0

I'm trying to combine Posenet with my custom model. The input of this custom model is: 

Tensor("dense_Dense1_input:0", shape=(?, 34), dtype=float32)

However, the problem is that the outputs of the posenet to be received is different

The input of this custom model has the values ​​of 17 keypoints of Posenet listed in xy order.

The output of this posenet is: 

<tf.Tensor 'tf_op_layer_heatmap/heatmap:0' shape=(?, 23, 23, 17) dtype=float32>, <tf.Tensor 'offset_2/BiasAdd:0' shape=(?, 23, 23, 34) dtype=float32>, <tf.Tensor 'displacement_fwd_2/BiasAdd:0' shape=(?, 23, 23, 32) dtype=float32>, <tf.Tensor 'displacement_bwd_2/BiasAdd:0' shape=(?, 23, 23, 32) dtype=float32>

How can I make these two models together?

this is graph of the models:

enter image description here

enter image description here

enter image description here

enter image description here

JGH
  • 73
  • 4
  • What do you mean by together? Also please full code before asking these questions. – Zabir Al Nazi Apr 27 '20 at 07:55
  • Kindly share the model architectures along with the code. –  Apr 27 '20 at 08:25
  • @ZabirAlNazi I uploaded graphs of these models. Can't you check these out? – JGH Apr 27 '20 at 09:19
  • @Rohit I uploaded graphs of these models. Can't you check these out? – JGH Apr 27 '20 at 09:19
  • check [this answer](https://stackoverflow.com/questions/61446429/how-do-i-connect-two-keras-models-into-one-model/61450306#61450306) which is what you are looking. – Vishnuvardhan Janapati Apr 27 '20 at 13:00
  • @VishnuvardhanJanapati I need to calculate keypoints from those 4 outputs. Calculating that, I need to list 17 keypoint xy coordinates and put them into the input of the target model. Can I create a layer that calculates each 4 outputs? – JGH Apr 28 '20 at 08:42

0 Answers0