import imghdr import io import os from io import BytesIO import cv2 import numpy as np import pandas as pd from flask import request, jsonify, render_template import base64 import time import shutil from app.utils.matcher import match_top_n_ocr_to_front_back import tempfile from PIL import Image from pillow_heif import register_heif_opener from app.utils.matcher import match_ocr_to_front_back_by_permuted_ocr, lcs_score register_heif_opener() # register HEIC ALLOWED_EXTENSIONS = {'png', 'jpg', 'jpeg', 'gif'} from app.utils.pill_detection import process_image def safe_get(row, key): val = row.get(key, "") if pd.isna(val): return "" return str(val).strip() def get_fallback_html(): """Fallback to HTML if fail""" return """ Medical Detection APP

Medical Detection APP

服務正常運行中

後端 API 已啟動並可接收請求

使用簡化模板顯示

查看除錯資訊 API 狀態

如果您是開發者，請檢查模板文件是否正確配置

""" def register_routes(app, data_status): """註冊所有路由到 Flask app""" # 從 app 取得數據，如果沒有則創建空的 DataFrame df = getattr(app, 'df', pd.DataFrame()) color_dict = getattr(app, 'color_dict', {}) shape_dict = getattr(app, 'shape_dict', {}) @app.route("/") def index(): try: return render_template("index.html") except Exception as e: print(f"Error rendering template: {e}") return get_fallback_html() @app.route("/healthz") def healthz(): return "ok", 200 @app.route("/debug") def debug(): import json info = { "color_counts": getattr(app, 'color_counts', {}), "status": "running", "cwd": os.getcwd(), "template_folder": app.template_folder, "template_exists": os.path.exists(app.template_folder), "static_folder": app.static_folder, "static_exists": os.path.exists(app.static_folder), "data_status": data_status, "flask_info": { "template_folder": app.template_folder, "static_folder": app.static_folder, "static_url_path": app.static_url_path } } # 列出文件 try: if os.path.exists(app.template_folder): info["template_files"] = os.listdir(app.template_folder) else: info["template_files"] = ["Template folder not found"] except Exception as e: info["template_files"] = [f"Error: {str(e)}"] try: if os.path.exists(app.static_folder): info["static_files"] = os.listdir(app.static_folder) else: info["static_files"] = ["Static folder not found"] except Exception as e: info["static_files"] = [f"Error: {str(e)}"] # 檢查具體文件路徑 info["file_paths"] = { "index.html": os.path.join(app.template_folder, "index.html"), "index.css": os.path.join(app.static_folder, "index.css"), "index.js": os.path.join(app.static_folder, "index.js"), } info["file_exists"] = { path_name: os.path.exists(path) for path_name, path in info["file_paths"].items() } info["color_dict_keys"] = list(color_dict.keys()) info["shape_dict_keys"] = list(shape_dict.keys()) return f""" Debug Info

