Spaces:
Running
on
Zero
Running
on
Zero
| import sys | |
| from pathlib import Path | |
| import torch | |
| from PIL import Image | |
| from .. import MODEL_REPO_ID, logger | |
| from ..utils.base_model import BaseModel | |
| roma_path = Path(__file__).parent / "../../third_party/RoMa" | |
| sys.path.append(str(roma_path)) | |
| from romatch.models.model_zoo import roma_model | |
| device = torch.device("cuda" if torch.cuda.is_available() else "cpu") | |
| class Roma(BaseModel): | |
| default_conf = { | |
| "name": "two_view_pipeline", | |
| "model_name": "roma_outdoor.pth", | |
| "model_utils_name": "dinov2_vitl14_pretrain.pth", | |
| "max_keypoints": 3000, | |
| "coarse_res": (560, 560), | |
| "upsample_res": (864, 1152), | |
| } | |
| required_inputs = [ | |
| "image0", | |
| "image1", | |
| ] | |
| # Initialize the line matcher | |
| def _init(self, conf): | |
| model_path = self._download_model( | |
| repo_id=MODEL_REPO_ID, | |
| filename="{}/{}".format(Path(__file__).stem, self.conf["model_name"]), | |
| ) | |
| dinov2_weights = self._download_model( | |
| repo_id=MODEL_REPO_ID, | |
| filename="{}/{}".format(Path(__file__).stem, self.conf["model_utils_name"]), | |
| ) | |
| logger.info("Loading Roma model") | |
| # load the model | |
| weights = torch.load(model_path, map_location="cpu") | |
| if self.conf["model_name"] == "gim_roma_100h.ckpt": | |
| if "state_dict" in weights.keys(): | |
| weights = weights["state_dict"] | |
| for k in list(weights.keys()): | |
| if k.startswith("model."): | |
| weights[k.replace("model.", "", 1)] = weights.pop(k) | |
| dinov2_weights = torch.load(dinov2_weights, map_location="cpu") | |
| if str(device) == "cpu": | |
| amp_dtype = torch.float32 | |
| else: | |
| amp_dtype = torch.float16 | |
| self.net = roma_model( | |
| resolution=self.conf["coarse_res"], | |
| upsample_preds=True, | |
| weights=weights, | |
| dinov2_weights=dinov2_weights, | |
| device=device, | |
| amp_dtype=amp_dtype, | |
| ) | |
| self.net.upsample_res = self.conf["upsample_res"] | |
| logger.info("Load Roma model done.") | |
| def _forward(self, data): | |
| img0 = data["image0"].cpu().numpy().squeeze() * 255 | |
| img1 = data["image1"].cpu().numpy().squeeze() * 255 | |
| img0 = img0.transpose(1, 2, 0) | |
| img1 = img1.transpose(1, 2, 0) | |
| img0 = Image.fromarray(img0.astype("uint8")) | |
| img1 = Image.fromarray(img1.astype("uint8")) | |
| W_A, H_A = img0.size | |
| W_B, H_B = img1.size | |
| # Match | |
| warp, certainty = self.net.match(img0, img1, device=device) | |
| # Sample matches for estimation | |
| matches, certainty = self.net.sample( | |
| warp, certainty, num=self.conf["max_keypoints"] | |
| ) | |
| kpts1, kpts2 = self.net.to_pixel_coordinates(matches, H_A, W_A, H_B, W_B) | |
| pred = { | |
| "keypoints0": kpts1, | |
| "keypoints1": kpts2, | |
| "mconf": certainty, | |
| } | |
| return pred | |