search

breadbread1984 / MTCNN-serving-tf2.0

This project implements the serving code of MTCNN facial detection
3 stars 4 forks source link

bad input shape #2

Open ma7555 opened 3 years ago

ma7555 commented 3 years ago

I am receiving this error:

OP_REQUIRES failed at conv_ops_fused_impl.h:716 : Invalid argument: input must be 4-dimensional[1,196,24,24,3]

problem is rnet receives dimensions [1, 196, 24, 24, 3]

I can see that predict_24_batch actually has the correct dimensions of 196, 24, 24, 3 -- Not sure where did that extra axis came from..

predict_24_batch = tf.image.crop_and_resize(inputs, boxes, tf.zeros((rectangles.shape[0],), dtype = tf.int32),(24,24));

mind giving any input?

Tried on both TF2.0 and TF2.3

DSOCT18 commented 3 years ago

I am receiving this error:

OP_REQUIRES failed at conv_ops_fused_impl.h:716 : Invalid argument: input must be 4-dimensional[1,196,24,24,3]

problem is rnet receives dimensions [1, 196, 24, 24, 3]

I can see that predict_24_batch actually has the correct dimensions of 196, 24, 24, 3 -- Not sure where did that extra axis came from..

predict_24_batch = tf.image.crop_and_resize(inputs, boxes, tf.zeros((rectangles.shape[0],), dtype = tf.int32),(24,24));

mind giving any input?

Tried on both TF2.0 and TF2.3

Did you find a solution for this error?