Why is model output size 1?

[yzhilong]

---

Layer (type) Output Shape Param #

input_1 (InputLayer) (None, 10, 3, 128, 160) 0

---

prednet_1 (PredNet) (None, 10, 4) 6915948

---

timedistributed_1 (TimeDistr (None, 10, 1) 5

---

flatten_1 (Flatten) (None, 10) 0

---

dense_2 (Dense) (None, 1) 11

Total params: 6,915,964
Trainable params: 6,915,948
Non-trainable params: 16

This is the model summary I got when I loaded in PredNet. I thought that the network outputs an image? If so why is the size of the output layer 1? Shouldn't it be a 2D array or be a size that can be reshaped into an array? I'm a beginner who's quite confused by this