image_utils.py

import torch
from torchvision import transforms as T
import numpy as np
import cv2
from PIL import Image


def opencv_to_pillow(img):
    img = Image.fromarray(cv2.cvtColor(img, cv2.COLOR_BGR2RGB))
    return img


def imagenet_preprocess(filename):
    input_image = Image.open(filename)
    preprocess = T.Compose(
        [
            T.Resize(256),
            T.CenterCrop(224),
            T.ToTensor(),
            T.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]),
        ]
    )
    input_tensor = preprocess(input_image)
    input_batch = input_tensor.unsqueeze(0)
    return input_batch


def yolox_preprocess(filename):
    img = cv2.imread(filename, 1)
    img = cv2.resize(img, (640, 640), interpolation=cv2.INTER_NEAREST)
    img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
    img = np.transpose(img, (2, 0, 1))
    input_batch = np.expand_dims(img, 0)

    return torch.from_numpy(input_batch)


def read_image(name, hw=(224, 224)):
    img = cv2.imread(name, 1)
    img = cv2.resize(img, hw, interpolation=cv2.INTER_NEAREST)
    return img


def get_imagenet_labels(index):
    import json

    labels = []
    with open("data/imagenet_labels.json") as f:
        labels = json.load(f)
    return labels[index]