papi_cc / app /utils.py
CJR61's picture
Claude Sonnet 4.6
Replace B3 ONNX with EfficientNet-B0 (3-fold CV, orig fold_3)
a5ec00a
Raw History Blame
2.3 kB
import cv2
import numpy as np
import os
import csv
import uuid
from datetime import datetime
ALLOWED_EXTENSIONS = {".jpg", ".jpeg", ".png", ".bmp", ".tiff", ".tif"}
def save_uploaded_file(save_dir, file):
# Validate file extension
_, ext = os.path.splitext(file.filename or "")
if ext.lower() not in ALLOWED_EXTENSIONS:
raise ValueError(f"Unsupported file type '{ext}'. Allowed: {ALLOWED_EXTENSIONS}")
# Ensure upload directory exists
os.makedirs(save_dir, exist_ok=True)
# Generate unique ID (date prefix + uuid4) to avoid race condition
image_id = datetime.now().strftime("%Y%m%d_%H%M%S") + "_" + uuid.uuid4().hex[:8]
saved_path = os.path.join(save_dir, f"{image_id}.jpg")
# Save file
with open(saved_path, "wb") as f:
f.write(file.file.read())
return saved_path, image_id
def is_valid_image(file_path):
try:
image = cv2.imread(file_path)
if image is None:
return False
return True
except Exception:
return False
def preprocess_image(image_path):
# Read image — Resize(256) + CenterCrop(224) to match training val_transform
image = cv2.imread(image_path)
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
h, w = image.shape[:2]
if h < w:
new_h, new_w = 256, int(w * 256 / h)
else:
new_h, new_w = int(h * 256 / w), 256
image = cv2.resize(image, (new_w, new_h))
y0 = (new_h - 224) // 2
x0 = (new_w - 224) // 2
image = image[y0:y0+224, x0:x0+224]
image = image / 255.0
mean = np.array([0.485, 0.456, 0.406])
std = np.array([0.229, 0.224, 0.225])
image = (image - mean) / std
image = np.transpose(image, (2, 0, 1)) # HWC -> CHW
input_tensor = np.expand_dims(image, axis=0).astype(np.float32)
return input_tensor
def append_to_csv(csv_path, image_id, prediction, probability, similarity):
# Ensure CSV file exists
os.makedirs(os.path.dirname(csv_path), exist_ok=True)
file_exists = os.path.isfile(csv_path)
with open(csv_path, mode="a", newline="") as f:
writer = csv.writer(f)
if not file_exists:
writer.writerow(["ID", "Prediction", "Probability", "Similarity"])
writer.writerow([image_id, prediction, probability, round(float(similarity), 3)])