🔍 Debug Information

System Status

{json.dumps(info, indent=2, ensure_ascii=False)}

Quick Links

← Back to Home

API Status

Test CSS File

Test JS File

""" @app.route("/api/color-stats") def api_color_stats(): buckets = ["白色", "透明", "黑色", "棕色", "紅色", "橘色", "皮膚色", "黃色", "綠色", "藍色", "紫色", "粉紅色", "灰色"] counts = getattr(app, "color_counts", {}) result = {c: int(counts.get(c, 0)) for c in buckets} return jsonify({"counts": result, "total_colors": len(buckets)}) @app.route("/upload", methods=["POST"]) def upload_image(): temp_path = None try: t0 = time.perf_counter() # === 1. 解析 JSON 並確認欄位 === data = request.get_json() if not data or "image" not in data: return jsonify({"ok": False, "error": "缺少 image 欄位"}), 400 b64_data = data["image"] # === 2. 嘗試 base64 header 並解碼 === if b64_data.startswith("data:"): b64_data = b64_data.split(",")[1] image_bytes = base64.b64decode(b64_data) # === 3. 嘗試用 Pillow 解析圖片格式 === image = None try: image = Image.open(io.BytesIO(image_bytes)).convert("RGB") except Exception as e: print(f"[UPLOAD] Pillow 無法辨識圖片格式: {e}") fmt = imghdr.what(None, image_bytes) print(f"[UPLOAD] imghdr 檢測結果: {fmt}") return jsonify({"ok": False, "error": "不支援的圖片格式"}), 400 # === 4. 暫存為圖片檔案（JPEG）=== import tempfile temp_file = tempfile.NamedTemporaryFile(delete=False, suffix=".jpg") temp_path = temp_file.name image.save(temp_path, format="JPEG") temp_file.close() # === 5. 呼叫核心辨識邏輯 === result = process_image(temp_path) or {} t5 = time.perf_counter() if isinstance(result, dict) and "error" in result: print(f" [UPLOAD] 無法偵測藥物: {result['error']}") return jsonify({ "ok": False, "error": "無法偵測藥物，請重新上傳圖片", "result": {"文字辨識": [], "顏色": [], "外型": "", "cropped_image": ""} }), 200 # 回傳 200，表示 API 正常運作，只是無結果 # === 6. 回傳 + 結束 === print( f"[UPLOAD] 推論成功：文字={result['文字辨識']}最佳版本={result['最佳版本']}信心分數={result['信心分數']} 顏色={result['顏色']} 外型={result['外型']}") print(f" [UPLOAD] 完成，總耗時 {(t5 - t0):.2f} s") return jsonify({"ok": True, "result": result}), 200 except Exception as e: import traceback traceback.print_exc() print(f" [UPLOAD] 失敗：{e}") return jsonify({ "ok": False, "error": f"{e}", "result": {"文字辨識": [], "顏色": [], "外型": "", "cropped_image": ""} }), 200 finally: try: if temp_path and os.path.exists(temp_path): os.remove(temp_path) except Exception as e: print(f" [UPLOAD] 臨時檔清理失敗：{e}") @app.route("/api/status") def api_status(): return jsonify({ "status": "running", "version": "1.0.0", "data_loaded": hasattr(app, 'df') and app.df is not None, "data_rows": len(app.df) if hasattr(app, 'df') and app.df is not None else 0, "endpoints": ["/", "/healthz", "/debug", "/api/status"] }) MIN_TOP1_ACCEPT = 0.30 # Top-1 分數低於此值 → 請重拍 HARD_THRESHOLD = 0.80 # 正常門檻 @app.route("/match", methods=["POST"]) def match_drug(): """藥物比對路由""" try: data = request.get_json() texts = data.get("texts", []) colors = data.get("colors", []) shape = data.get("shape", "") if df.empty: print(" [MATCH] 錯誤：資料庫未載入") return jsonify({"error": "資料庫未載入"}), 500 # 尋找候選藥物 candidates = set() # --- 顏色交集 --- color_sets = [] for color in colors: ids = set(color_dict.get(color, [])) # print(f" - 顏色篩選：{color} ➜ {len(ids)} 筆") color_sets.append(ids) if color_sets: candidates = set.intersection(*color_sets) # print(f" 顏色交集後 ➜ {len(candidates)} 筆") else: candidates = set() # --- 外型交集 --- if shape: before_shape = len(candidates) shape_ids = set(shape_dict.get(shape, [])) candidates &= shape_ids # print(f" 外型交集：{shape} ➜ 從 {before_shape} 筆減為 {len(candidates)} 