Code with resnet50 model binary and app.py

README.md

@@ -11,3 +11,35 @@ short_description: ResNet50 model trained on ImageNet Dataset of 1000 classes
 ---
 
 Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+
+<pre>
+Epoch: [10] Total time: 0:34:49
+Test:   [  0/196]  eta: 0:15:22  loss: 2.3581 (2.3581)  acc1: 63.6719 (63.6719)  acc5: 89.0625 (89.0625)  time: 4.7075  data: 4.4695  max mem: 12905
+Test:   [100/196]  eta: 0:00:29  loss: 3.3925 (2.8562)  acc1: 38.2812 (51.5161)  acc5: 66.7969 (79.0068)  time: 0.2379  data: 0.0003  max mem: 12905
+Test:  Total time: 0:00:54
+Test:  Acc@1 47.506 Acc@5 74.236
+
+Epoch: [11] Total time: 0:34:49
+Test:   [  0/196]  eta: 0:15:56  loss: 2.4094 (2.4094)  acc1: 64.4531 (64.4531)  acc5: 89.0625 (89.0625)  time: 4.8820  data: 4.6439  max mem: 12905
+Test:   [100/196]  eta: 0:00:30  loss: 3.4837 (2.9939)  acc1: 36.3281 (49.0215)  acc5: 67.1875 (76.7481)  time: 0.2523  data: 0.0147  max mem: 12905
+Test:  Total time: 0:00:54
+Test:  Acc@1 46.916 Acc@5 73.774
+
+Epoch: [12] Total time: 0:34:49
+Test:   [  0/196]  eta: 0:16:04  loss: 2.1682 (2.1682)  acc1: 69.1406 (69.1406)  acc5: 92.5781 (92.5781)  time: 4.9221  data: 4.6836  max mem: 12905
+Test:   [100/196]  eta: 0:00:29  loss: 3.5695 (2.8364)  acc1: 35.1562 (52.0614)  acc5: 63.6719 (79.2659)  time: 0.2545  data: 0.0168  max mem: 12905
+Test:  Total time: 0:00:54
+Test:  Acc@1 48.304 Acc@5 74.626
+
+Epoch: [13] Total time: 0:34:48
+Test:   [  0/196]  eta: 0:17:28  loss: 2.2632 (2.2632)  acc1: 68.3594 (68.3594)  acc5: 89.4531 (89.4531)  time: 5.3511  data: 5.1130  max mem: 12905
+Test:   [100/196]  eta: 0:00:30  loss: 3.5760 (2.9239)  acc1: 39.4531 (52.5101)  acc5: 66.7969 (78.9720)  time: 0.2380  data: 0.0003  max mem: 12905
+Test:  Total time: 0:00:54
+Test:  Acc@1 48.908 Acc@5 74.982
+
+Epoch: [14] Total time: 0:34:49
+Test:   [  0/196]  eta: 0:16:47  loss: 2.2360 (2.2360)  acc1: 65.6250 (65.6250)  acc5: 90.2344 (90.2344)  time: 5.1387  data: 4.9008  max mem: 12905
+Test:   [100/196]  eta: 0:00:30  loss: 3.4053 (2.8317)  acc1: 38.6719 (52.1117)  acc5: 69.1406 (79.7030)  time: 0.2428  data: 0.0051  max mem: 12905
+Test:  Total time: 0:00:54
+Test:  Acc@1 48.958 Acc@5 75.700
+</pre>

app.py

@@ -0,0 +1,52 @@
+import torch
+import torchvision.transforms as transforms
+from torchvision import models
+import gradio as gr
+from PIL import Image
+
+# Load the ResNet50 model
+model = models.resnet50()
+model.fc = torch.nn.Linear(model.fc.in_features, 1000)  # Update for 1000 classes
+model.load_state_dict(torch.load("model_32.pth", map_location="cpu"))
+model.eval()
+
+# ImageNet class labels
+imagenet_classes = [line.strip() for line in open("imagenet_classes.txt")]
+
+# Define image preprocessing
+def preprocess_image(image):
+    transform = transforms.Compose([
+        transforms.Resize((224, 224)),
+        transforms.ToTensor(),
+        transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]),
+    ])
+    return transform(image).unsqueeze(0)
+
+# Define prediction function
+def predict(image):
+    image = preprocess_image(image)
+    with torch.no_grad():
+        outputs = model(image)
+        probabilities = torch.nn.functional.softmax(outputs[0], dim=0)
+        top5_prob, top5_catid = torch.topk(probabilities, 5)
+    
+    results = []
+    for i in range(top5_prob.size(0)):
+        results.append((imagenet_classes[top5_catid[i]], top5_prob[i].item()))
+    return results
+
+# Gradio Interface
+def classify_image(image):
+    results = predict(image)
+    return {"Class": [r[0] for r in results], "Probability": [r[1] for r in results]}
+
+interface = gr.Interface(
+    fn=classify_image,
+    inputs=gr.Image(type="pil"),
+    outputs=gr.Label(num_top_classes=5),
+    title="ResNet50 Image Classifier",
+    description="Upload an image to get the top 5 predictions using a ResNet50 model trained on ImageNet."
+)
+
+if __name__ == "__main__":
+    interface.launch()

model_32.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:acf3cff2eb11da67bfcad4323a784db7305f1c0afd7e3a6bf411c38903d0f51b
+size 204824981