Nonymous-1029/spark_v30_best.pth

CALIPOSE: SparkV30Net Pre-Trained Weights

Pre-trained PyTorch weights for SparkV30Net, the Direct Regression architecture introduced in our paper, "CALIPOSE: Direct Regression vs Modular Tracking for Event-Based Spacecraft Pose Estimation".

Loading the Weights Programmatically

You can easily download and load the pre-trained weights into your evaluation script using the huggingface_hub library:

from huggingface_hub import hf_hub_download
import torch
from train import SparkV30Net  # Assuming your script is named train.py

# Download checkpoint from your Hugging Face repository
checkpoint_path = hf_hub_download(
    repo_id="Pruthvi-1029/spark_v30_best.pth", 
    filename="spark_v30_best.pth"
)

# Load into model
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model = SparkV30Net().to(device)
checkpoint = torch.load(checkpoint_path, map_location=device)
model.load_state_dict(checkpoint)
model.eval()

print("SparkV30Net weights successfully loaded from Hugging Face!")