torchvision nms

要求格式[x1,y1,x2,y2]

if __name__ == '__main__':

    import torch
    a=torch.Tensor([[1,1,2,2],[1,1,3.100001,3],[1,1,3.1,3]])
    b=torch.Tensor([[0.9],[0.98],[0.980005]])

    from torchvision.ops import nms

    ccc=nms(a,b,0.4)
    print(ccc)
    print(a[ccc])

 

from __future__ import print_function
import os
import argparse
import torch
import torch.backends.cudnn as cudnn
import numpy as np
from torchvision.ops import nms

from data import cfg_mnet, cfg_re50, cfg_peleenet
from layers.functions.prior_box import PriorBox
from utils.nms.py_cpu_nms import py_cpu_nms
import cv2
from models.retinaface import RetinaFace
from utils.box_utils import decode, decode_landm
import time

parser = argparse.ArgumentParser(description='Retinaface')

# parser.add_argument('-m', '--trained_model', default='weights/mobilenet0.25_Final.pth',type=str)
# parser.add_argument('-m', '--trained_model', default='weights/0.8296_0.9090_0.7629_12.5676_1.0e-06_8.pth',type=str)
parser.add_argument('-m', '--trained_model', default='158/0.8289_0.9060_0.7638_23.3821_1.0e-04_22.pth',type=str)
parser.add_argument('--network', default='peleenet', help='Backbone network mobile0.25 or resnet50')
parser.add_argument('--cpu', action="store_true", default=False, help='Use cpu inference')
parser.add_argument('--confidence_threshold', default=0.9, type=float, help='confidence_threshold')
parser.add_argument('--top_k', default=20, type=int, help='top_k')
parser.add_argument('--nms_threshold', default=0.3, type=float, help='nms_threshold')
parser.add_argument('--keep_top_k', default=12, type=int, help='keep_top_k')
parser.add_argument('--save_image', action="store_true", default=True, help='show detection results')
parser.add_argument('--vis_thres', default=0.6, type=float, help='visualization_threshold')
args = parser.parse_args()



def check_keys(model, pretrained_state_dict):
    ckpt_keys = set(pretrained_state_dict.keys())
    model_keys = set(model.state_dict().keys())
    used_pretrained_keys = model_keys & ckpt_keys
    unused_pretrained_keys = ckpt_keys - model_keys
    missing_keys = model_keys - ckpt_keys
    print('Missing keys:{}'.format(len(missing_keys)))
    print('Unused checkpoint keys:{}'.format(len(unused_pretrained_keys)))
    print('Used keys:{}'.format(len(used_pretrained_keys)))
    assert len(used_pretrained_keys) > 0, 'load NONE from pretrained checkpoint'
    return True


def remove_prefix(state_dict, prefix):
    ''' Old style model is stored with all names of parameters sharing common prefix 'module.' '''
    print('remove prefix \'{}\''.format(prefix))
    f = lambda x: x.split(prefix, 1)[-1] if x.startswith(prefix) else x
    return {f(key): value for key, value in state_dict.items()}


def load_model(model, pretrained_path, load_to_cpu):
    print('Loading pretrained model from {}'.format(pretrained_path))
    if load_to_cpu:
        pretrained_dict = torch.load(pretrained_path, map_location=lambda storage, loc: storage)
    else:
        device = torch.cuda.current_device()
        pretrained_dict = torch.load(pretrained_path, map_location=lambda storage, loc: storage.cuda(device))
    if "state_dict" in pretrained_dict.keys():
        pretrained_dict = remove_prefix(pretrained_dict['state_dict'], 'module.')
    else:
        pretrained_dict = remove_prefix(pretrained_dict, 'module.')
    check_keys(model, pretrained_dict)
    model.load_state_dict(pretrained_dict, strict=False)
    return model


if __name__ == '__main__':
    torch.set_grad_enabled(False)
    cfg = None
    if args.network == "mobile0.25":
        cfg = cfg_mnet
    elif args.network == "peleenet":
        cfg = cfg_peleenet
    elif args.network == "resnet50":
        cfg = cfg_re50
    # net and model
    cfg['num_classes'] = 1
    net = RetinaFace(cfg=cfg, phase = 'test')
    net = load_model(net, args.trained_model, args.cpu)
    net.eval()
    print('Finished loading model!')
    # print(net)
    cudnn.benchmark = True
    device = torch.device("cpu" if args.cpu else "cuda")
    net = net.to(device)

