How to load model and make predictions on new data?

[laag115]

If I download a new song that is not in the Dataset, how could I see what genre it is using the Model.final output file?

[pl425]

??

[eggfly]

Use preproccess.py to load a mp3 or .au etc.. to mfcc or mel-spec data, then start a tensorflow session, put some code like this: saver.restore(sess, "../output/model.final") xs, ys = /* your preproccessed data */ acc = sess.run(accuracy, feed_dict={x: xs, y: ys, keep_prob: 1.}) loss = sess.run(cost, feed_dict={x: xs, y: ys, keep_prob: 1.}) print("Minibatch Loss= " + \"{:.6f}".format(loss) + ", Accuracy= " + "{:.5f}".format(acc))

when accuracy is 1 then model prediction is right, otherwise accuracy is zero.

[jqzhao7]

I'm sorry to disturb you. Since the link of the data set of this project in the current year is invalid and the author cannot be contacted, I can only ask you in this way. If you still have the data set of this project, could you please send it to me [email protected]