aiforge/templates/repo/modelsafety/indicators_descr.tmpl

<span>
  <i class="question circle icon indicators_descr" style="cursor:pointer;"></i>
  <div class="ui flowing popup bottom left transition hidden" style="inset: 564px auto auto 1px; width: 920px">
      <table class="ui celled table">
          <thead>
              <tr>
                  <th style="width:80px;">指标</th>
                  <th>说明</th>
              </tr>
          </thead>
          <tbody>
              <tr>
                  <td data-label="指标">ACC</td>
                  <td data-label="说明" indicator="ACC">Accuracy：精确度，计算模型预测准确率，该指标越高，说明评测结果越好。</td>
              </tr>
             <!-- <tr>
                  <td data-label="指标">CAV</td>
                  <td data-label="说明" indicator="CAV">Classification Accuracy Variance：用于评估深度学习模型性能的最重要指标为准确性。该指标值越高，说明评测结果越好。</td>
              </tr>  -->
              <tr>
                  <td data-label="指标">ASS</td>
                  <td data-label="说明" indicator="ASS">Average Structural Similarity：所有攻击成功对抗样本与其原始样本间的平均相似性。该指标越大，说明评测结果越好。</td>
              </tr>
              <tr>
                  <td data-label="指标">ALDp</td>
                  <td data-label="说明" indicator="ALDp">Average LpDistortion：为所有攻击成功的对抗样本的平均归一化Lp失真度，ALDp值越小，对抗样本的不可感知性越强。</td>
              </tr>
              
              <tr>
                  <td data-label="指标">ACAC</td>
                  <td data-label="说明" indicator="ACAC">Average Confidence of Adversarial Class(ACAC)：这个数值越高，攻击算法的攻击能力越强。</td>
              </tr> 
              <!--  <tr>
                  <td data-label="指标">ACTC</td>
                  <td data-label="说明" indicator="ACTC">Average Confidence of True Class(ACTC)：这个数值越低，攻击算法的攻击能力越强。</td>
              </tr>  -->
              <tr>
                  <td data-label="指标">PSD</td>
                  <td data-label="说明" indicator="PSD">Perturbation Sensitivity Distance：用于评测人类对扰动的感知能力。该指标值越大，说明评测结果越差。</td>
              </tr>
              <tr>
                  <td data-label="指标">CACC</td>
                  <td data-label="说明" indicator="CACC">Clean ACC：该值计算的是原始未被攻击的样本，使用模型和groundtruth相比较的模型的本身的一个。</td>
              </tr>
          </tbody>
      </table>
  </div>
</span>
<script>
  ;(function() {
    setTimeout(function(){
      $('.indicators_descr').popup({
          inline: true,
          hoverable: true,
          position: 'top center',
          delay: {
            show: 300,
            hide: 800
          }
      });
    }, 1000);
  })();
</script>