    # vc = cv2.VideoCapture(0)  # 讀入視頻文件
    d = 0
    exit_flag = False
    c = 0
    images = []
    images_origin = []
    time1 = time.time()

    im_width=320*2
    im_height=180*2

    # img_raw = cv2.resize(img_raw, (width, height))
    # im_height, im_width, _ = img_raw.shape
    scale = torch.Tensor([im_width, im_height, im_width, im_height])
    scale = scale.to(device)
    priorbox = PriorBox(cfg, image_size=(im_height, im_width))
    priors = priorbox.forward()
    priors = priors.to(device)

    path=r"\\192.168.25.73\Team-CV\dataset\chumao_train\video\20190528_124414\JPEGImages/"
    list_path=r"F:\data\VOCdevkit2007\VOC2007/"
    # path=r"\\192.168.25.73\Team-CV\dataset\chumao_train\coco_0719\JPEGImages/"
    # path=r"\\192.168.25.73\Team-CV\dataset\chumao_train_1024\head_big\0806\JPEGImages/"
    # path=r"D:\Team-CV\dataset\chumao_train_1024\chumao_head\JPEGImages/"

    # testset_list =path[:-7] + "wider_val.txt"
    #
    # with open(testset_list, 'r') as fr:
    #     test_dataset = fr.read().split()

    g = os.walk(list_path)
    test_dataset = ['%s/%s' % (i[0], j) for i in g if i[0].endswith('JPEGImages') for j in i[-1] if
                      j.endswith('jpg')]

    num_images = len(test_dataset)


    for i, file in enumerate(test_dataset):
        img_raw=cv2.imread(file)
        img_raw=cv2.resize(img_raw,(im_width, im_height))

        img = np.float32(img_raw)
        tic = time.time()
        img -= (104, 117, 123)
        img = img.transpose(2, 0, 1)
        img = torch.from_numpy(img).unsqueeze(0)
        img = img.to(device)

        loc, conf = net(img)  # forward pass

        prior_data = priors.data
        boxes = decode(loc.data.squeeze(0), prior_data, cfg['variance'])
        boxes = boxes * scale
        # boxes = boxes.cpu().numpy()
        scores = torch.sigmoid(conf.squeeze(0).squeeze(1))#.data.cpu().numpy()


        # ignore low scores
        inds =scores >= args.confidence_threshold
        boxes = boxes[inds]
        # landms = landms[inds]
        scores = scores[inds]

        scores,order= scores.topk(min((scores.size(0), args.top_k)), 0, True, True)

        boxes = boxes[order]
        # landms = landms[order]

        # do NMS
        # dets = np.hstack((boxes, scores[:, np.newaxis])).astype(np.float32, copy=False)
        # dets = soft_nms(dets, score_thresh=args.confidence_threshold)

        keep = nms(boxes, scores,args.nms_threshold)
        # keep top-K faster NMS

        boxes = boxes[keep]
        scores = scores[keep]

        dets = boxes[:args.keep_top_k, :]
        scores = scores[:args.keep_top_k]
        # landms = landms[:args.keep_top_k, :]

        # dets = np.concatenate((dets, landms), axis=1)
        print('net forward time: {:.4f}'.format(time.time() - tic))
        # show image
        if args.save_image:
            for index, b in enumerate(dets):
                if scores[index] < args.vis_thres:
                    continue
                text = "{:.4f}".format(scores[index])
                b = list(map(int, b))
                cv2.rectangle(img_raw, (b[0], b[1]), (b[2], b[3]), (0, 0, 255), 1)
                cx = b[0]
                cy = b[1] + 12
                cv2.putText(img_raw, text, (cx, cy),
                            cv2.FONT_HERSHEY_DUPLEX, 0.5, (255, 255, 255))

            # save image
            cv2.imshow("sdf",img_raw)
            cv2.waitKey()
            # name = "test.jpg"
            # cv2.imwrite(name, img_raw)