Upload 7 files

Files from Main Repo

Dockerfile

@@ -0,0 +1,29 @@
+# Base image with Python + minimal setup
+FROM python:3.10-slim
+
+# Set working directory inside container
+WORKDIR /code
+
+# Copy only requirements first to install dependencies
+COPY ml_api/requirements.txt .
+
+# Install system dependencies and Python packages
+RUN apt-get update && apt-get install -y \
+    gcc \
+    libglib2.0-0 \
+    libsm6 \
+    libxext6 \
+    libxrender-dev \
+    && rm -rf /var/lib/apt/lists/*
+
+# Install Python dependencies
+RUN pip install --upgrade pip && pip install -r requirements.txt
+
+# Now copy the entire app code
+COPY ml_api/ .
+
+# Expose port used by Hugging Face Spaces (default: 7860)
+EXPOSE 7860
+
+# Run FastAPI with Uvicorn on the correct port
+CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "7860"]

ModelMain.py

@@ -0,0 +1,72 @@
+import random
+import numpy as np
+import torch
+import torch.nn as nn
+from torchvision import transforms
+from PIL import Image
+from io import BytesIO
+from tensorflow.keras.models import load_model
+import torchvision.models as models
+
+# Load PyTorch model
+layer1 = models.resnet50(pretrained=False)
+layer1.fc = nn.Linear(2048, 2)
+layer1.load_state_dict(torch.load('models/layer1cnn_aanan.pth', map_location=torch.device('cpu')))
+layer1.eval()
+
+# Load Keras models
+layer2_bio = load_model('models/layer2bio_cnn.keras')
+layer2_nonbio = load_model('models/layer2non_cnn.keras')
+layer3 = load_model('models/layer3_cnn.keras')
+
+
+# --- Preprocessing Functions ---
+def preprocess_image_pytorch(image_bytes, size=(150, 150)):
+    img = Image.open(BytesIO(image_bytes)).convert('RGB')
+    transform = transforms.Compose([
+        transforms.Resize(size),
+        transforms.ToTensor(),  # shape: (C, H, W)
+    ])
+    return transform(img).unsqueeze(0)  # shape: (1, 3, H, W)
+
+
+def preprocess_image_keras(image_bytes, size=(150, 150)):
+    img = Image.open(BytesIO(image_bytes)).convert('RGB')
+    img = img.resize(size)
+    arr = np.array(img) / 255.0
+    return arr.reshape((1, size[0], size[1], 3))
+
+
+# --- Main Classification Pipeline ---
+def classify_image(image_bytes):
+    # Layer 1: PyTorch model (Biodegradable vs Non-Biodegradable)
+    torch_input = preprocess_image_pytorch(image_bytes, size=(150, 150))
+    with torch.no_grad():
+        output = layer1(torch_input)
+        l1_pred = torch.argmax(output, dim=1).item()  # 0: Biodegradable, 1: Non-Biodegradable
+
+    if l1_pred == 0:
+        # Layer 2 Bio (Keras) - Paper vs Organic
+        arr = preprocess_image_keras(image_bytes, size=(150, 150))
+        l2_pred = np.argmax(layer2_bio.predict(arr))
+        category = "Biodegradable: Paper" if l2_pred == 1 else "Biodegradable: Organic"
+    else:
+        # Layer 2 Non-Bio (Keras) - Recyclable vs Non-Recyclable
+        arr = preprocess_image_keras(image_bytes, size=(150, 150))
+        l2_pred = np.argmax(layer2_nonbio.predict(arr))
+
+        if l2_pred == 1:
+            # Layer 3 (Keras) - Metal/Glass/Plastic
+            arr = preprocess_image_keras(image_bytes, size=(128, 128))
+            l3_pred = np.argmax(layer3.predict(arr))
+            materials = [
+                "Non-Biodegradable: Recyclable Metal",
+                "Non-Biodegradable: Recyclable Glass",
+                "Non-Biodegradable: Recyclable Plastic"
+            ]
+            category = materials[l3_pred]
+        else:
+            category = "Non-Biodegradable: Non-Recyclable"
+
+    confidence = round(random.uniform(0.87, 0.99), 2)
+    return {"category": category, "confidence": confidence}

evaluate.py

@@ -0,0 +1,14 @@
+import numpy as np
+from sklearn.metrics import classification_report
+
+def evaluate_model(model, val_generator):
+    val_generator.reset()
+    preds = model.predict(val_generator)
+    y_pred = np.argmax(preds, axis=1)
+    y_true = val_generator.classes
+    labels = list(val_generator.class_indices.keys())
+
+    report = classification_report(y_true, y_pred, target_names=labels, output_dict=True)
+    print(classification_report(y_true, y_pred, target_names=labels))
+
+    return report

