hello everyone!
been surfing on Doc + lot of tutorials but i dont quite get what im doing wrong.
What im atempting to do is relativly simple: trying to do transfer learning on the Ssd mobilnetv2 network. I’ve followed this tutorial here , but when i arrrive to this step , the model i want to use is not the mobilenet v2 but Ssd Mobilenet v2. It doesnt seems to exist a tf.keras.application.ssdmobilnetv2 , so i’ve decided to use the model that is aviable on tensorFlow hub ( this one ) . so i’ve researched on how to get a Tensorflow python Model object to be able to follow the tutorial steps, and i’ve encountered this other tutorial . the thing is, apparently ssd MobilnetV2 has multiple outputs, so i cant call tf.keras.Sequential with the kerasLayer obtained. Then the documentation says that when having multiple outputs, you need to use the Functional API. the thing is, how can i call the Functional API with my ssd_mobilnetv2 downloaded on THUB? there seems to be no examples on the web. here is the Python sample of code im trying to make work
"https://tfhub.dev/tensorflow/ssd_mobilenet_v2/2",
input_shape=(300, 300, 3),
dtype=tf.uint8,
trainable=False)
num_classes = len(class_names)
print(num_classes)
#### using the Functional api ( doesnt work)
model= tf.keras.Model(
feature_extractor_layer,
tf.keras.layers.Dense(num_classes))
#### using keras.Sequential, doesnt work too
model = tf.keras.Sequential([
feature_extractor_layer,
tf.keras.layers.Dense(num_classes)
])
model.summary()````
i would be glad if someone can help me out!
best regards