added app.py

README.md

@@ -1,6 +1,6 @@
 ---
 title: AvianVision
-emoji: 👁
+emoji: 
 colorFrom: pink
 colorTo: purple
 sdk: gradio

app.py

@@ -0,0 +1,80 @@
+from models import EfficientNet
+from utils import get_device
+import torch
+import json
+import gradio as gr
+import torch
+from torchvision import transforms
+from PIL import Image
+import json
+import timm
+from torch import nn
+import torch.nn.functional as F
+
+def load_efficientnet_model(model_path: str, device=get_device()):
+    """
+    Load a PyTorch model checkpoint.
+
+    Args:
+        model_path: The path of the checkpoint file.
+        device: The device to load the model onto.
+
+    Returns:
+        The model loaded onto the specified device.
+    """
+    # Initialize model
+    model = EfficientNet()
+
+    # Load model weights onto the specified device
+    model.load_state_dict(torch.load(model_path, map_location=device)['model_state_dict'])
+
+    # Set model to evaluation mode
+    model.eval()
+
+    return model
+
+with open('idx_to_class.json', 'r') as f:
+    idx_to_class = json.load(f)
+
+
+def predict_image(array):
+    """
+    Predict the class of an image.
+
+    Args:
+        array: The image data as an array.
+
+    Returns:
+        The predicted class.
+    """
+    # Convert the image to a PIL Image object
+    input_image = Image.fromarray(array)
+
+    # Load the model
+    model = load_efficientnet_model('/home/vedmani/Downloads/efficientnet_epoch=18_loss=0.0020_val_f1score=0.8993.pth')
+
+    # Transform the image
+    transform = transforms.Compose([
+        transforms.Resize(size=(150, 150)),
+        transforms.ToTensor(),
+        transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5)),
+    ])
+    image = transform(input_image).unsqueeze(0)
+    image.to(get_device())
+
+    # Predict the class
+    with torch.no_grad():
+        output = model(image)
+        # Apply softmax to the outputs to convert them into probabilities
+        probabilities = F.softmax(output, dim=1)
+        predicted = probabilities.argmax().item()
+        predicted_class = idx_to_class[str(predicted)]  # Make sure your keys in json are string type
+
+    return predicted_class
+
+
+# Create the image classifier
+image_classifier = gr.Interface(fn=predict_image, inputs="image", outputs="text", allow_flagging='Never')
+
+# Launch the image classifier
+image_classifier.launch(share=True)

efficientnet_epoch=18_loss=0.0020_val_f1score=0.8993.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:86679a9101ac637ac321200f69d807b92d8c7419879111a8f598cd24d7987445
+size 49005075

idx_to_class.json

@@ -0,0 +1 @@
+{"0": "Asian-Green-Bee-Eater", "1": "Brown-Headed-Barbet", "2": "Cattle-Egret", "3": "Common-Kingfisher", "4": "Common-Myna", "5": "Common-Rosefinch", "6": "Common-Tailorbird", "7": "Coppersmith-Barbet", "8": "Forest-Wagtail", "9": "Gray-Wagtail", "10": "Hoopoe", "11": "House-Crow", "12": "Indian-Grey-Hornbill", "13": "Indian-Peacock", "14": "Indian-Pitta", "15": "Indian-Roller", "16": "Jungle-Babbler", "17": "Northern-Lapwing", "18": "Red-Wattled-Lapwing", "19": "Ruddy-Shelduck", "20": "Rufous-Treepie", "21": "Sarus-Crane", "22": "White-Breasted-Kingfisher", "23": "White-Breasted-Waterhen", "24": "White-Wagtail"}