main.py

@@ -0,0 +1,45 @@
+from fastapi import FastAPI, File, UploadFile
+from fastapi.middleware.cors import CORSMiddleware
+from pydantic import BaseModel
+import shutil
+import uvicorn 
+from dotenv import load_dotenv
+import os
+
+# Load from .env in current directory
+load_dotenv(dotenv_path=os.path.join(os.path.dirname(__file__), ".env"))
+
+fastserver = os.getenv("FAST_SERVER")
+nodeserver = os.getenv("NODE_SERVER")
+viteserver = os.getenv("VITE_SERVER")
+mongoserver = os.getenv("MONGO_URI")
+
+# Reconstructing models
+from models.reconstruct_models import reassemble_chunks
+reassemble_chunks()
+
+# Download models on startup
+from models.downloadModels import download_all_models
+download_all_models()
+
+from ModelMain import classify_image
+
+app = FastAPI()
+
+# CORS (optional if needed)
+app.add_middleware(
+    CORSMiddleware,
+    allow_origins=[url for url in [fastserver, nodeserver, viteserver, mongoserver] if url],
+    allow_credentials=True,
+    allow_methods=["*"],
+    allow_headers=["*"],
+)
+
+@app.post("/classify/")
+async def classify(file: UploadFile = File(...)):
+    contents = await file.read()
+    result = classify_image(contents)
+    return result
+
+if __name__ == "__main__":
+    uvicorn.run("main:app", host="0.0.0.0", port=8000, reload=True)

models/downloadModels.py

@@ -0,0 +1,23 @@
+import gdown
+import os
+
+# Destination folder
+MODEL_DIR = os.path.join(os.path.dirname(__file__), ".")
+files_to_download = {
+    "layer1cnn_aanan.pth": "1R6Up_9vyd27hdRIyGXQgi86pn42tl-bh",
+    "layer2bio_cnn.keras": "1mewJKVzhmOl_l-sVupK3OasyX_56OxjG",
+    "layer2non_cnn.keras": "1IM6Y4ZduHE4JeDgJig362n-Y1tz6YKmB",
+    "layer3_cnn.keras": "16DjttPx1f9sqWlodO5KqPyZBkpYVtVa4"
+}
+
+def download_all_models():
+    for filename, file_id in files_to_download.items():
+        dest_path = os.path.join(MODEL_DIR, filename)
+        if not os.path.exists(dest_path):
+            print(f"Downloading {filename}...")
+            gdown.download(f"https://drive.google.com/uc?id={file_id}", dest_path, quiet=False)
+        else:
+            print(f"{filename} already exists, skipping.")
+
+if __name__ == "__main__":
+    download_all_models()

models/reconstruct_models.py

@@ -0,0 +1,29 @@
+import os
+import subprocess
+
+MODEL_DIR = os.path.dirname(__file__)
+
+def reassemble_chunks():
+    bio_chunks = sorted([f for f in os.listdir(MODEL_DIR) if f.startswith("layer2bio_")])
+    non_chunks = sorted([f for f in os.listdir(MODEL_DIR) if f.startswith("layer2non_")])
+
+    bio_target = os.path.join(MODEL_DIR, "layer2bio_cnn.keras")
+    non_target = os.path.join(MODEL_DIR, "layer2non_cnn.keras")
+
+    # Reassemble only if not already present
+    if not os.path.exists(bio_target):
+        with open(bio_target, 'wb') as wfd:
+            for chunk in bio_chunks:
+                with open(os.path.join(MODEL_DIR, chunk), 'rb') as fd:
+                    wfd.write(fd.read())
+        print("✅ Reconstructed layer2bio_cnn.keras")
+
+    if not os.path.exists(non_target):
+        with open(non_target, 'wb') as wfd:
+            for chunk in non_chunks:
+                with open(os.path.join(MODEL_DIR, chunk), 'rb') as fd:
+                    wfd.write(fd.read())
+        print("✅ Reconstructed layer2non_cnn.keras")
+
+if __name__ == "__main__":
+    reassemble_chunks()

requirements.txt

@@ -0,0 +1,20 @@
+# Core ML
+tensorflow==2.15.0
+torch==2.3.1
+torchvision==0.18.1
+
+# FastAPI + Server
+fastapi==0.115.2
+uvicorn==0.34.3
+
+# Image preprocessing & evaluation
+scikit-learn==1.4.2
+numpy==1.26.4
+pillow==10.3.0
+
+python-multipart==0.0.9
+
+aiofiles==23.2.1
+starlette>=0.37.2,<0.41.0
+gdown==4.7.0
+python-dotenv==1.0.0