|
it
Pav = 1/N Σ Pi, Disp = 1/N Σ(Pi - Pav)2, Error = (Disp/N)1/2. |
"Daisy" Dataset demo. Press "iterate" to collect 1000 daisy-class (985) probabilities. See console for more detailes. CNN "GraphModels" from tfhub.dev are used.
CNN Pav _____________ v2_0.35 0.29 v2_0.50 0.54 v1_1.00 0.74 v1_1.00* 0.48 v2_1.00 0.47 v2_1.40 0.56 ResNet 0.74* I get Pav = 0.48 for the same v1_1_224 "LayersModel" at 'https://storage.googleapis.com/tfjs-models/tfjs/mobilenet_v1_1.0_224/model.json'. See "Daisy" weight quantization test.
Unfortunately "daisy"-classification test is not perfect but simple and fast :)
The only daisy is not representative?