關(guān)于用yolov5進(jìn)行驗(yàn)證集的map計(jì)算

1、yolov5map計(jì)算import timeimport torch.backends.cudnn as cudnnfrom numpy import randomcheck_img_size, non_max_suppression, apply_classifier, scale_coords,xyxy2xywh, plot_one_box, strip_optimizer, set_logging)from utils.torch_utils import select_device, load_classifier, time_synchronizedfrom cfg_mAP impo

2、rt Cfgout, source, weights, view_img, save_txt, imgsz = # Initializeset_logging()device = select_device(opt.device)if os.path.exists(out):# Load modelmodel = attempt_load(weights, map_location=device) # load FP32 modelimgsz = check_img_size(imgsz, s=model.stride.max() # check img_sizeif half:model.h

3、alf() # to FP16# Second-stage classifierclassify = Falseif classify:# Set Dataloadervid_path, vid_writer = None, Noneif webcam:img = torch.zeros(1, 3, imgsz, imgsz), device=device) # init img_ = model(img.half() if half else img) if device.type != cpu else None # run oncetest_time=for path, img, im0

4、s, vid_cap in dataset:pred = apply_classifier(pred, modelc, img, im0s)# Process detectionsfor i, det in enumerate(pred): # detections per imageif webcam: # batch_size = 1p, s, im0 = pathi, %g: % i, im0si.copy()else:p, s, im0 = path, , im0simg_name = Path(p).nametxt = open(opt.eval_imgs_name_txt, a)t

5、xt.write(img_name:-4)txt.write(n)txt.close()save_path = str(Path(out) / Path(p).name)gn = torch.tensor(im0.shape)1, 0, 1, 0 # normalization gain whwhif det is not None and len(det):# Rescale boxes from img_size to im0 sizedet:, :4 = scale_coords(img.shape2:, det:, :4, im0.shape).round()# Print resul

6、tsfor c in det:, -1.unique():for *xyxy, conf, cls in reversed(det):txt = open(opt.eval_classtxt_path + /%s % namesint(cls), a)obj_conf = conf.cpu().numpy()txt.write( .join(str(a) for a in new_box)f.write(%g * 5 + n) % (cls, *xywh) # label formatif save_img or view_img: # Add bbox to imageplot_one_bo

7、x(xyxy, im0, label=label, color=colorsint(cls), line_thickness=3)# Save results (image with detections)if save_img:if dataset.mode = images:cv2.imwrite(save_path, im0)else:if vid_path != save_path: # new videovid_path = save_pathfourcc = mp4v # output video codecfps = vid_cap.get(cv2.CAP_PROP_FPS)if

8、 save_txt or save_img:print(Results saved to %s % Path(out)if platform.system() = Darwin and not opt.update: # MacOSos.system(open + save_path)print(Done. (%.3fs) % (time.time() - t0)mean_time=sum(test_time)/len(test_time)print(mean time:, mean_time)print(frame: , 1/mean_time)dir = ./data_test/imgs_

9、name_manual.txtif os.path.exists(predictions_manual):if os.path.exists(class_txt_manual):shutil.rmtree(class_txt_manual) # delete output folderos.makedirs(class_txt_manual) # make new output folderif os.path.exists(cachedir_manual):shutil.rmtree(cachedir_manual) # delete output folderos.makedirs(cac

10、hedir_manual) # make new output folderparser = argparse.ArgumentParser()parser.add_argument(-output, type=str, default=./data_test/predictions_manual,help=output folder) # output folderprint(opt)with torch.no_grad():if opt.update: # update all models (to fix SourceChangeWarning)for opt.weights in yolov5s.pt, yolov5m.pt, yolov5l.pt, yolov5x.pt:detect()strip_optimizer(opt.weights)else:detect()關(guān)于這段代碼的操作運(yùn)會(huì)有個(gè)UnicodeDecodeError:utf-