File size: 5,253 Bytes
a561d8f
 
 
 
 
 
 
 
 
b116e78
a561d8f
 
 
 
f7fcf35
 
a561d8f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
ff93443
 
 
a561d8f
 
 
fd159d4
a561d8f
 
 
 
 
 
 
 
 
 
 
 
 
 
fb4246b
 
a561d8f
 
 
 
 
 
 
ff93443
a8aa8cb
a561d8f
 
 
 
 
 
 
 
 
 
 
ff93443
a561d8f
 
a8aa8cb
a561d8f
 
 
 
 
 
 
 
 
 
5ff4a6c
a561d8f
a8aa8cb
a561d8f
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
import io
import gradio as gr
import requests, validators
import torch
import pathlib
from PIL import Image
import datasets
from transformers import AutoFeatureExtractor, AutoModelForImageClassification
import os
import IPython


os.environ["KMP_DUPLICATE_LIB_OK"]="TRUE"

feature_extractor = AutoFeatureExtractor.from_pretrained("saved_model_files")
model = AutoModelForImageClassification.from_pretrained("saved_model_files")

labels = ['angular_leaf_spot', 'bean_rust', 'healthy']

def classify(im):
  '''FUnction for classifying plant health status'''

  features = feature_extractor(im, return_tensors='pt')
  with torch.no_grad():
    logits = model(**features).logits
  probability = torch.nn.functional.softmax(logits, dim=-1)
  probs = probability[0].detach().numpy()
  confidences = {label: float(probs[i]) for i, label in enumerate(labels)}

  return confidences
    
def get_original_image(url_input):
  '''Get image from URL'''
  if validators.url(url_input):

      image = Image.open(requests.get(url_input, stream=True).raw)
      
      return image

def detect_plant_health(url_input,image_input,webcam_input):
    
    if validators.url(url_input):
        image = Image.open(requests.get(url_input, stream=True).raw)
         
    elif image_input:
        image = image_input
        
    elif webcam_input:
        image = webcam_input
    
    #Make prediction
    label_probs = classify(image)
    
    return label_probs
        
def set_example_image(example: list) -> dict:
    return gr.Image.update(value=example[0])

def set_example_url(example: list) -> dict:
    return gr.Textbox.update(value=example[0]), gr.Image.update(value=get_original_image(example[0]))


title = """<h1 id="title">Plant Health Classification with ViT</h1>"""

description = """
This Plant Health classifier app was built to detect the health of plants using images of leaves by fine-tuning a Vision Transformer (ViT) [google/vit-base-patch16-224](https://huggingface.co/google/vit-base-patch16-224) on the [Beans](https://huggingface.co/datasets/beans) dataset.
The finetuned model has an accuracy of 98.4% on the test (unseen) dataset and 100% on the validation dataset.

How to use the app:
- Upload an image via 3 options, uploading the image from local device, using a URL (image from the web) or a webcam
- The app will take a few seconds to generate a prediction with the following labels:
  - *angular_leaf_spot* 
  - *bean_rust*
  - *healthy*
- Feel free to click the image examples as well.  
"""
urls = ["https://www.healthbenefitstimes.com/green-beans/","https://huggingface.co/nateraw/vit-base-beans/resolve/main/angular_leaf_spot.jpeg", "https://huggingface.co/nateraw/vit-base-beans/resolve/main/bean_rust.jpeg"]
images = [[path.as_posix()] for path in sorted(pathlib.Path('images').rglob('*.p*g'))]

twitter_link = """
[![](https://img.shields.io/twitter/follow/nickmuchi?label=@nickmuchi&style=social)](https://twitter.com/nickmuchi)
"""

css = '''
h1#title {
  text-align: center;
}
'''
demo = gr.Blocks(css=css)

with demo:
    gr.Markdown(title)
    gr.HTML('<center><img src="images/Healthy.png" width=250px height=250px></center>')

    gr.Markdown(description)
    gr.Markdown(twitter_link)
    
    with gr.Tabs():
        with gr.TabItem('Image Upload'):
            with gr.Row():
                with gr.Column():
                  img_input = gr.Image(type='pil',shape=(450,450))
                  label_from_upload= gr.Label(num_top_classes=3)
                
            with gr.Row(): 
                example_images = gr.Examples(examples=images,inputs=[img_input])                            
                                                   
                
            img_but = gr.Button('Classify')
            
        with gr.TabItem('Image URL'):
            with gr.Row():
                with gr.Column():
                    url_input = gr.Textbox(lines=2,label='Enter valid image URL here..')
                    original_image = gr.Image(shape=(450,450))
                    url_input.change(get_original_image, url_input, original_image)
                with gr.Column():
                    label_from_url = gr.Label(num_top_classes=3)
                
            with gr.Row():
                example_url = gr.Examples(examples=urls,inputs=[url_input])
                
            
            url_but = gr.Button('Classify')
            
        with gr.TabItem('WebCam'):
            with gr.Row():
                with gr.Column():
                  web_input = gr.Image(source='webcam',type='pil',shape=(450,450),streaming=True)
                with gr.Column():
                  label_from_webcam= gr.Label(num_top_classes=3)

            cam_but = gr.Button('Classify')
            
    url_but.click(detect_plant_health,inputs=[url_input,img_input,web_input],outputs=[label_from_url],queue=True)
    img_but.click(detect_plant_health,inputs=[url_input,img_input,web_input],outputs=[label_from_upload],queue=True)
    cam_but.click(detect_plant_health,inputs=[url_input,img_input,web_input],outputs=[label_from_webcam],queue=True)

    gr.Markdown("![visitor badge](https://visitor-badge.glitch.me/badge?page_id=nickmuchi-plant-health)")

    
demo.launch(debug=True,enable_queue=True)