筆") # === 無候選處理 === if not candidates: # print(" [MATCH] 沒有符合的候選藥物") return jsonify({"error": "找不到符合顏色與外型的藥品"}), 404 # 篩選數據 df_sub = df[df["用量排序"].isin(candidates)] if "用量排序" in df.columns else df # print(f"[MATCH] 經過篩選剩下 {len(df_sub)} 筆藥物") # 如果沒有文字或文字為空 if not texts or texts == ["None"]: # print(" [MATCH] 無文字情境，搜尋純顏色/外型比對結果") results = [] for _, row in df_sub.iterrows(): if str(row.get("文字", "")).strip() not in ["F:NONE|B:NONE", "F:None|B:None"]: continue # 尋找藥物圖片 picture_path = os.path.join("data/pictures", f"{row.get('批價碼', '')}.jpg") picture_base64 = "" if os.path.exists(picture_path): try: with open(picture_path, "rb") as f: picture_base64 = f"data:image/jpeg;base64,{base64.b64encode(f.read()).decode('utf-8')}" except Exception as e: print(f"Error reading picture {picture_path}: {e}") results.append({ "name": safe_get(row, "學名"), "symptoms": safe_get(row, "適應症"), "precautions": safe_get(row, "用藥指示與警語"), "side_effects": safe_get(row, "副作用"), "drug_image": picture_base64 }) return jsonify({"candidates": results}) top_matches = match_top_n_ocr_to_front_back(texts, df_sub, threshold=HARD_THRESHOLD, top_n=4) # === 門檻沒過：降門檻取 Top-1 回傳（low_confidence） === if not top_matches: print("[MATCH] 門檻未通過，啟用 Top-1 回傳（low_confidence）") fallback = match_ocr_to_front_back_by_permuted_ocr(texts, df_sub, threshold=0.0) # 從 front/back 取分數最高者 best, best_side = None, None if fallback: for side in ("front", "back"): if side in fallback and fallback[side].get("row") is not None: if (best is None) or (fallback[side]["score"] > best["score"]): best = fallback[side]; best_side = side # 低信心單一結果回傳 if best and best["score"] >= MIN_TOP1_ACCEPT: row = best["row"] if isinstance(row, pd.Series): row = row.to_dict() picture_path = os.path.join("data/pictures", f"{row.get('批價碼', '')}.jpg") picture_base64 = "" if os.path.exists(picture_path): with open(picture_path, "rb") as f: picture_base64 = f"data:image/jpeg;base64,{base64.b64encode(f.read()).decode('utf-8')}" return jsonify({ "name": safe_get(row, "學名"), "symptoms": safe_get(row, "適應症"), "precautions": safe_get(row, "用藥指示與警語"), "side_effects": safe_get(row, "副作用"), "drug_image": picture_base64, "score": round(best["score"], 3), "side": best_side, "low_confidence": True }), 200 # 重拍 return jsonify({ "error": "影像過於模糊或光線不足，建議重拍（請讓藥面填滿畫面、避免反光、對焦清晰）。", "need_retake": True }), 422 # === 正常門檻有結果：組成多筆 candidates 回傳 === results = [] seen = set() # 用來記錄已經加入的藥物 for match in top_matches: row = match["row"] if isinstance(row, pd.Series): row = row.to_dict() # 用「批價碼」作為唯一識別 drug_id = row.get("批價碼", "") if not drug_id or drug_id in seen: continue seen.add(drug_id) picture_path = os.path.join("data/pictures", f"{drug_id}.jpg") picture_base64 = "" if os.path.exists(picture_path): try: with open(picture_path, "rb") as f: picture_base64 = f"data:image/jpeg;base64,{base64.b64encode(f.read()).decode('utf-8')}" except Exception as e: print(f"Error reading picture {picture_path}: {e}") results.append({ "name": safe_get(row, "學名"), "symptoms": safe_get(row, "適應症"), "precautions": safe_get(row, "用藥指示與警語"), "side_effects": safe_get(row, "副作用"), "drug_image": picture_base64, "score": round(match["score"], 3), "match": match["match"], "side": match["side"] }) print(f"🟢 [MATCH] Top-{len(results)} 比對完成，準備回傳") return jsonify({"candidates": results}), 200 except Exception as e: import traceback traceback.print_exc() return jsonify({"error": "Internal server error", "details": str(e)}), 500