Thank you for your open-source work. I'm working on running the CoT data from scratch and need to use Depth Anything V2 to generate the depth maps ourselves. The extracted archive only contains the original images, without any depth maps. My question is: should the required depth maps here be in the range of 0 to 255? Are they generated through the following calculations:
depth = depth_anything.infer_image(raw_image, args.input_size)
depth = (depth - depth.min()) / (depth.max() - depth.min()) * 255.0
depth = depth.astype(np.uint8)
depth = (cmap(depth)[:, :, :3] * 255)[:, :, ::-1].astype(np.uint8)?
Thank you for your open-source work. I'm working on running the CoT data from scratch and need to use Depth Anything V2 to generate the depth maps ourselves. The extracted archive only contains the original images, without any depth maps. My question is: should the required depth maps here be in the range of 0 to 255? Are they generated through the following calculations:
depth = depth_anything.infer_image(raw_image, args.input_size)
depth = (depth - depth.min()) / (depth.max() - depth.min()) * 255.0
depth = depth.astype(np.uint8)
depth = (cmap(depth)[:, :, :3] * 255)[:, :, ::-1].astype(np.uint8)?