Posts

Implementing ResNet with Base Line Accuracy

  Implementing ResNet with Base Line Accuracy    ResNet Base Line Accuracies   https://keras.io/api/applications/     Model   Size (MB)   Top-1 Accuracy   Top-5 Accuracy   Parameters   Depth   Time ( ms ) per inference step (CPU)   Time ( ms ) per inference step (GPU)   ResNet50   98   0.749   0.921   25,636,712   -   58.20   4.55   ResNet101   171   0.764   0.928   44,707,176   -   89.59   5.19     Results Summary & Comparison (Reasons and the Implementations described below)   Model   Size (MB)   Base Line Top-1 Accuracy   Actual Base Line Top 1 Accuracy   ResNet50   98   74.9 %   68.084 %   ResNet101   171   76.4 %   69.976 %     ResNet 50 Validation Test Results   Data Set: ImageNet Validation data set (50,000 images)    Validation Accuracy (Top1) = 68.084 % with caffe input preprocessing     ResNet 101 Validation Test Results   Data Set: ImageNet Validation data set (50,000 images)   Validation Accuracy (Top1) = 69.976 % with caffe input preprocessing     Base Line Setup Deta