hloc/matchers/imp.py

# -*- coding: UTF-8 -*-
import sys
from pathlib import Path

import torch

from .. import device, logger
from ..utils.base_model import BaseModel

pram_path = Path(__file__).parent / "../../third_party/pram"
sys.path.append(str(pram_path))

from nets.gml import GML


class IMP(BaseModel):
    default_conf = {
        "match_threshold": 0.2,
        "features": "sfd2",
        "model_name": "imp_gml.920.pth",
        "sinkhorn_iterations": 20,
    }
    required_inputs = [
        "image0",
        "keypoints0",
        "scores0",
        "descriptors0",
        "image1",
        "keypoints1",
        "scores1",
        "descriptors1",
    ]

    def _init(self, conf):
        self.conf = {**self.default_conf, **conf}
        weight_path = pram_path / "weights" / self.conf["model_name"]
        self.net = GML(self.conf).eval().to(device)
        self.net.load_state_dict(
            torch.load(weight_path, map_location="cpu")["model"], strict=True
        )
        logger.info("Load IMP model done.")

    def _forward(self, data):
        data["descriptors0"] = data["descriptors0"].transpose(2, 1).float()
        data["descriptors1"] = data["descriptors1"].transpose(2, 1).float()

        return self.net.produce_matches(data, p=0.2)