ThreadNode / api /main.py
abdurrehmansaeed's picture
Upload 11 files
4031985 verified
Raw History Blame Contribute Delete
2.82 kB
from fastapi import FastAPI, File, UploadFile, HTTPException
from fastapi.middleware.cors import CORSMiddleware
from fastapi.responses import JSONResponse
from fastapi.staticfiles import StaticFiles
import torch
import torch.nn as nn
from torchvision import models, transforms
from PIL import Image
import pickle
import io
# --- 1. CONFIGURATION ---
MODEL_PATH = "../models/fashion_hierarchy_model_v3.pth"
VOCAB_PATH = "../data/mlb_vocab.pkl"
THRESHOLD = 0.85
# --- 2. INITIALIZE FASTAPI ---
app = FastAPI(title="Apparel Taxonomy API", version="1.0")
app.add_middleware(
CORSMiddleware,
allow_origins=["*"],
allow_credentials=True,
allow_methods=["*"],
allow_headers=["*"],
)
# --- 3. HARDWARE & TRANSFORMS ---
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
print(f"Server spinning up. Computation routed to: {device.type.upper()}")
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])
])
# --- 4. LOAD STATE INTO MEMORY ---
try:
with open(VOCAB_PATH, 'rb') as f:
mlb = pickle.load(f)
num_classes = len(mlb.classes_)
model = models.efficientnet_b0(weights=None)
in_features = model.classifier[1].in_features
model.classifier[1] = nn.Linear(in_features, num_classes)
model.load_state_dict(torch.load(MODEL_PATH, map_location=device))
model.to(device)
model.eval()
print("V3 Engine successfully loaded and locked into memory.")
except Exception as e:
print(f"CRITICAL BOOT ERROR: {e}")
# --- 5. THE INFERENCE ROUTE ---
@app.post("/predict")
async def predict_apparel(file: UploadFile = File(...)):
if not file.content_type.startswith("image/"):
raise HTTPException(status_code=400, detail="Invalid file type. Please upload an image.")
try:
image_bytes = await file.read()
image = Image.open(io.BytesIO(image_bytes)).convert("RGB")
input_tensor = transform(image).unsqueeze(0).to(device)
with torch.no_grad():
outputs = model(input_tensor)
probs = torch.sigmoid(outputs)
preds = (probs > THRESHOLD).float().cpu().numpy()
predicted_tags = mlb.inverse_transform(preds)[0]
return JSONResponse(content={
"status": "success",
"threshold_applied": THRESHOLD,
"tags": list(predicted_tags)
})
except Exception as e:
raise HTTPException(status_code=500, detail=f"Inference Engine Error: {str(e)}")
# --- 6. SERVE THE UI ---
app.mount("/", StaticFiles(directory="../frontend", html=True), name="frontend")