Runtime: TensorFlow.js 4.22.0, unmodified npm tf.min.js, Apache-2.0.
Weights: official TensorFlow.js MobileNet v1 0.25 224px Layers model.
Source manifest: https://storage.googleapis.com/tfjs-models/tfjs/mobilenet_v1_0.25_224/model.json
All 55 manifest-declared weight shards downloaded unmodified from the same official source directory.
Labels: @tensorflow-models/mobilenet 2.1.1 ImageNet class dictionary, Apache-2.0.
Inference: resize to 224x224 RGB; normalize pixels to [-1,1]; pretrained 1000-class softmax output.
Output is subject recognition, never photographic-quality assessment, face identity, closed-eye detection or automatic culling.
Source: https://github.com/tensorflow/tfjs-models/tree/master/mobilenet