Closed Bean95zx closed 1 year ago
Hi, it is likely that mask['output_refined'] contains an empty tensor. This can be caused if there is no identified objects in the image. For images similar to the ones the model was trained on, this can be resolved by ensuring the shape of the image is reasonable (e.g. 1024 x 1024). If the image contains a significant domain shift then further fine-tuning is needed.
why show this error wheni use the costumn image ?