LibreDepthAnythingV2l-depth

Depth Anything V2 ViT-L encoder (335M params) + DPT head, repackaged for LibreYOLO's 'depth' task.

Non-commercial use only. These weights are CC BY-NC 4.0. The Small encoder (LibreDepthAnythingV2s-depth) is Apache-2.0 and has no commercial restriction.

Source

Derived from DepthAnything/Depth-Anything-V2 at commit 03876f8651c73a60fe4c2c48294e09fcb6838fcf. Copyright (c) 2024 Depth Anything V2 authors (TikTok / ByteDance). Licensed under the Creative Commons Attribution-NonCommercial 4.0 International License (CC BY-NC 4.0).

The ViT-L encoder is a DINOv2 backbone from facebookresearch/dinov2. Copyright (c) Meta Platforms, Inc. and affiliates. Licensed under the Apache License, Version 2.0.

Modifications

State-dict key remapping only. Learned parameters are unchanged. See weights/convert_depth_anything_v2_weights.py in the LibreYOLO source repository.

Usage

from libreyolo import LibreYOLO

model = LibreYOLO("LibreYOLO/LibreDepthAnythingV2l-depth")
results = model.predict("image.jpg")
depth_map = results[0].depth_map  # DepthMap(shape=(H, W), higher = closer)

License

Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0). See the LICENSE and NOTICE files in this repository. Commercial use is not permitted.