Update app.py

app.py

@@ -1,45 +1,43 @@
+# app.py
 from fastapi import FastAPI, File, UploadFile
 from transformers import pipeline
 from PIL import Image
 import io
-import torch
 
 app = FastAPI()
 
-# Load the CLIP model and processor directly using the pipeline
-# This happens once when your Space starts up
+# Load the model pipeline only once when the Space starts.
+# This makes it ready for all incoming requests.
 classifier = pipeline("zero-shot-image-classification",
-                      model="openai/clip-vit-base-patch32",
-                      device= -1)  # The '-1' forces it to use the CPU
-
-# These are the candidate labels for your task.
-# The model will assign a score to each one.
-candidate_labels = [
-    "a photo of a fish",
-    "a photo of a person",  # Adding a non-fish example can help
-    "a photo of a tree"
-]
+                      model="google/vit-base-patch16-224",
+                      device=-1)  # -1 forces the model to use the CPU
 
 @app.get("/")
-def read_root():
-    return {"message": "Fish Detector API is running. Send POST to /detect"}
+def root():
+    return {"message": "Fish Detector API is running. Use POST /detect"}
 
 @app.post("/detect")
 async def detect(file: UploadFile = File(...)):
-    # 1. Read the image file
+    # 1. Read and open the image from the uploaded file
     contents = await file.read()
     image = Image.open(io.BytesIO(contents)).convert("RGB")
 
-    # 2. Use the pipeline to run zero-shot classification
-    # We use the 'candidate_labels' we defined above
+    # 2. Define the labels for zero-shot classification.
+    #    These are the categories our model will choose from.
+    candidate_labels = [
+        "a photo of a fish, with fins and scales",
+        "a photo of a person, an animal, or an object that is not a fish",
+        "an empty photo with no fish in it"
+    ]
+
+    # 3. Run the image through the pipeline to get predictions
     predictions = classifier(image, candidate_labels=candidate_labels)
 
-    # 3. Determine the result
-    # The top prediction is the one with the highest score.
-    # We check if its label is 'a photo of a fish'.
-    is_fish = predictions[0]['label'] == 'a photo of a fish'
+    # 4. The top prediction is the one with the highest score.
+    #    We check if its label is the fish-related one.
+    is_fish = predictions[0]['label'] == candidate_labels[0]
 
-    # 4. Return the appropriate text response
+    # 5. Return the result as a plain text string
     if is_fish:
         return "fish"
     else: