Maximef/Yologo

1import argparse2import time3from pathlib import Path4 5import cv26import torch7import torch.backends.cudnn as cudnn8from numpy import random9 10from models.experimental import attempt_load11from utils.datasets import LoadStreams, LoadImages12from utils.general import check_img_size, check_requirements, check_imshow, non_max_suppression, apply_classifier, \13    scale_coords, xyxy2xywh, strip_optimizer, set_logging, increment_path14from utils.plots import plot_one_box15from utils.torch_utils import select_device, load_classifier, time_synchronized, TracedModel16from recommendation import SimilarityRecommender17 18 19def detect(save_img=False):20    recommender = SimilarityRecommender("./TopBrands.xlsx")21    source, weights, view_img, save_txt, imgsz, trace = opt.source, opt.weights, opt.view_img, opt.save_txt, opt.img_size, not opt.no_trace22    save_img = not opt.nosave and not source.endswith('.txt')  # save inference images23    webcam = source.isnumeric() or source.endswith('.txt') or source.lower().startswith(24        ('rtsp://', 'rtmp://', 'http://', 'https://'))25 26    # Directories27    save_dir = Path(increment_path(Path(opt.project) / opt.name, exist_ok=opt.exist_ok))  # increment run28    (save_dir / 'labels' if save_txt else save_dir).mkdir(parents=True, exist_ok=True)  # make dir29 30    # Initialize31    set_logging()32    device = select_device(opt.device)33    half = device.type != 'cpu'  # half precision only supported on CUDA34 35    # Load model36    model = attempt_load(weights, map_location=device)  # load FP32 model37    stride = int(model.stride.max())  # model stride38    imgsz = check_img_size(imgsz, s=stride)  # check img_size39 40    if trace:41        model = TracedModel(model, device, opt.img_size)42 43    if half:44        model.half()  # to FP1645 46    # Second-stage classifier47    classify = False48    if classify:49        modelc = load_classifier(name='resnet101', n=2)  # initialize50        modelc.load_state_dict(torch.load('weights/resnet101.pt', map_location=device)['model']).to(device).eval()51 52    # Set Dataloader53    vid_path, vid_writer = None, None54    if webcam:55        view_img = check_imshow()56        cudnn.benchmark = True  # set True to speed up constant image size inference57        dataset = LoadStreams(source, img_size=imgsz, stride=stride)58    else:59        dataset = LoadImages(source, img_size=imgsz, stride=stride)60 61    # Get names and colors62    names = model.module.names if hasattr(model, 'module') else model.names63    colors = [[random.randint(0, 255) for _ in range(3)] for _ in names]64 65    # Run inference66    if device.type != 'cpu':67        model(torch.zeros(1, 3, imgsz, imgsz).to(device).type_as(next(model.parameters())))  # run once68    old_img_w = old_img_h = imgsz69    old_img_b = 170 71    t0 = time.time()72    for path, img, im0s, vid_cap in dataset:73        img = torch.from_numpy(img).to(device)74        img = img.half() if half else img.float()  # uint8 to fp16/3275        img /= 255.0  # 0 - 255 to 0.0 - 1.076        if img.ndimension() == 3:77            img = img.unsqueeze(0)78 79        # Warmup80        if device.type != 'cpu' and (old_img_b != img.shape[0] or old_img_h != img.shape[2] or old_img_w != img.shape[3]):81            old_img_b = img.shape[0]82            old_img_h = img.shape[2]83            old_img_w = img.shape[3]84            for i in range(3):85                model(img, augment=opt.augment)[0]86 87        # Inference88        t1 = time_synchronized()89        pred = model(img, augment=opt.augment)[0]90        t2 = time_synchronized()91 92        # Apply NMS93        pred = non_max_suppression(pred, opt.conf_thres, opt.iou_thres, classes=opt.classes, agnostic=opt.agnostic_nms)94        t3 = time_synchronized()95 96        # Apply Classifier97        if classify:98            pred = apply_classifier(pred, modelc, img, im0s)99 100        # Process detections101        for i, det in enumerate(pred):  # detections per image102            if webcam:  # batch_size >= 1103                p, s, im0, frame = path[i], '%g: ' % i, im0s[i].copy(), dataset.count104            else:105                p, s, im0, frame = path, '', im0s, getattr(dataset, 'frame', 0)106 107            p = Path(p)  # to Path108            save_path = str(save_dir / p.name)  # img.jpg109            txt_path = str(save_dir / 'labels' / p.stem) + ('' if dataset.mode == 'image' else f'_{frame}')  # img.txt110            gn = torch.tensor(im0.shape)[[1, 0, 1, 0]]  # normalization gain whwh111            if len(det):112                # Rescale boxes from img_size to im0 size113                det[:, :4] = scale_coords(img.shape[2:], det[:, :4], im0.shape).round()114 115                # Print results116                for c in det[:, -1].unique():117                    n = (det[:, -1] == c).sum()  # detections per class118                    s += f"{n} {names[int(c)]}{'s' * (n > 1)}, "  # add to string119 120                # Write results121                for *xyxy, conf, cls in reversed(det):122                    name_class = names[int(cls.item())]123                    name_recommanded = recommender.make_recommendation(name_class)124                    if save_txt:  # Write to file125                        xywh = (xyxy2xywh(torch.tensor(xyxy).view(1, 4)) / gn).view(-1).tolist()  # normalized xywh126                        line = (cls, *xywh, conf) if opt.save_conf else (cls, *xywh)  # label format127                        with open(txt_path + '.txt', 'a') as f:128                            f.write(('%g ' * len(line)).rstrip() % line + '\n')129 130                    if save_img or view_img:  # Add bbox to image131                        label = f'{names[int(cls)]} {conf:.2f} || Recommendation : {name_recommanded.lower()}'132                        plot_one_box(xyxy, im0, label=label, color=colors[int(cls)], line_thickness=1)133 134            # Print time (inference + NMS)135            print(f'{s}Done. ({(1E3 * (t2 - t1)):.1f}ms) Inference, ({(1E3 * (t3 - t2)):.1f}ms) NMS')136 137            # Stream results138            if view_img:139                cv2.imshow(str(p), im0)140                cv2.waitKey(1)  # 1 millisecond141 142            # Save results (image with detections)143            if save_img:144                if dataset.mode == 'image':145                    cv2.imwrite(save_path, im0)146                    print(f" The image with the result is saved in: {save_path}")147                else:  # 'video' or 'stream'148                    if vid_path != save_path:  # new video149                        vid_path = save_path150                        if isinstance(vid_writer, cv2.VideoWriter):151                            vid_writer.release()  # release previous video writer152                        if vid_cap:  # video153                            fps = vid_cap.get(cv2.CAP_PROP_FPS)154                            w = int(vid_cap.get(cv2.CAP_PROP_FRAME_WIDTH))155                            h = int(vid_cap.get(cv2.CAP_PROP_FRAME_HEIGHT))156                        else:  # stream157                            fps, w, h = 30, im0.shape[1], im0.shape[0]158                            save_path += '.mp4'159                        vid_writer = cv2.VideoWriter(save_path, cv2.VideoWriter_fourcc(*'mp4v'), fps, (w, h))160                    vid_writer.write(im0)161 162    if save_txt or save_img:163        s = f"\n{len(list(save_dir.glob('labels/*.txt')))} labels saved to {save_dir / 'labels'}" if save_txt else ''164        #print(f"Results saved to {save_dir}{s}")165 166    print(f'Done. ({time.time() - t0:.3f}s)')167 168 169if __name__ == '__main__':170    parser = argparse.ArgumentParser()171    parser.add_argument('--weights', nargs='+', type=str, default='yolov7.pt', help='model.pt path(s)')172    parser.add_argument('--source', type=str, default='inference/images', help='source')  # file/folder, 0 for webcam173    parser.add_argument('--img-size', type=int, default=640, help='inference size (pixels)')174    parser.add_argument('--conf-thres', type=float, default=0.25, help='object confidence threshold')175    parser.add_argument('--iou-thres', type=float, default=0.45, help='IOU threshold for NMS')176    parser.add_argument('--device', default='', help='cuda device, i.e. 0 or 0,1,2,3 or cpu')177    parser.add_argument('--view-img', action='store_true', help='display results')178    parser.add_argument('--save-txt', action='store_true', help='save results to *.txt')179    parser.add_argument('--save-conf', action='store_true', help='save confidences in --save-txt labels')180    parser.add_argument('--nosave', action='store_true', help='do not save images/videos')181    parser.add_argument('--classes', nargs='+', type=int, help='filter by class: --class 0, or --class 0 2 3')182    parser.add_argument('--agnostic-nms', action='store_true', help='class-agnostic NMS')183    parser.add_argument('--augment', action='store_true', help='augmented inference')184    parser.add_argument('--update', action='store_true', help='update all models')185    parser.add_argument('--project', default='runs/detect', help='save results to project/name')186    parser.add_argument('--name', default='exp', help='save results to project/name')187    parser.add_argument('--exist-ok', action='store_true', help='existing project/name ok, do not increment')188    parser.add_argument('--no-trace', action='store_true', help='don`t trace model')189    opt = parser.parse_args()190    print(opt)191    #check_requirements(exclude=('pycocotools', 'thop'))192 193    with torch.no_grad():194        if opt.update:  # update all models (to fix SourceChangeWarning)195            for opt.weights in ['yolov7.pt']:196                detect()197                strip_optimizer(opt.weights)198        else:199            detect()200