt["model"])
model = model.to(device).eval()
print(f"[server] Modelo carregado em {device}")
# ── App ───────────────────────────────────────────────────────────────────────
app = FastAPI(title="Scoliosis Analysis API", version="1.0.0")
app.add_middleware(
CORSMiddleware,
allow_origins=["*"],
allow_methods=["POST", "GET"],
allow_headers=["*"],
)
# ── Schemas ───────────────────────────────────────────────────────────────────
class AnalyzeRequest(BaseModel):
estudoId: str
imageUrl: str
class CobbAngles(BaseModel):
thoracic_proximal: float
thoracic_main: float
lumbar: float
class AnalyzeResponse(BaseModel):
estudoId: str
cobb_angles: CobbAngles
max_angle: float
classification: str
processing_time_ms: float
# ── Helpers ───────────────────────────────────────────────────────────────────
def classify(angle: float) -> str:
if angle < 10: return "NORMAL"
if angle < 25: return "LEVE"
if angle < 40: return "MODERADA"
return "GRAVE"
def download_image(url: str) -> np.ndarray:
try:
resp = requests.get(url, timeout=15)
resp.raise_for_status()
except requests.RequestException as e:
raise HTTPException(status_code=400, detail=f"Erro ao descarregar imagem: {e}")
arr = np.frombuffer(resp.content, np.uint8)
img = cv2.imdecode(arr, cv2.IMREAD_GRAYSCALE)
if img is None:
raise HTTPException(status_code=400, detail="Não foi possível descodificar a imagem.")
return cv2.cvtColor(img, cv2.COLOR_GRAY2RGB)
@torch.no_grad()
def run_inference(img: np.ndarray) -> list[float]:
tensor = TRANSFORM(image=img)["image"].unsqueeze(0).to(device)
return model(tensor)[0].cpu().tolist()
# ── Endpoints ─────────────────────────────────────────────────────────────────
@app.get("/health")
def health():
return {"status": "ok", "device": str(device)}
@app.post("/analyze", response_model=AnalyzeResponse)
def analyze(body: AnalyzeRequest):
t0 = time.perf_counter()
img = download_image(body.imageUrl)
angles = run_inference(img)
t1 = time.perf_counter()
max_angle = max(angles)
return AnalyzeResponse(
estudoId = body.estudoId,
cobb_angles = CobbAngles(
thoracic_proximal = round(angles[0], 2),
thoracic_main = round(angles[1], 2),
lumbar = round(angles[2], 2),
),
max_angle = round(max_angle, 2),
classification = classify(max_angle),
processing_time_ms = round((t1 - t0) * 1000, 1),
)
|