Update README.md

README.md

@@ -3,18 +3,24 @@ tags:
 - fastai
 ---
 
-# Amazing!
+from huggingface_hub import from_pretrained_fastai
+from fastai.vision.all import *
+def label_func(f): return f.name[:2]
 
-🥳 Congratulations on hosting your fastai model on the Hugging Face Hub!
 
-# Some next steps
-1. Fill out this model card with more information (see the template below and the [documentation here](https://huggingface.co/docs/hub/model-repos))!
+learn = from_pretrained_fastai("smaciu/bee-wings-classifier")
+
+
+def predict_image(image_path):
+    img = PILImage.create(image_path)
+    bee,_,probs = learn.predict(img)
+    return bee, max(probs)
+
+cat, prob = predict_image('path_to_image')
+print(f"Honey bee from: {cat}. {100*prob.item():.2f}%")
 
-2. Create a demo in Gradio or Streamlit using 🤗 Spaces ([documentation here](https://huggingface.co/docs/hub/spaces)).
 
-3. Join the fastai community on the [Fastai Discord](https://discord.com/invite/YKrxeNn)!
 
-Greetings fellow fastlearner 🤝! Don't forget to delete this content from your model card.
 
 
 ---
@@ -29,4 +35,48 @@ More information needed
 More information needed
 
 ## Training and evaluation data
+
+from fastai.vision.all import *
+
+from huggingface_hub import from_pretrained_fastai
+
+from datasets import load_dataset
+
+def label_func(f): return f.name[:2]
+
+dataset = load_dataset("smaciu/bee-wings-large", split='train')
+
+
+def get_items(o): 
+    return range(len(dataset))
+
+def get_x(i): 
+    return dataset[i]['image']
+
+def get_y(i): 
+    return dataset[i]['label']
+
+bee_wing_stats =([0.7641, 0.7641, 0.7641], [0.1771, 0.1771, 0.1771]) # dataset mean and std to normalise 
+
+# 2. Create a FastAI DataBlock
+dls = DataBlock(
+                blocks=(ImageBlock, CategoryBlock),
+                get_items=get_items,
+                get_x=get_x,                             # function to get the image file path
+                get_y=get_y,                             # function to get the image label
+                splitter=RandomSplitter(valid_pct=0.2),  # random split with 20% in validation set
+                item_tfms=Resize(460),                   # item transforms
+                batch_tfms= [Normalize.from_stats(*bee_wing_stats)]).dataloaders(dataset,bs=64, num_workers=num_cpus(), pin_memory=True).to('device')
+
+cbfs = [
+        #ShowGraphCallback,
+        #ReduceLROnPlateau(monitor='valid_loss', min_delta=0.1, patience=2),
+        ]
+
+learn = from_pretrained_fastai("smaciu/bee-wings-classifier").to_channelslast()
+
+learn.dls = dls
+
+learn.fit_one_cycle(1, lr_max=1.4858086386360967e-08)
+
 More information needed