Does this include support for box and labels as inputs?

#1
by gptQtPi - opened

prompt_encoder_mask_decoder.onnx seems to only include support for points as inputs. These are the three inputs I see-
```
Input Name: input_points
Input Shape: ['batch_size', 'point_batch_size', 'nb_points_per_image', 2]
Input Type: tensor(float)

Input Name: image_embeddings
Input Shape: ['batch_size', 256, 64, 64]
Input Type: tensor(float)

Input Name: image_positional_embeddings
Input Shape: ['batch_size', 256, 64, 64]
Input Type: tensor(float)

Does this mean we can't use box and labels?

@fxmarty , who worked on the SAM export in Optimum, should be able to answer this question :)

Sign up or log in to comment