def compute_stats(self, confm_list, val_loss): TP_list, TN_list, FP_list, FN_list = extract_stats_from_confm( confm_list) mean_accuracy = compute_accuracy(TP_list, TN_list, FP_list, FN_list) self.stats.val.acc = np.nanmean(mean_accuracy) self.stats.val.loss = val_loss.avg
def compute_stats(self, confm_list, train_loss): TP_list, TN_list, FP_list, FN_list = extract_stats_from_confm( confm_list) mean_accuracy = compute_accuracy(TP_list, TN_list, FP_list, FN_list) self.stats.train.acc = np.nanmean(mean_accuracy) self.stats.train.loss = float(train_loss.avg.cpu().data)
def compute_stats(self, confm_list, val_loss): TP_list, TN_list, FP_list, FN_list = extract_stats_from_confm(confm_list) mean_IoU = compute_mIoU(TP_list, FP_list, FN_list) mean_accuracy = compute_accuracy_segmentation(TP_list, FN_list) self.stats.val.acc = np.nanmean(mean_accuracy) self.stats.val.mIoU_perclass = mean_IoU self.stats.val.mIoU = np.nanmean(mean_IoU) if val_loss is not None: self.stats.val.loss = val_loss.avg
def compute_stats(self, confm_list, train_loss): TP_list, TN_list, FP_list, FN_list = extract_stats_from_confm( confm_list) mean_accuracy = compute_accuracy(TP_list, TN_list, FP_list, FN_list) mean_precision = compute_precision(TP_list, FP_list) mean_recall = compute_recall(TP_list, FN_list) mean_f1score = compute_f1score(TP_list, FP_list, FN_list) self.stats.train.acc = np.nanmean(mean_accuracy) self.stats.train.recall = np.nanmean(mean_recall) self.stats.train.precision = np.nanmean(mean_precision) self.stats.train.f1score = np.nanmean(mean_f1score) if train_loss is not None: self.stats.train.loss = train_loss.avg
def compute_stats(self, confm_list, val_loss): TP_list, TN_list, FP_list, FN_list = extract_stats_from_confm( confm_list) mean_accuracy = compute_accuracy(TP_list, TN_list, FP_list, FN_list) mean_precision = compute_precision(TP_list, FP_list) mean_recall = compute_recall(TP_list, FN_list) # self.stats.val.acc = np.nanmean(mean_accuracy) self.stats.val.acc = np.sum(TP_list) / (np.sum(FP_list) + np.sum(TP_list)) self.stats.val.acc = np.nanmean(mean_accuracy) self.stats.val.recall = np.nanmean(mean_recall) self.stats.val.precision = np.nanmean(mean_precision) self.stats.val.f1score = np.nanmean(mean_f1score) if val_loss is not None: self.stats.val.loss = val_loss.avg
def compute_stats(self, confm_list, val_loss): TP_list, TN_list, FP_list, FN_list = extract_stats_from_confm(confm_list) tp = np.sum(TP_list) tn = np.sum(TN_list) fp = np.sum(FP_list) fn = np.sum(FN_list) r = tp / (tp + fn) p = tp / (tp + fp) self.stats.val.acc = (tp + tn) / (tp + tn + fp + fn) self.stats.val.recall = r self.stats.val.precision = p self.stats.val.f1score = 2 * (r * p) / (r + p) if val_loss is not None: self.stats.val.loss = val_loss.avg
def compute_stats(self, confm_list, val_loss): TP_list, TN_list, FP_list, FN_list = extract_stats_from_confm( confm_list) TP = np.sum(TP_list) TN = np.sum(TN_list) FP = np.sum(FP_list) FN = np.sum(FN_list) precision = TP / (TP + FP) recall = TP / (TP + FN) self.stats.val.acc = np.sum(TP_list) / (np.sum(FP_list) + np.sum(TP_list)) self.stats.val.recall = recall self.stats.val.precision = precision print("VAL results: accuracy", self.stats.val.acc, "precision", precision, "recall", recall) self.stats.val.f1score = 2 * (recall * precision) / (recall + precision) if val_loss is not None: self.stats.val.loss = val_loss.avg