now that I have finished classifying my photo collection using Alexnet network, I now see that the function of embedding in tensorBoard feels very cool, and I especially hope to see the effect of my network classification through that kind of visual interface. But I really can"t do it about embedding. I hope God can give me some advice!

in my network, first save the picture set to a .npy file, and then grab it randomly in the following training process. It is not very clear where embedding should be loaded in the code. I want to save my original data, that is, pictures, into embedding to see the effect, but there has been a problem when loading. I hope God can help me!

the code is as follows:

def alexnet_main():
    loopNum = 5
    -sharp 
    files = np.load("label.npy", encoding="bytes")[()]

    -sharp-embedding
    log_dir = "model"
    metadata = os.path.join(log_dir,"metadata.tsv")
    j = 0
    with open(metadata, "w") as metadata_file:
        for i in files:
            metadata_file.write("%d\n" % j)
            j = j+1


    -sharp 
    keys = [i for i in files]


    myinput = tf.placeholder(dtype=tf.float32, shape=[None, 224, 224, 3], name="input")
    mylabel = tf.placeholder(dtype=tf.float32, shape=[None, 10], name="label")
    
    -sharp keepprob0.6
    myoutput = alexnet(myinput, 0.6)

    -sharp loss
    loss = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=myoutput, labels=mylabel))

    -sharp 0.09
    optimizer = tf.train.GradientDescentOptimizer(learning_rate=0.09).minimize(loss)

    -sharp 
    valaccuracy = tf.reduce_mean(
        tf.cast(
            tf.equal(
                tf.argmax(myoutput, 1),
                tf.argmax(mylabel, 1)),
            tf.float32))

    -sharp tensorflowsaver
    saver = tf.train.Saver()
    init = tf.global_variables_initializer()
    all_vars = tf.global_variables()

    -sharp
    with tf.Session() as sess:
        sess.run(init)
        saver = tf.train.Saver(all_vars)
        saver.restore(sess, r"model/model.ckpt")-sharp

        -sharp 100epoch
        totalAcc = 0
        for loop in range(loopNum):

            -sharp 
            indices = np.arange(1100)-sharpmodify
            random.shuffle(indices)

            -sharp batch size50
            -sharp 11001000100
            for i in range(0, 0+1000, 50):
                photo = []
                label = []
                -sharpprint("1:",label)
                for j in range(0, 20):
                    photo.append(cv2.resize(cv2.imread(keys[indices[i + j]]), (224, 224))/225)
                    -sharpprint(i+j)
                    label.append(files[keys[indices[i + j]]])
                    -sharpprint("2:",label)

                -sharpembedding,
                target = tf.convert_to_tensor(photo)
                embedding_var = tf.Variable(photo,"data_embedding")
                config = projector.ProjectorConfig()
                embedding = config.embeddings.add()
                embedding.tensor_name = embedding_var.name
                embedding.metadata_path = metadata
                embedding.sprite.single_image_dim.extend([28,28])
                projector.visualize_embeddings(tf.summary.FileWriter(log_dir), config)


                m = getOneHotLabel(label, depth=10)
                a, b = sess.run([optimizer, loss], feed_dict={myinput: photo, mylabel: m})

            acc = 0
            -sharp 20200
            for i in range(1000, 1000+100, 20):
                photo = []
                label = []
                for j in range(i, i + 5):
                    photo.append(cv2.resize(cv2.imread(keys[indices[j]]), (224, 224))/225)
                    label.append(files[keys[indices[j]]])
                m = getOneHotLabel(label, depth=10)
                acc += sess.run(valaccuracy, feed_dict={myinput: photo, mylabel: m})
            -sharp 550
            print("Epoch ", loop, ": validation rate: ", acc/5)
            totalAcc += acc/5
        print("final ",totalAcc/loopNum)
        -sharp 
        saver.save(sess, "model/model.ckpt")
        to_visualise = myinput
        to_visualise = vector_to_mnist(to_visualise)
        to_visualise = invert_grayscale(to_visualise)
        sprite_image = create_sprite_image(to_visualise)
        plt.imsave(metadata,sprite_image)
        plt.imshow(sprite_image)

if __name__ == "__main__":
    alexnet_main()

where

alexnet