# YOLOv5 🚀 by Ultralytics, AGPL-3.0 license import argparse import csv import os import platform import sys from pathlib import Path import glob import numpy as np import pickle import torch # ============================================================================= # FILE = Path(__file__).resolve() # ROOT = FILE.parents[0] # YOLOv5 root directory # if str(ROOT) not in sys.path: # sys.path.append(str(ROOT)) # add ROOT to PATH # ROOT = Path(os.path.relpath(ROOT, Path.cwd())) # relative # ============================================================================= sys.path.append('./ytracking') from models.common import DetectMultiBackend from utils.dataloaders import LoadImages from utils.general import (LOGGER, Profile, check_img_size, check_requirements, colorstr, cv2, increment_path, non_max_suppression, scale_boxes, strip_optimizer) from utils.torch_utils import select_device, smart_inference_mode '''集成跟踪模块,输出跟踪结果文件 .npy''' # from ultralytics.engine.results import Boxes # Results # from ultralytics.utils import IterableSimpleNamespace, yaml_load from tracking.utils.plotting import Annotator, colors from tracking.utils import Boxes, IterableSimpleNamespace, yaml_load from tracking.trackers import BOTSORT, BYTETracker from tracking.utils.showtrack import drawtracks # tracker_yaml = r"./tracking/trackers/cfg/botsort.yaml" def init_trackers(tracker_yaml=None, bs=1): """ Initialize trackers for object tracking during prediction. """ # tracker_yaml = r"./tracking/trackers/cfg/botsort.yaml" TRACKER_MAP = {'bytetrack': BYTETracker, 'botsort': BOTSORT} cfg = IterableSimpleNamespace(**yaml_load(tracker_yaml)) trackers = [] for _ in range(bs): tracker = TRACKER_MAP[cfg.tracker_type](args=cfg, frame_rate=30) trackers.append(tracker) return trackers @smart_inference_mode() def run( weights=r"D:/Project/ieemoo-ai/tools/ckpts/best_158734_cls11_noaug10.pt", # model path or triton URL source=r"D:/Project/ieemoo-ai/testdata/88.mp4", # file/dir/URL/glob/screen/0(webcam) project=r'./runs/detect', # save results to project/name name='exp', # save results to project/name tracker_yaml="D:/Project/ieemoo-ai/ytracking/tracking/trackers/cfg/botsort.yaml", imgsz=(640, 640), # inference size (height, width) conf_thres=0.25, # confidence threshold iou_thres=0.45, # NMS IOU threshold max_det=1000, # maximum detections per image device='', # cuda device, i.e. 0 or 0,1,2,3 or cpu bs=1, # batch_size save_txt=False, # save results to *.txt save_img=True, # do not save images/videos classes=None, # filter by class: --class 0, or --class 0 2 3 agnostic_nms=False, # class-agnostic NMS augment=False, # augmented inference visualize=False, # visualize features update=False, # update all models exist_ok=False, # existing project/name ok, do not increment line_thickness=3, # bounding box thickness (pixels) hide_labels=False, # hide labels hide_conf=False, # hide confidencesL half=False, # use FP16 half-precision inference dnn=False, # use OpenCV DNN for ONNX inference vid_stride=1, # video frame-rate stride ): # Load model device = select_device(device) model = DetectMultiBackend(weights, device=device, dnn=dnn, fp16=half) stride, names, pt = model.stride, model.names, model.pt imgsz = check_img_size(imgsz, s=stride) # check image size # Run inference model.warmup(imgsz=(1 if pt or model.triton else bs, 3, *imgsz)) # warmup ##=============================生成文件夹 save_dir,存储检测跟踪图像 source = str(source) save_dir = Path(project) / Path(source).stem if save_dir.exists(): print(Path(source).stem) # return save_dir = increment_path(Path(project) / name, exist_ok=exist_ok) # increment run (save_dir / 'labels' if save_txt else save_dir).mkdir(parents=True, exist_ok=True) # make dir else: save_dir.mkdir(parents=True, exist_ok=True) # Dataloader seen, dt = 0, (Profile(), Profile(), Profile()) dataset = LoadImages(source, img_size=imgsz, stride=stride, auto=pt, vid_stride=vid_stride) ## ================================================= 生成跟踪器对象 tracker = init_trackers(tracker_yaml, bs)[0] track_boxes = np.empty((0, 