| import os |
| import cv2 |
| import numpy as np |
| import torch |
|
|
| from einops import rearrange |
| from .zoedepth.models.zoedepth.zoedepth_v1 import ZoeDepth |
| from .zoedepth.utils.config import get_config |
| from modules import devices |
| from annotator.annotator_path import models_path |
|
|
|
|
| class ZoeDetector: |
| model_dir = os.path.join(models_path, "zoedepth") |
|
|
| def __init__(self): |
| self.model = None |
| self.device = devices.get_device_for("controlnet") |
|
|
| def load_model(self): |
| remote_model_path = "https://huggingface.co/lllyasviel/Annotators/resolve/main/ZoeD_M12_N.pt" |
| modelpath = os.path.join(self.model_dir, "ZoeD_M12_N.pt") |
| if not os.path.exists(modelpath): |
| from basicsr.utils.download_util import load_file_from_url |
| load_file_from_url(remote_model_path, model_dir=self.model_dir) |
| conf = get_config("zoedepth", "infer") |
| model = ZoeDepth.build_from_config(conf) |
| model.load_state_dict(torch.load(modelpath, map_location=model.device)['model']) |
| model.eval() |
| self.model = model.to(self.device) |
|
|
| def unload_model(self): |
| if self.model is not None: |
| self.model.cpu() |
|
|
| def __call__(self, input_image): |
| if self.model is None: |
| self.load_model() |
| self.model.to(self.device) |
|
|
| assert input_image.ndim == 3 |
| image_depth = input_image |
| with torch.no_grad(): |
| image_depth = torch.from_numpy(image_depth).float().to(self.device) |
| image_depth = image_depth / 255.0 |
| image_depth = rearrange(image_depth, 'h w c -> 1 c h w') |
| depth = self.model.infer(image_depth) |
|
|
| depth = depth[0, 0].cpu().numpy() |
|
|
| vmin = np.percentile(depth, 2) |
| vmax = np.percentile(depth, 85) |
|
|
| depth -= vmin |
| depth /= vmax - vmin |
| depth = 1.0 - depth |
| depth_image = (depth * 255.0).clip(0, 255).astype(np.uint8) |
|
|
| return depth_image |
|
|
