import os from pathlib import Path import hydra import numpy as np import torch from ultralytics.yolo.data import build_dataloader from ultralytics.yolo.data.dataloaders.v5loader import create_dataloader from ultralytics.yolo.engine.validator import BaseValidator from ultralytics.yolo.utils import DEFAULT_CONFIG, colorstr, ops, yaml_load from ultralytics.yolo.utils.checks import check_file, check_requirements from ultralytics.yolo.utils.metrics import ConfusionMatrix, DetMetrics, box_iou from ultralytics.yolo.utils.plotting import output_to_target, plot_images from ultralytics.yolo.utils.torch_utils import de_parallel class DetectionValidator(BaseValidator): def __init__(self, dataloader=None, save_dir=None, pbar=None, logger=None, args=None): super().__init__(dataloader, save_dir, pbar, logger, args) self.data_dict = yaml_load(check_file(self.args.data)) if self.args.data else None self.is_coco = False self.class_map = None self.targets = None self.metrics = DetMetrics(save_dir=self.save_dir, plot=self.args.plots) self.iouv = torch.linspace(0.5, 0.95, 10) # iou vector for mAP@0.5:0.95 self.niou = self.iouv.numel() def preprocess(self, batch): batch["img"] = batch["img"].to(self.device, non_blocking=True) batch["img"] = (batch["img"].half() if self.args.half else batch["img"].float()) / 255 self.nb, _, self.height, self.width = batch["img"].shape # batch size, channels, height, width self.targets = torch.cat((batch["batch_idx"].view(-1, 1), batch["cls"].view(-1, 1), batch["bboxes"]), 1) self.targets = self.targets.to(self.device) height, width = batch["img"].shape[2:] self.targets[:, 2:] *= torch.tensor((width, height, width, height), device=self.device) # to pixels self.lb = [self.targets[self.targets[:, 0] == i, 1:] for i in range(self.nb)] if self.args.save_hybrid else [] # for autolabelling return batch def init_metrics(self, model): head = model.model[-1] if self.training else model.model.model[-1] self.is_coco = self.data.get('val', '').endswith(f'coco{os.sep}val2017.txt') # is COCO dataset self.class_map = ops.coco80_to_coco91_class() if self.is_coco else list(range(1000)) self.args.save_json |= self.is_coco and not self.training # run on final val if training COCO self.nc = head.nc self.names = model.names self.metrics.names = self.names self.confusion_matrix = ConfusionMatrix(nc=self.nc) self.seen = 0 self.jdict = [] self.stats = [] def get_desc(self): return ('%22s' + '%11s' * 6) % ('Class', 'Images', 'Instances', 'Box(P', "R", "mAP50", "mAP50-95)") def postprocess(self, preds): preds = ops.non_max_suppression(preds, self.args.conf_thres, self.args.iou_thres, labels=self.lb, multi_label=True, agnostic=self.args.single_cls, max_det=self.args.max_det) return preds def update_metrics(self, preds, batch): # Metrics for si, pred in enumerate(preds): labels = self.targets[self.targets[:, 0] == si, 1:] nl, npr = labels.shape[0], pred.shape[0] # number of labels, predictions shape = batch["ori_shape"][si] # path = batch["shape"][si][0] correct_bboxes = torch.zeros(npr, self.niou, dtype=torch.bool, device=self.device) # init self.seen += 1 if npr == 0: if nl: self.stats.append((correct_bboxes, *torch.zeros((2, 0), device=self.device), labels[:, 0])) if self.args.plots: self.confusion_matrix.process_batch(detections=None, labels=labels[:, 0]) continue # Predictions if self.args.single_cls: pred[:, 5] = 0 predn = pred.clone() ops.scale_boxes(batch["img"][si].shape[1:], predn[:, :4], shape) # native-space pred # Evaluate if nl: tbox = ops.xywh2xyxy(labels[:, 1:5]) # target boxes ops.scale_boxes(batch["img"][si].shape[1:], tbox, shape) # native-space labels labelsn = torch.cat((labels[:, 0:1], tbox), 1) # native-space labels correct_bboxes = self._process_batch(predn, labelsn) # TODO: maybe remove these `self.` arguments as they already are member variable if self.args.plots: self.confusion_matrix.process_batch(predn, labelsn) self.stats.append((correct_bboxes, pred[:, 4], pred[:, 5], labels[:, 0])) # (conf, pcls, tcls) # Save if self.args.save_json: self.pred_to_json(predn, batch["im_file"][si]) # if self.args.save_txt: # save_one_txt(predn, save_conf, shape, file=save_dir / 'labels' / f'{path.stem}.txt') def get_stats(self): stats = [torch.cat(x, 0).cpu().numpy() for x in zip(*self.stats)] # to numpy if len(stats) and stats[0].any(): self.metrics.process(*stats) self.nt_per_class = np.bincount(stats[-1].astype(int), minlength=self.nc) # number of targets per class fitness = {"fitness": self.metrics.fitness()} metrics = dict(zip(self.metric_keys, self.metrics.mean_results())) return {**metrics, **fitness} def print_results(self): pf = '%22s' + '%11i' * 2 + '%11.3g' * len(self.metric_keys) # print