使用model_main.py

问题描述

我正在使用model_main.py来训练和评估tensorflow更快的rcnn incetion v2模型,但是评估结果始终为0

Model_main.py用作:

!python /content/models/research/object_detection/model_main.py --pipeline_config_path=/content/models/research/object_detection/samples/configs/demo.config --model_dir=/content/models/research/training --alsologtostderr

配置文件中的评估参数:

    eval_config: {
  num_examples: 30
  num_visualizations: 6
  metrics_set: "coco_detection_metrics"
}

eval_input_reader: {
  label_map_path: "/content/tensorflow-object-detection-faster-rcnn/data/eval/demo.pbtxt"
  shuffle: false
  num_readers: 1
  tf_record_input_reader {
    input_path: "/content/tensorflow-object-detection-faster-rcnn/data/eval/data.record"
  }
}

每次图像评估运行中...评估注释类型 bBox 完成 (t = 0.56s)。累积评估结果...完成(t = 0.07s)。平均 精度(AP)@ [IoU = 0.50:0.95 |面积=全部| maxDets = 100] = 0.000 平均精度(AP)@ [IoU = 0.50 |面积=全部| maxDets = 100] = 0.001平均精度(AP)@ [IoU = 0.75 |面积=全部| maxDets = 100] = 0.000平均精度(AP)@ [IoU = 0.50:0.95 | 面积=小| maxDets = 100] = 0.000平均精度(AP)@ [ IoU = 0.50:0.95 | area = medium | maxDets = 100] = 0.001平均精度 (AP)@ [IoU = 0.50:0.95 |面积=大| maxDets = 100] = -1.000平均值 召回(AR)@ [IoU = 0.50:0.95 |面积=全部| maxDets = 1] = 0.000 平均召回率(AR)@ [IoU = 0.50:0.95 |面积=全部| maxDets = 10] = 0.003平均召回率(AR)@ [IoU = 0.50:0.95 |面积=全部| maxDets = 100] = 0.015平均召回率(AR)@ [IoU = 0.50:0.95 | 面积=小| maxDets = 100] = 0.006平均召回率(AR)@ [ IoU = 0.50:0.95 | area = medium | maxDets = 100] = 0.031平均召回率
(AR)@ [IoU = 0.50:0.95 |面积=大| maxDets = 100] = -1.000 INFO:tensorflow:在2020-09-13-02:34:19 I0913完成的评估 02:34:19.032517 140132493051776 Evaluation.py:275]完成评估 在2020-09-13-02:34:19 INFO:tensorflow:为全局步骤保存命令 5000:DetectionBoxes_Precision / mAP = 0.00017041541, DetectionBoxes_Precision / mAP(大)= -1.0, DetectionBoxes_Precision / mAP(中)= 0.0011315575, DetectionBoxes_Precision / mAP(小)= 7.34418e-05, DetectionBoxes_Precision / mAP @ .50IoU = 0.0011445266, DetectionBoxes_Precision / mAP @ .75IoU = 2.8234763e-05, DetectionBoxes_Recall / AR @ 1 = 0.0,DetectionBoxes_Recall / AR @ 10 = 0.0031914893,DetectionBoxes_Recall / AR @ 100 = 0.0148936175,DetectionBoxes_Recall / AR @ 100(大)= -1.0, DetectionBoxes_Recall / AR @ 100(中)= 0.03125, DetectionBoxes_Recall / AR @ 100(小)= 0.006451613, 损失/ BoxClassifierLoss / classification_loss = 0.014203117, 损失/ BoxClassifierLoss / localization_loss = 0.0038770703, 损失/ rpnLoss / localization_loss = 0.042681895, 损失/ rpnLoss / objectness_loss = 0.093742944,损失/总损失= 0.15450501,global_step = 5000,learning_rate = 2e-06,损失= 0.15450501 I0913 02:34:19.032845 140132493051776 estimator.py:2049]保存全局步骤5000的字典:DetectionBoxes_Precision / mAP = 0.00017041541,DetectionBoxes_Precision / mAP(大)= -1.0,DetectionBoxes_Precision / mAP(中)= 0.0011315575, DetectionBoxes_Precision / mAP(小)= 7.34418e-05, DetectionBoxes_Precision / mAP @ .50IoU = 0.0011445266, DetectionBoxes_Precision / mAP @ .75IoU = 2.8234763e-05, DetectionBoxes_Recall / AR @ 1 = 0.0,DetectionBoxes_Recall / AR @ 10 = 0.0031914893,DetectionBoxes_Recall / AR @ 100 = 0.0148936175,DetectionBoxes_Recall / AR @ 100(大)= -1.0, DetectionBoxes_Recall / AR @ 100(中)= 0.03125, DetectionBoxes_Recall / AR @ 100(小)= 0.006451613, 损失/ BoxClassifierLoss / classification_loss = 0.014203117, 损失/ BoxClassifierLoss / localization_loss = 0.0038770703, 损失/ rpnLoss / localization_loss = 0.042681895, 损失/ rpnLoss / objectness_loss = 0.093742944,损失/总损失= 0.15450501,global_step = 5000,learning_rate = 2e-06,损失= 0.15450501

步数:5000 tensorflow版本1.15

解决方法

暂无找到可以解决该程序问题的有效方法,小编努力寻找整理中!

如果你已经找到好的解决方法,欢迎将解决方案带上本链接一起发送给小编。

小编邮箱:dio#foxmail.com (将#修改为@)