9), dtype=np.float32) features_dict = {} for path, im, im0s, vid_cap, s in dataset: # img preprocess with dt[0]: im = torch.from_numpy(im).to(model.device) im = im.half() if model.fp16 else im.float() # uint8 to fp16/32 im /= 255 # 0 - 255 to 0.0 - 1.0 if len(im.shape) == 3: im = im[None] # expand for batch dim # Inference with dt[1]: visualize = increment_path(save_dir / Path(path).stem, mkdir=True) if visualize else False pred = model(im, augment=augment, visualize=visualize) # NMS with dt[2]: pred = non_max_suppression(pred, conf_thres, iou_thres, classes, agnostic_nms, max_det=max_det) # Process predictions for i, det in enumerate(pred): # per image seen += 1 p, im0, frame = path, im0s.copy(), getattr(dataset, 'frame', 0) im0_ant = im0.copy() p = Path(p) # to Path save_path = str(save_dir / p.name) # im.jpg s += '%gx%g ' % im.shape[2:] # print string annotator = Annotator(im0_ant, line_width=line_thickness, example=str(names)) if save_img else None if len(det): # Rescale boxes from img_size to im0 size det[:, :4] = scale_boxes(im.shape[2:], det[:, :4], im0.shape).round() # boxes_and_imgs.append((det.cpu().numpy(), im0, frame)) ## ================================================================ writed by WQG det_tracking = Boxes(det, im0.shape).cpu().numpy() tracks = tracker.update(det_tracking, im0) if len(tracks): track_boxes = np.concatenate([track_boxes, tracks], axis=0) feat_dict = {int(x.idx): x.curr_feat for x in tracker.tracked_stracks if x.is_activated} frame_id = track_boxes[0, 7] features_dict.update({int(frame_id): feat_dict}) if annotator is not None: for *xyxy, id, conf, cls, fid, bid in reversed(tracks): name = ('' if id == -1 else f'id:{int(id)} ') + names[int(cls)] label = None if hide_labels else (name if hide_conf else f'{name} {conf:.2f}') if id >= 0 and cls == 0: color = colors(int(cls), True) elif id >= 0 and cls != 0: color = colors(int(id), True) else: color = colors(19, True) # 19为调色板的最后一个元素 annotator.box_label(xyxy, label, color=color) # Save tracking image if annotator is not None: save_path_img, ext = os.path.splitext(save_path) imgpath = save_path_img + f"_{dataset.frame}.png" cv2.imwrite(Path(imgpath), annotator.result()) # Print time (inference-only) LOGGER.info(f"{s}{'' if len(det) else '(no detections), '}{dt[1].dt * 1E3:.1f}ms") ## ======================================================================== written by WQG ''' track_boxes: Array, [x1, y1, x2, y2, track_id, score, cls, frame_index, box_id] ''' if save_img: filename = os.path.split(save_path_img)[-1] '''====== save in './run/detect/' ======''' imgshow = drawtracks(track_boxes) showpath_1 = save_path_img + "_show.png" cv2.imwrite(Path(showpath_1), imgshow) '''====== save tracks data ======''' tracks_dir = Path('D:/Project/ieemoo-ai/ytracking/tracking/tracking/data/tracks/') if not tracks_dir.exists(): tracks_dir.mkdir(parents=True, exist_ok=True) tracks_path = tracks_dir.joinpath(filename + ".npy") np.save(tracks_path, track_boxes) '''====== save reid features data ======''' feats_dir = Path('D:/Project/ieemoo-ai/ytracking/tracking/data/trackfeats/') if not feats_dir.exists(): feats_dir.mkdir(parents=True, exist_ok=True) feats_path = feats_dir.joinpath(f'{filename}.pkl') with open(feats_path, 'wb') as file: pickle.dump(features_dict, file) # Print results t = tuple(x.t / seen * 1E3 for x in dt) # speeds per image LOGGER.info(f'Speed: %.1fms pre-process, %.1fms inference, %.1fms NMS per image at shape {(1, 3, *imgsz)}' % t) if save_txt or save_img: s = f"\n{len(list(save_dir.glob('labels/*.txt')))} labels saved to {save_dir / 'labels'}" if save_txt else '' LOGGER.info(f"Results saved to {colorstr('bold', save_dir)}{s}") if update: strip_optimizer(weights[0]) # update model (to fix SourceChangeWarning) def main(): ROOT = Path(Path.cwd()) check_requirements(ROOT / 'requirements.txt', exclude=('tensorboard', 'thop')) optdict = {'weights': r"D:/Project/ieemoo-ai/tools/ckpts/best_158734_cls11_noaug10.pt", 'source': r"D:/Project/ieemoo-ai/testdata/88.mp4", } run(**optdict) if __name__ == '__main__': main()