I am using Inception-ResNet-v2 pretrained version to classify the images. I need human-readable class labels for this. I found one in following site: https://gist.github.com/yrevar/942d3a0ac09ec9e5eb3a.
However, when I try to validate these labels with the images, I find it doesn't map to correct labels. One such instance is I tried to classify "Panda" image- the class label it matches is : "barracouta, snoek" with score - 0.927924 and "giant panda, panda, panda bear, coon bear, Ailuropoda melanoleuca" with score - 0.001053.
Please provide me a source where I can find correct mappings of class label to human-readable text for this model.
https://gist.github.com/yrevar/942d3a0ac09ec9e5eb3a this human readable label works, but initialize the class label list with an "unused background", before loading these labels, as inception resnet v2 model is trained for 1001 classes and this has 1000.