cis519-Image2GPS/ImageToGPSproject_resnet18_layer

• lat mean = 39.95169318421053
• lat std = 0.0007139636196696079
• lon mean = -75.19131129824562
• lon std = 0.0006948352800088026

---

To load & evaluate our model:

!pip install geopy > delete.txt
!pip install datasets > delete.txt
!pip install torch torchvision datasets > delete.txt
!pip install huggingface_hub > delete.txt
!rm delete.txt
!pip install transformers
!pip install geopy

import getpass
from huggingface_hub import notebook_login
# Securely input the Hugging Face token
token = getpass.getpass("Enter your Hugging Face token: ")
# Log in to Hugging Face Hub
notebook_login(token)

from huggingface_hub import hf_hub_download
import torch
from huggingface_hub import HfApi, HfFolder, Repository
# Specify the repository and the filename of the model you want to load
repo_id = "cis519-Image2GPS/ImageToGPSproject_resnet18_layer"  # Replace with your repo name
filename = "resnet_gps_regressor_complete.pth"
model_path = hf_hub_download(repo_id=repo_id, filename=filename)
# Load the model using torch
model_test = torch.load(model_path)
model_test.eval()  # Set the model to evaluation mode

from datasets import load_dataset, Image
dataset_test = load_dataset("gydou/released_img", split="train")

import torchvision.transforms as transforms
import numpy as np
from geopy.distance import geodesic
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
transform = transforms.Compose([
    transforms.RandomResizedCrop(224),  # Random crop and resize to 224x224
    transforms.RandomHorizontalFlip(),  # Random horizontal flip
    # transforms.RandomRotation(degrees=15),  # Random rotation between -15 and 15 degrees
    transforms.ColorJitter(brightness=0.2, contrast=0.2, saturation=0.2, hue=0.1),  # Random color jitter
    transforms.ToTensor(),
    transforms.Normalize(mean=[0.485, 0.456, 0.406],
                         std=[0.229, 0.224, 0.225])
])
inference_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])
])
with torch.no_grad():
    for data in dataset_test:
        image = inference_transform(data["image"]).unsqueeze(0).to(device)
        outputs = model_test(image)
        # print("Predicted latitude & longitude:", outputs.cpu().numpy())
lat_mean = 39.95169318421053
lat_std = 0.0007139636196696079
lon_mean = -75.19131129824562
lon_std = 0.0006948352800088026
all_distances = []
model_test.eval()
with torch.no_grad():
    for data in dataset_test:
        image = transform(data["image"]).unsqueeze(0).to(device)
        outputs = model_test(image).cpu().numpy()
        preds_denorm = outputs * np.array([lat_std, lon_std]) + np.array([lat_mean, lon_mean])
        actual = [data["Latitude"], data["Longitude"]]
        distance = geodesic(actual, preds_denorm[0]).meters
        all_distances.append(distance)
mean_error = np.mean(all_distances)
rmse_error = np.sqrt(np.mean(np.square(all_distances)))
print('mean_error: ', mean_error)
print('rmse_error: ', rmse_error)