maonang/inference copy.py

import cv2
import torch
import numpy as np
from PIL import Image
from ultralytics import YOLO
from flask import Flask, request, send_file, make_response
from flask_cors import CORS
import time
from io import BytesIO

# 3个输入参数
img_path = "img0002.jpg"
iou = 0.1
conf = 0.25
originalImgPath = ""
result1Path = ""
result2Path = ""

app = Flask(__name__)
# 解决跨域问题
cors = CORS(
    app,
    resources={
        r"/api/*": {
            "origins": ["http://localhost:5173"],
            "methods": ["GET", "POST"],
        }
    },
)
# 初始化模型
resResult = ""


def split_image(img_path, size=(800, 800)):
    img = cv2.imread(img_path)
    height, width = img.shape[:2]
    rows = (height + size[1] - 1) // size[1]
    cols = (width + size[0] - 1) // size[0]
    img_list = []
    indexes = []
    for r in range(rows):
        for c in range(cols):
            y1 = r * size[1]
            y2 = min((r + 1) * size[1], height)
            x1 = c * size[0]
            x2 = min((c + 1) * size[0], width)
            split = img[y1:y2, x1:x2]
            img_list.append(split)
            indexes.append((r, c))
    return img_list, indexes, (height, width)


def combine_images(pred_imgs, indexes, size=(800, 800), img_shape=(3000, 4000)):
    combined_img = np.zeros((img_shape[0], img_shape[1], 3), dtype=np.uint8)
    for idx, (r, c) in enumerate(indexes):
        y1 = r * size[1]
        y2 = min((r + 1) * size[1], img_shape[0])
        x1 = c * size[0]
        x2 = min((c + 1) * size[0], img_shape[1])
        combined_img[y1:y2, x1:x2] = pred_imgs[idx][: y2 - y1, : x2 - x1]
    return combined_img


follicle_groups_detector = YOLO("follicle_groups.pt")
follicles_detector = YOLO("follicles.pt")


# 保存检测图片
@app.route("/api/checkPng", methods=["POST"])
def checkPng():
    global follicle_groups_detector
    global follicles_detector
    global originalImgPath
    global result1Path
    global result2Path
    global resResult
    pngFile = request.files["file"]
    img = Image.open(pngFile.stream)
    iou = request.form.get("iou")
    iou = float(iou)
    conf = request.form.get("conf")
    conf = float(conf)
    nowtime1 = time.strftime("%Y-%m-%d_%H-%M-%S", time.localtime(time.time()))
    originalImgPath = "images/original" + str(nowtime1) + ".png"
    result1Path = "images/result1" + str(nowtime1) + ".png"
    result2Path = "images/result2" + str(nowtime1) + ".png"
    img.save(originalImgPath)
    results = follicle_groups_detector(originalImgPath, iou=iou, conf=conf)
    for r in results:
        num_follicle_groups = len(r.boxes)
        im_array = r.plot()
        im = Image.fromarray(im_array[..., ::-1])
        im.save(result1Path)  # 输出结果图1

    img_list, indexes, (height, width) = split_image(result1Path)
    print(f"Number of image blocks: {len(img_list)}")
    num_small_follicles = 0
    num_big_follicles = 0
    pred_imgs = []
    for img in img_list:
        results = follicles_detector(img, iou=iou, conf=conf)
        for r in results:
            num_small_follicles += torch.sum(r.boxes.cls == 0).item()
            num_big_follicles += torch.sum(r.boxes.cls == 1).item()
            im_array = r.plot()
            pred_imgs.append(im_array)

    # 输出的3个结果文本
    print("毛囊群数量：", num_follicle_groups)
    print("大毛囊数量：", num_big_follicles)
    print("小毛囊数量：", num_small_follicles)

    combined_img = combine_images(
        pred_imgs, indexes, size=(800, 800), img_shape=(height, width)
    )
    combined_image_pil = Image.fromarray(combined_img[..., ::-1])
    combined_image_pil.save(result2Path)  # 输出结果图2
    resResult = {
        "hasError": False,
        "num_follicle_groups": num_follicle_groups,
        "num_big_follicles": num_big_follicles,
        "num_small_follicles": num_small_follicles,
        "originalImgPath": originalImgPath,
        "result1Path": result1Path,
        "result2Path": result2Path,
    }
    return resResult


# 检测结果返回
@app.route("/api/checkResult", methods=["GET"])
def checkResult():
    global resResult
    return resResult


# 图片查询
@app.route("/api/getPng/<pngPath>", methods=["GET"])
def getPng(pngPath):
    # 打开或处理图像
    img = Image.open("images/" + pngPath)
    # 对图像进行处理，例如调整大小
    img = img.resize((800, 600))
    # 创建一个 BytesIO 对象来保存图像数据
    img_byte_arr = BytesIO()
    img.save(img_byte_arr, format="PNG")
    img_byte_arr.seek(0)

    # 使用 send_file 返回图像数据
    # return send_file(img_byte_arr, mimetype='image/jpeg')

    # 或者使用 make_response 来创建一个响应对象
    response = make_response(img_byte_arr.getvalue())
    response.headers.set("Content-Type", "image/jpeg")
    response.headers.set("Content-Disposition", "attachment", filename=pngPath)

    return response


if __name__ == "__main__":
    app.run(host="127.0.0.1", port=8006, debug=True)