update app.py

app.py

@@ -51,5 +51,28 @@ def pred(img):
         pred = torch.softmax(logits, dim=-1)
     return f"prediction: {class_names[pred.argmax(dim=-1).item()]} | confidence: {pred.max():.3f}"
 
-demo = gr.Interface(pred, gr.Image(), outputs="text")
-demo.launch()
+demo = gr.Blocks()
+
+with demo: 
+    gr.Markdown("""
+        # Welcome to FreshVision 📷
+        _FreshVision is a machine learning model to classify freshness for fruits. This model 
+        utilizes transfer learning from pre-trained model from PyTorch [EfficientNetB0](https://pytorch.org/vision/stable/models/generated/torchvision.models.efficientnet_b0.html).
+        This model has been trained on [kaggle datasets](https://www.kaggle.com/datasets/sriramr/fruits-fresh-and-rotten-for-classification) using NVIDIA T4 GPU._
+        
+        ## Model capabilities:
+        - Classify freshness from fruits image (apple, orange, and banana) with two labels: _Fresh_ and _Rotten/spoiled_
+        ## Model drawbacks: 
+        - Sometimes perform false prediction on some fruits condition, this is due to low variability on the image datasets
+        - Can't perform accurate prediction on multiple objects/combined condition (e.g. two bananas with different freshness condition)
+        - This models can't identify non-fruits objects , since it's only trained with fruits dataset. 
+                
+        ## **How to get the best result with this model:** 
+        1. The image should only contain fruits that the model can recognize (apple, orange, and banana)
+        2. The image should only contain one object (one fruit)
+        3. Ensure the object is captured with sufficient light so that the surface of the fruits is exposed properly
+""")
+    gr.Interface(pred, gr.Image(), outputs="text")
+
+if __name__ == "__main__":
+    demo.launch()