Have a ComfyUI workflow? Get a readable Workflow Doc in your browser.Make a workflow readable.Try it now ↗
Master's thesis: Woodland Segmentation in High-Resolution Aerial Imagery Using Few-Shot Learning.
Model source
Source description
Master's thesis: Woodland Segmentation in High-Resolution Aerial Imagery Using Few-Shot Learning.
Source code: github.com/zmgul/few-shot-woodland-segmentation
Prototypical network (ProtoNet) based on , with woodland as the novel class. 4 checkpoints published — one per experiment, each being the best-fold weights from 5-fold cross-validation.
Sources
1 sourceVerified Aug 25
Model artifacts
4 artifactsSource excerpts
2 excerpts| Experiment | Configuration | Checkpoint | fgIoU |
|---|---|---|---|
| 1 — Backbone | Swin-T (frozen, ImageNet) | best_ProtoNet_SwinT_none_fold1_seed-42.pt | 0.603 |
| 2 — Layer unfreezing (CNN) | ResNet-50 layer3 | best_ProtoNet_ResNet50_layer3_fold4_seed-42.pt | 0.587 |
| 3 — Stage unfreezing (ViT) | Swin-T stage3 | best_ProtoNet_SwinT_ImageNetV1_stage3_fold1_seed-42.pt | 0.605 |
| 4 — Pre-training ★ | SeCo + ResNet-50 (frozen) | best_ProtoNet_ResNet50_SeCo_none_fold1_seed-42.pt | 0.656 |
Key finding: Pre-training source dominates other factors — self-supervised SeCo outperforms all supervised sources, including domain-aligned Million-AID. Δ_pretrain = 0.103 (≈ 2× backbone effect, ≈ 8× fine-tuning effect).
LandCover.ai v1 — 41 high-resolution aerial images (25–50 cm/pixel) over Poland. 4 labeled land-cover classes + background. Woodland (33.3%) is held out as the novel class; building, water, road are base classes used for episodic training.
import torch
from huggingface_hub import hf_hub_download
from src.model import ProtoNet # from GitHub repo
# Backbone must match the checkpoint; pretrained init is overwritten by load_state_dict
model = ProtoNet(
backbone_name="resnet50",
pretrained="imagenet_v1",
unfreeze_from="none",
)
ckpt_path = hf_hub_download(
repo_id="zmgul/few-shot-woodland-segmentation",
filename="experiments/checkpoints/best_ProtoNet_ResNet50_SeCo_none_fold1_seed-42.pt",
)
state = torch.load(ckpt_path, map_location="cpu")
model.load_state_dict(state["model_state_dict"])
model.eval()
# 5 support tiles + masks, 1 query tile → binary woodland mask
# See notebooks/05_qualitative_results.ipynb for end-to-end inference example.
@mastersthesis{gul2026woodland,
author = {Gül, Zehra Merve},
title = {Woodland Segmentation in High-Resolution Aerial Imagery Using Few-Shot Learning},
school = {İstanbul University, Institute of Graduate Studies in Sciences},
year = {2026},
type = {Master's thesis}
}
MIT (code). LandCover.ai dataset and external pre-trained weights are subject to their original licenses.
experiments/checkpoints/best_ProtoNet_SwinT_ImageNetV1_stage3_fold1_seed-42.pt
pt · 108 MB · SHA-256 6ee1d7835ef4…65af · Hugging Face
Downloadexperiments/checkpoints/best_ProtoNet_SwinT_none_fold1_seed-42.pt
pt · 108 MB · SHA-256 d26d8cff9a22…5671 · Hugging Face
Download--- license: mit language: en library_name: pytorch tags: - few-shot-learning - semantic-segmentation - remote-sensing - aerial-imagery - prototypical-networks - landcover-ai pipeline_tag: image-segmentation datasets: - landcover-ai --- # Few-Shot Woodland Segmentation **Master's thesis:** Woodland Segmentation in High-Resolution Aerial Imagery Using Few-Shot Learning. **Source code:** [github.com/zmgul/few-shot-woodland-segmentation](https://github.com/zmgul/few-shot-woodland-segmentation) Prototypical network (ProtoNet) based **1-way 5-shot binary segmentation** on [LandCover.ai v1](https://landcover.ai/), with woodland as the novel class. 4 checkpoints published — one per experiment, each being the best-fold weights from 5-fold cross-validation. ## Results | Experiment | Configuration | Checkpoint | fgIoU | |---|---|---|---| | 1 — Backbone | Swin-T (frozen, ImageNet) | `best_ProtoNet_SwinT_none_fold1_seed-42.pt` | 0.603 | | 2 — Layer unfreezing (CNN) | ResNet-50 layer3 | `best_ProtoNet_ResNet50_layer3_fold4_seed-42.pt` | 0.587 | | 3 — Stage unfreezing (ViT) | Swin-T stage3 | `best_ProtoNet_SwinT_ImageNetV1_stage3_fold1_seed-42.pt` | 0.605 | | 4 — Pre-training ★ | SeCo + ResNet-50 (frozen) | `best_ProtoNet_ResNet50_SeCo_none_fold1_seed-42.pt` | **0.656** | **Key finding:** Pre-training source dominates other factors — self-supervised **SeCo** outperforms all supervised sources, including domain-aligned Million-AID. Δ_pretrain = 0.103 (≈ 2× backbone effect, ≈ 8× fine-tuning effect). ## Dataset [LandCover.ai v1](https://landcover.ai/) — 41 high-resolution aerial images (25–50 cm/pixel) over Poland. 4 labeled land-cover classes + background. Woodland (33.3%) is held out as the **novel class**; building, water, road are **base classes** used for episodic training. ## Usage ```python import torch from huggingface_hub import hf_hub_download from src.model import ProtoNet # from GitHub repo # Backbone must match the checkpoint; pretrained init is overwritten by load_state_dict model = ProtoNet( backbone_name="resnet50", pretrained="imagenet_v1", unfreeze_from="none", ) ckpt_path = hf_hub_download( repo_id="zmgul/few-shot-woodland-segmentation", filename="experiments/checkpoints/best_ProtoNet_ResNet50_SeCo_none_fold1_seed-42.pt", ) state = torch.load(ckpt_path, map_location="cpu") model.load_state_dict(state["model_state_dict"]) model.e...
Source context: 0 downloads · 0 likes · Pipeline image-segmentation · Library pytorch · Repo zmgul/few-shot-woodland-segmentation