---
license: mit
library_name: libreyolo
tags:
  - depth-estimation
  - monocular-depth
  - midas
---

# LibreMiDaSs-depth

MiDaS EfficientNet-Lite3 (MiDaS v2.1 small, 256 px) monocular depth model, repackaged in LibreYOLO checkpoint format.

## Source

Derived from [isl-org/MiDaS](https://github.com/isl-org/MiDaS), release asset
`midas_v21_small_256.pt`.
Copyright (c) Intel ISL. Licensed under the MIT License.

## Modifications

State-dict key remapping only. Learned parameters are unchanged.
See `weights/convert_midas_weights.py` in the
[LibreYOLO source repository](https://github.com/LibreYOLO/libreyolo).

## A note on the training data

MiDaS was trained on a mixture of twelve datasets, several of which carry
non-commercial terms of their own. LibreYOLO redistributes these weights under
the MIT licence its publisher applied to them. If your use is commercial,
satisfy yourself about the training-data terms before relying on it.

## Usage

```python
from libreyolo import LibreYOLO

model = LibreYOLO("LibreMiDaSs-depth.pt")
depth = model.predict("image.jpg")[0].depth_map
```

## License

MIT License. See the [`LICENSE`](./LICENSE) and [`NOTICE`](./NOTICE) files in
this repository.