Pengcheng Shi

icon indicating a website link https://scholar.google.com/citations?user=aJQOvncAAAAJ&hl=en icon indicating an email [email protected]

icon company Harbin Institute of Technology (Shenzhen) icon location Shenzhen

Results 2 issues of Pengcheng Shi

Implement TTA Batch Processing to Improve Inference Speed

2 comment

**Summary:** Proposing the integration of Test Time Augmentation (TTA) with batch processing in nnUNet to enhance inference efficiency, particularly evident in larger 3D datasets. Demonstrated improvements of 5%-8% in speed...

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

2 comment

Issue: I have encountered a memory consumption issue in my Docker setup during inference tasks. Specifically, when the shm-size is set to 30GB, the system can successfully process softmax operations...