Bidirectional Projection-Based Multi-Modal Fusion Transformer for Early Detection of Cerebral Palsy in Infants
Code release is forthcoming.
The file path should be changed.
Download weights of SegFormer (MiT-B1) pre-trained on ImageNet-1K, and put them in a folder
https://github.com/NVlabs/SegFormer?tab=readme-ov-file
torch==1.10.0+cu113
mmcv==1.6.1
mmcv_full==1.6.1
numpy==1.24.4
opencv_python==4.7.0.72
Pillow==8.2.0
scikit_learn==0.24.1
scipy==1.13.1
Our model is based on the SegFormer,
E. Xie et al. “SegFormer: Simple and efficient design for semantic segmentation with transformers”. NeurIPS 34 (2021), pp. 12077–12090