ozkanib/Multi_Label_Classification

1#!/usr/bin/env python2# coding: utf-83 4# In[ ]:5 6 7#|default_exp export8 9 10# In[1]:11 12 13#|export 14 15# In[416from fastai.vision import *17 18#|export19import gradio as gr20 21 22# In[14]:23 24 25#|export26def get_x(df): return path/'train'/df['fname'] 27def get_y(df): return df['labels'].split()28def splitter(df): 29    valid_idx=df.index[df['is_valid']].tolist()30    train_idx=df.index[df['is_valid']].tolist()31    return train_idx,valid_idx32def accuracy_multip(preds,targets,tresh=0.5,sigmoid=True): 33    if sigmoid: preds=preds.sigmoid()34    return ((preds>tresh)==targets.bool()).float().mean()35 36 37# In[18]:38 39 40#|export41learn=load_learner('MLCmodel.pkl')42 43 44# In[34]:45 46 47#|export48categories=['aeroplane', 'bicycle', 'bird', 'boat', 'bottle', 'bus', 'car', 'cat', 'chair', 'cow', 'diningtable', 'dog', 'horse', 'motorbike', 'person', 'pottedplant', 'sheep', 'sofa', 'train', 'tvmonitor']49 50 51# In[38]:52 53 54#|export55def classify_image(image):56    pred,index,probs=learn.predict(image)57    index=torch.where(index==True)58    return dict(zip(pred,map(float,probs[index])))59 60 61# In[75]:62 63 64#|export65image=gr.inputs.Image(shape=(100,100))66label=gr.outputs.Label()67examples=['human_bicycle2.jpg','human_bicycle.jpg','aero_bicycle.jpg']68inter=gr.Interface(fn=classify_image,inputs=image,outputs=label,examples=examples)69inter.launch(inline=False)70 71