---

license: apache-2.0 pipeline_tag: image-segmentation tags:

• remote-sensing

• semantic-segmentation

• rooftop-segmentation

• boundary-aware

• sam2

• dinov2

---

FDS-UNeXT

Checkpoints for "Boundary-Aware Adaptation of Frozen Foundation Models for Fine-Grained Rooftop Segmentation" (submitted to Remote Sensing).

FDS-UNeXT adapts two frozen foundation models, the SAM 2.1 Hiera-L encoder and DINOv2 ViT-L/14, to four-class rooftop material segmentation with 6.97 M trainable parameters (1.33 % of 523.1 M). Code: https://github.com/Jim-myoenk/FDS-UNeXT.

Format: trainable tensors only

A full checkpoint is 2.0 GB, but 516.2 M of its 523.1 M parameters are the frozen trunks and are identical to the public SAM 2.1 and DINOv2 weights. Each file here stores only the 6.97 M tensors that training updated, which makes it 26.7 MB. The round trip is exact: merging reproduces every tensor of the original file.

You still need the two public trunks:

wget -P weights https://dl.fbaipublicfiles.com/segment_anything_2/092824/sam2.1_hiera_large.pt
wget -P weights https://dl.fbaipublicfiles.com/dinov2/dinov2_vitl14/dinov2_vitl14_pretrain.pth

Files

File	Trained on	Used for
lrmd-fold0.pth … lrmd-fold4.pth	LRMD, 0.14 m, five folds	Table 3, Figure 10
potsdam-seed42.pth, potsdam-seed3407.pth	ISPRS Potsdam, 0.05 m	Table 4, Figure 11
egcs-seed42.pth, egcs-seed3407.pth, egcs-seed666.pth	E-GCS, 0.3 m	Table 4, Figures 9 and 11

Usage

huggingface-cli download Jim-myoenk/FDS-UNeXT lrmd-fold0.pth --local-dir deltas

python tools/merge_trainable.py \
  --delta deltas/lrmd-fold0.pth \
  --hiera_path weights/sam2.1_hiera_large.pt \
  --dinov2_path weights/dinov2_vitl14_pretrain.pth \
  --out runs/fold0/FDS-UNeXT-best.pth

python eval.py \
  --hiera_path weights/sam2.1_hiera_large.pt \
  --dinov2_path weights/dinov2_vitl14_pretrain.pth \
  --checkpoint runs/fold0/FDS-UNeXT-best.pth \
  --split_path dataset/roof_dataset_v2/split_fold0.json --split test \
  --num_classes 4 --bf_tolerance 8 --dy_edge_offset 6 \
  --class_names background,high-albedo-steel,low-albedo-steel,tile

Classes: 0 background (concrete roofs, asphalt, soil, vegetation, roads, water, vehicles), 1 high-albedo color-coated steel, 2 low-albedo color-coated steel, 3 tile.

Results

Dataset	mIoU (%)	Boundary F-score (%)
LRMD, five folds	70.00 ± 6.58	72.92 ± 3.69
Potsdam, two seeds	75.56 ± 0.49	72.67 ± 0.67
E-GCS, three seeds	95.15 ± 0.37	75.67 ± 1.14

The boundary F-score uses an 8-pixel tolerance. Metrics skip classes that are absent from the reference tile.

Citation

@article{zhao2026fdsunext,
  title   = {Boundary-Aware Adaptation of Frozen Foundation Models for
             Fine-Grained Rooftop Segmentation},
  author  = {Zhao, Yuancheng and Wang, Jie and Li, Faguo and Liu, Erte and
             Meng, Xiangying and Chen, Yueyue},
  journal = {Remote Sensing},
  year    = {2026},
  note    = {Under review}
}

License: Apache-2.0 for this model and for the two upstream trunks.