app.py

import timm
import transformers
from torch import nn
import numpy as np
import gradio as gr
import PIL

# Instantiate classification model
from fastai.vision.all import *
model_multi = load_learner('vit_tiny_patch16.pkl')

def binary_label(path):
    return 'No-anomaly' if (parent_label(path) == 'No-Anomaly') else 'Anomaly'

model_binary = load_learner('vit_tiny_patch16_binary.pkl')

# Instantiate segmentation model
from transformers import SegformerFeatureExtractor, SegformerForSemanticSegmentation
from torchvision.transforms import Grayscale

seg_feature_extractor = SegformerFeatureExtractor.from_pretrained('zklee98/segformer-b1-solarModuleAnomaly-v0.1')
seg_model = SegformerForSemanticSegmentation.from_pretrained('zklee98/segformer-b1-solarModuleAnomaly-v0.1')

def get_seg_overlay(image, seg):
    color_seg = np.zeros((seg.shape[0], seg.shape[1], 3), dtype=np.uint8) # height, width, 3
    palette = np.array(sidewalk_palette())
    for label, color in enumerate(palette):
        color_seg[seg == label, :] = color

    # Show image + mask
    img = np.array(image) * 0.5 + color_seg * 0.5
    img = img.astype(np.uint8)
    #img = PIL.Image.open(img)

    return img

#@title `def sidewalk_palette()`

def sidewalk_palette():
    """Sidewalk palette that maps each class to RGB values."""
    return [
        [0, 0, 0],
        [216, 82, 24],
        [255, 255, 0],
        [125, 46, 141],
        [118, 171, 47],
        [161, 19, 46],
        [255, 0, 0],
        [0, 128, 128],
        [190, 190, 0],
        [0, 255, 0],
        [0, 0, 255],
        [170, 0, 255],
        [84, 84, 0],
        [84, 170, 0],
        [84, 255, 0],
        [170, 84, 0],
        [170, 170, 0],
        [170, 255, 0],
        [255, 84, 0],
        [255, 170, 0],
        [255, 255, 0],
        [33, 138, 200],
        [0, 170, 127],
        [0, 255, 127],
        [84, 0, 127],
        [84, 84, 127],
        [84, 170, 127],
        [84, 255, 127],
        [170, 0, 127],
        [170, 84, 127],
        [170, 170, 127],
        [170, 255, 127],
        [255, 0, 127],
        [255, 84, 127],
        [255, 170, 127],
    ]



def predict(classification_mode, image):

    if (classification_mode == 'Binary Classification'):
        model = model_binary
    else:
        model = model_multi

    labels = model.dls.vocab
    # Classification model prediction
    #image  = PILImage.create(image)
    pred, pred_idx, probs = model.predict(image)

    seg_img = None
    percentage_affected = '0%'
    if (pred.upper() != 'NO-ANOMALY'):
        addChannel = Grayscale(num_output_channels=3)
        image = addChannel(image)

        inputs = seg_feature_extractor(images=image, return_tensors="pt")
        outputs = seg_model(**inputs)
        logits = outputs.logits  # shape (batch_size, num_labels, height/4, width/4)

        # First, rescale logits to original image size
        upsampled_logits = nn.functional.interpolate(
            logits,
            size=image.size[::-1], # (height, width)
            mode='bilinear',
            align_corners=False)

        # Second, apply argmax on the class dimension
        pred_seg = upsampled_logits.argmax(dim=1)[0]

        seg_img = get_seg_overlay(image, pred_seg)

        classified_pixels = np.unique(pred_seg.numpy(), return_counts=True)
        pixels_count = dict({classified_pixels[0][0]: classified_pixels[1][0],
                             classified_pixels[0][1]: classified_pixels[1][1]})
        percentage_affected = round((pixels_count[1]/960)*100, 1)
        percentage_affected = str(percentage_affected) + '%'

        seg_img = PIL.Image.fromarray(seg_img)

    return ({labels[i]: float(probs[i]) for i in range(len(labels))}, seg_img, percentage_affected)


description = """
<center><img src="https://huggingface.co/spaces/zklee98/SolarPanelAnomaly/resolve/main/dronePV_picture.jpg" width=270px> </center><br><br><br><br>
<center>This program identifies the type of anomaly found in solar panel using an image classification model and the percentage of the affected area using an image segmentation model.</center>
<center><i>(Models are trained on <a href="https://ai4earthscience.github.io/iclr-2020-workshop/papers/ai4earth22.pdf">InfraredSolarModules</a> dataset, and hence expect infrared image as input)</center></i>
"""

gr.Interface(fn=predict,
             inputs= [gr.Dropdown(choices=['Binary Classification', 'Multiclass Classification'], label='Classification Mode:', 
                                  info='Choose to classify between anomaly and no-anomaly OR between 12 different types of anomalies.'),
                      gr.Image(type='pil', label='Input infrared image: ')],
             outputs=[gr.outputs.Label(num_top_classes=3, label='Detected:').style(container=False),
                      gr.Image(type='pil', label=' ').style(height=240, width=144),
                      gr.Textbox(label='Affected area:').style(container=False)],
             title='Solar Panel Anomaly Detector',
             description=description,
             examples=['/images/10000.jpg', '/images/1093.jpg'],
                       #[PILImage.create('https://huggingface.co/spaces/zklee98/SolarPanelAnomaly/resolve/main/images/1046.jpg')]],
                       #['/images/1093.jpg'],
                       #['/images/2503.jpg'],
                       #['/images/4978.jpg'],
                       #['/images/5039.jpg'],
                       #['/images/7001.jpg']],
             article= '<center>by <a href="https://www.linkedin.com/in/lzk/">Lee Zhe Kaai</a></center>').launch()