SegFormer (b0-sized) model fine-tuned on FLAME

The model was trained for a deep learning project titled Forest Fire Detection.

Model Description

The model is intended to be used for fire detection through image segmentation.

The provided pretrained model was finetuned on the FLAME dataset for 3 epochs with a learning rate of 1e-3 and was able to score an IOU score of 0.745 on the test examples.

How to use

Here is how to use this model to segment an image:

from transformers import SegformerImageProcessor, SegformerForSemanticSegmentation
from PIL import Image
import requests

processor = AutoFeatureExtractor.from_pretrained("millionhz/segformer-b0-finetuned-flame")
model = SegformerForSemanticSegmentation.from_pretrained("millionhz/segformer-b0-finetuned-flame")


url = <add url here>
image = Image.open(requests.get(url, stream=True).raw)

inputs = feature_extractor(images=image, return_tensors="pt")
outputs = model(**inputs)

License

The license for this model can be found here.