Model files for Nightfall Meteor's optional Video Depth Anything Small (518x294) host depth model. Meteor downloads these itself when you choose "Download Video Depth Anything" in its tray (Model menu). You don't need to download them by hand.
This isn't a Nightfall release. The app is in the latest release.
| File | Size | SHA-256 |
|---|---|---|
vda_s_518x294.onnx.data
| 115,901,952 bytes | f5cb22053bfeacfdc175a12ed12efcddecd5a20c827994c0a25567c898531a32
|
vda_s_518x294_step.onnx
| 10,354,815 bytes | 0982c66a3ec3f6551ef968f52063ab9132c096d254e184b6bc30a096ddad86de
|
vda_s_518x294_cold.onnx
| 275,651 bytes | 4048448bc7f2e056f938caf4188ddc7b509768320bb4e5dd294d62ae6c88f262
|
vda_s_518x294_step.onnx: the recurrent step, run on every frame with 31 earlier states.vda_s_518x294_cold.onnx: the cold start, run on the first frame after a reset to seed those states.vda_s_518x294.onnx.data: both graphs' weights, stored once as ONNX external data.
Source and licence: Video Depth Anything Small by ByteDance (Sili Chen, Hengkai Guo, Shengnan Zhu, Feihu Zhang, Zilong Huang, Jiashi Feng, Bingyi Kang), from the official Small checkpoint, which is Apache-2.0 (LICENSE). The Base and Large models are non-commercial and aren't used. Nightfall's model research exported it as a causal streaming model at 518x294 for TensorRT. The weights are unchanged. meteor/models/share_vda_weights.py repacks the two exported graphs to share one weights file, writing these exact bytes.