Pengcheng Shi comments

Results 5 comments of


                                            Pengcheng Shi

Implement TTA Batch Processing to Improve Inference Speed

Hi Fabian, Thanks for your feedback! Based on your suggestions, I have now made the "use_batch_tta" an optional parameter in the nnUNetPredictor class, which can be controlled via the parser...

Welcome to suggest more medical segmentation backbones

Thanks for your impressive work! We developed a Topology-Aware backbone based on the Vision GNN (ViG), offering new insights in medical image segmentation. Paper title: NexToU: Efficient Topology-Aware U-Net for...

Implement Argmax Inference Instead of Softmax when 'has_regions' is False

Hi, Thank you for your feedback. I encountered this issue while preparing for [AortaSeg24 Grand Challenge](https://aortaseg24.grand-challenge.org/). According to the Grand Challenge platform's requirements for algorithms ([Grand Challenge Documentation](https://grand-challenge.org/documentation/runtime-environment/)), 50% of...

Implement Argmax Inference Instead of Softmax when 'has_regions' is False

Hi, thanks for providing the detailed inference code for ToothFairy2. It has been really helpful for submitting inference code to the challenge platform. I will close this pull request.

Blosc2: Could not build empty array

I'm experiencing the same problem when using blosc2==3.6.1. However, when I downgrade to blosc2==3.5.0, the code runs without issues.