Usage
from transformers import AutoImageProcessor, SegformerForSemanticSegmentation
import torch, torch.nn.functional as F
from PIL import Image
import numpy as np
repo = "GlobalWheat/GWFSS_model_v1.0"
processor = AutoImageProcessor.from_pretrained(repo)
model = SegformerForSemanticSegmentation.from_pretrained(repo).eval()
img = Image.open("example.jpg").convert("RGB")
inputs = processor(images=img, return_tensors="pt")
with torch.no_grad():
logits = model(**inputs).logits
up = F.interpolate(logits, size=(img.height, img.width), mode="bilinear", align_corners=False)
pred = up.argmax(1)[0].cpu().numpy() # (H, W) class IDs
This version is based on huggingface Segformer which could be slightly different from the one we used for our paper. The paper version was implemented based on the mmsegmentation. You can find the model weight for mmsegmentation library in this repo as well.
Related Paper
This dataset is associated with the following paper: The Global Wheat Full Semantic Organ Segmentation (GWFSS) Dataset https://doi.org/10.1016/j.plaphe.2025.100084