import argparse
from pathlib import Path

import cv2
import numpy as np

from config import CLASSES, COLORS
from models.utils import blob, det_postprocess, letterbox, path_to_list
from models.utils import det_postprocess_wlm


def main(args: argparse.Namespace) -> None:
    if args.method == 'cudart':
        from models.cudart_api import TRTEngine
    elif args.method == 'pycuda':
        from models.pycuda_api import TRTEngine
    else:
        raise NotImplementedError

    print('start init engine')
    Engine = TRTEngine(args.engine)
    H, W = Engine.inp_info[0].shape[-2:]
    print(f'engine.shape: {Engine.inp_info[0].shape}')

    images = path_to_list(args.imgs)
    save_path = Path(args.out_dir)

    if not args.show and not save_path.exists():
        save_path.mkdir(parents=True, exist_ok=True)

    # for image in images:
    for i in range(0, len(images), 4):
        
        batch_imgs = images[i:i+4]
        #
        batch_tensor = np.empty((4, 3, 640, 640), dtype=np.float32)  # 直接在函数调用中创建并传递元组
        batch_draw = []
        batch_save_iamge = []

        idx = 0
        for image in batch_imgs:
            print(f'idx: {idx} image_name {image.name}')
            save_image = save_path / image.name
            bgr = cv2.imread(str(image))
            draw = bgr.copy()
            bgr, ratio, dwdh = letterbox(bgr, (W, H))
            rgb = cv2.cvtColor(bgr, cv2.COLOR_BGR2RGB)
            tensor = blob(rgb, return_seg=False)
            dwdh = np.array(dwdh * 2, dtype=np.float32)
            tensor = np.ascontiguousarray(tensor)
            batch_tensor[idx] = tensor
            batch_draw.append(draw)
            batch_save_iamge.append(save_image)
            idx += 1
        # inference
        # data = Engine(tensor)
        data = Engine(batch_tensor)
        
        batch_bboxes, batch_scores, batch_labels = det_postprocess_wlm(data)
        # bboxes, scores, labels = det_postprocess(data)
        
        # for image in batch_imgs:
        for i in range(batch_imgs.__len__()):
            bboxes = batch_bboxes[i]
            scores = batch_scores[i]
            labels = batch_labels[i]
            if bboxes.size == 0:
                # if no bounding box
                print(f'{image}: no object!')
                continue
            bboxes -= dwdh
            bboxes /= ratio

            draw = batch_draw[i]
            save_image = batch_save_iamge[i]
            for (bbox, score, label) in zip(bboxes, scores, labels):
                bbox = bbox.round().astype(np.int32).tolist()
                cls_id = int(label)
                cls = CLASSES[cls_id]
                color = COLORS[cls]
                cv2.rectangle(draw, bbox[:2], bbox[2:], color, 2)
                cv2.putText(draw,
                            f'{cls}:{score:.3f}', (bbox[0], bbox[1] - 2),
                            cv2.FONT_HERSHEY_SIMPLEX,
                            0.75, [225, 255, 255],
                            thickness=2)
                if args.show:
                    cv2.imshow('result', draw)
                    cv2.waitKey(0)
                else:
                    cv2.imwrite(str(save_image), draw)


def parse_args():
    parser = argparse.ArgumentParser()
    parser.add_argument('--engine', type=str, help='Engine file')
    parser.add_argument('--imgs', type=str, help='Images file')
    parser.add_argument('--show',
                        action='store_true',
                        help='Show the detection results')
    parser.add_argument('--out-dir',
                        type=str,
                        default='./output',
                        help='Path to output file')
    parser.add_argument('--method',
                        type=str,
                        default='cudart',
                        help='CUDART pipeline')
    args = parser.parse_args()
    return args


if __name__ == '__main__':
    args = parse_args()
    main(args)