UniDepth V2 ViT-L/14 — Finetuned on RSRD for Pothole Depth

Finetuned from lpiccinelli/unidepth-v2-vitl14 on the RSRD dataset for metric depth estimation of road surface depressions.

Usage

from unidepth.models import UniDepthV2
import torch

model = UniDepthV2.from_pretrained("lpiccinelli/unidepth-v2-vitl14")
ckpt = torch.load("checkpoint_3500.pth", map_location="cpu")
model.load_state_dict(ckpt["model"], strict=False)

Training

Dataset: RSRD-dense (~2,800 samples)
Hardware: NVIDIA A100 80GB
Steps: 5,000
Base model: UniDepth V2 ViT-L/14

License

CC BY-NC 4.0 (same as base model)

Citation

RAIA Fellowship — USP / UFBA

Downloads last month: -; Downloads are not tracked for this model. How to track

Inference Providers NEW

Depth Estimation

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for 10ge6/unidepth-v2-vitl14-rsrd-pothole

Base model

lpiccinelli/unidepth-v2-vitl14

Finetuned

(1)

this model