format self.logger.info(pf % ("all", self.seen, self.nt_per_class.sum(), *self.metrics.mean_results())) if self.nt_per_class.sum() == 0: self.logger.warning( f'WARNING ⚠️ no labels found in {self.args.task} set, can not compute metrics without labels') # Print results per class if (self.args.verbose or not self.training) and self.nc > 1 and len(self.stats): for i, c in enumerate(self.metrics.ap_class_index): self.logger.info(pf % (self.names[c], self.seen, self.nt_per_class[c], *self.metrics.class_result(i))) if self.args.plots: self.confusion_matrix.plot(save_dir=self.save_dir, names=list(self.names.values())) def _process_batch(self, detections, labels): """ Return correct prediction matrix Arguments: detections (array[N, 6]), x1, y1, x2, y2, conf, class labels (array[M, 5]), class, x1, y1, x2, y2 Returns: correct (array[N, 10]), for 10 IoU levels """ iou = box_iou(labels[:, 1:], detections[:, :4]) correct = np.zeros((detections.shape[0], self.iouv.shape[0])).astype(bool) correct_class = labels[:, 0:1] == detections[:, 5] for i in range(len(self.iouv)): x = torch.where((iou >= self.iouv[i]) & correct_class) # IoU > threshold and classes match if x[0].shape[0]: matches = torch.cat((torch.stack(x, 1), iou[x[0], x[1]][:, None]), 1).cpu().numpy() # [label, detect, iou] if x[0].shape[0] > 1: matches = matches[matches[:, 2].argsort()[::-1]] matches = matches[np.unique(matches[:, 1], return_index=True)[1]] # matches = matches[matches[:, 2].argsort()[::-1]] matches = matches[np.unique(matches[:, 0], return_index=True)[1]] correct[matches[:, 1].astype(int), i] = True return torch.tensor(correct, dtype=torch.bool, device=detections.device) def get_dataloader(self, dataset_path, batch_size): # TODO: manage splits differently # calculate stride - check if model is initialized gs = max(int(de_parallel(self.model).stride if self.model else 0), 32) return create_dataloader(path=dataset_path, imgsz=self.args.imgsz, batch_size=batch_size, stride=gs, hyp=dict(self.args), cache=False, pad=0.5, rect=self.args.rect, workers=self.args.workers, prefix=colorstr(f'{self.args.mode}: '), shuffle=False, seed=self.args.seed)[0] if self.args.v5loader else \ build_dataloader(self.args, batch_size, img_path=dataset_path, stride=gs, mode="val")[0] # TODO: align with train loss metrics @property def metric_keys(self): return ["metrics/precision(B)", "metrics/recall(B)", "metrics/mAP50(B)", "metrics/mAP50-95(B)"] def plot_val_samples(self, batch, ni): plot_images(batch["img"], batch["batch_idx"], batch["cls"].squeeze(-1), batch["bboxes"], paths=batch["im_file"], fname=self.save_dir / f"val_batch{ni}_labels.jpg", names=self.names) def plot_predictions(self, batch, preds, ni): plot_images(batch["img"], *output_to_target(preds, max_det=15), paths=batch["im_file"], fname=self.save_dir / f'val_batch{ni}_pred.jpg', names=self.names) # pred def pred_to_json(self, predn, filename): stem = Path(filename).stem image_id = int(stem) if stem.isnumeric() else stem box = ops.xyxy2xywh(predn[:, :4]) # xywh box[:, :2] -= box[:, 2:] / 2 # xy center to top-left corner for p, b in zip(predn.tolist(), box.tolist()): self.jdict.append({ 'image_id': image_id, 'category_id': self.class_map[int(p[5])], 'bbox': [round(x, 3) for x in b], 'score': round(p[4], 5)}) def eval_json(self, stats): if self.args.save_json and self.is_coco and len(self.jdict): anno_json = self.data['path'] / "annotations/instances_val2017.json" # annotations pred_json = self.save_dir / "predictions.json" # predictions self.logger.info(f'\nEvaluating pycocotools mAP using {pred_json} and {anno_json}...') try: # https://github.com/cocodataset/cocoapi/blob/master/PythonAPI/pycocoEvalDemo.ipynb check_requirements('pycocotools>=2.0.6') from pycocotools.coco import COCO # noqa from pycocotools.cocoeval import COCOeval # noqa for x in anno_json, pred_json: assert x.is_file(), f"{x} file not found" anno = COCO(str(anno_json)) # init annotations api pred = anno.loadRes(str(pred_json)) # init predictions api (must pass string, not Path) eval = COCOeval(anno, pred, 'bbox') if self.is_coco: eval.params.imgIds = [int(Path(x).stem) for x in self.dataloader.dataset.im_files] # images to eval eval.evaluate() eval.accumulate() eval.summarize() stats[self.metric_keys[-1]], stats[self.metric_keys[-2]] = eval.stats[:2] # update mAP50-95 and mAP50 except Exception as e: self.logger.warning(f'pycocotools unable to run: {e}') return stats @hydra.main(version_base=None, config_path=str(DEFAULT_CONFIG.parent), config_name=DEFAULT_CONFIG.name) def val(cfg): cfg.data = cfg.data or "coco128.yaml" validator = DetectionValidator(args=cfg) validator(model=cfg.model) if __name__ == "__main__": val()