Checks all the ops till the desired outputs are reached. From these ops it collects all the ops that are unsupported. Error out if there is at

Creates an adjacency representation of the directed graph formed from the list of TF ops. Nodes are the ops. Directed edge from an op "A" to an

Get the right translator function

Get the right translator function

Infer CoreML input shape from TensorFlow shape.

Infer CoreML output shape from TensorFlow shape.

Detect Multiply or add layers after convolution and recast as Batchnorm layer so that it can be fused in the framework.

Removes from the CoreML NN graph all the layers that are not connected to the output nodes.

Fold load constants that interact through 'add', 'multiply', 'activation', 'slice', 'reduce' or 'unary' layers. In other words, evaluate any su

Find a reduce layer with mode = 'average'/'max' and axis = 'HW' and replace it with global average/max pooling layer.

input: list of TF ops output: list of TF ops, in topological sort order such that an op is encountered only after all the ops that generated it

Add pixelValue to byteBuffer. @param pixelValue

Inserts the operations we need to evaluate the accuracy of our results. Args: result_tensor: The new final node that produces results.

Adds a new softmax and fully-connected layer for training. We need to retrain the top layer to identify our new classes, so this function add

Creates the operations to apply the specified distortions. During training it can help to improve the results if we run the images through si

Adds operations that perform JPEG decoding and resizing to the graph.. Args: input_width: Desired width of the image fed into the recognize