pretrained model

• https://huggingface.co/nvidia/mit-b0
• SegFormer (b0-sized) encoder pre-trained-only
• SegFormer encoder fine-tuned on Imagenet-1k. It was introduced in the paper SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformers by Xie et al. and first released in this repository.

training-set

• you can find the training-set here : https://codalab.lisn.upsaclay.fr/competitions/8769

training-arguments

• channels : rgb
• batch : 8
• epochs : 8
• learning-rate : 5e-6
• GPU : T4

results on test-set

• Mean IoU : 59.9
• more information here : https://codalab.lisn.upsaclay.fr/competitions/8769