Caption Quality Test for Cross-View Geo-Localization

Validate whether generated text captions improve retrieval R@1 in cross-view geo-localization (drone-to-satellite).

Architecture

Overview

Asymmetric dual-encoder with domain-specific frozen backbones and hierarchical text fusion for drone-to-satellite image retrieval.

┌──────────────────────────── QUERY BRANCH ────────────────────────────┐
│                                                                       │
│  drone_img ──► DINOv3 ViT-L/16 LVD ──► CLS token                    │
│  [B,3,256,256]   (frozen, 303M)         [B,1024]                     │
│                                            │                          │
│                                      proj_drone: Linear(1024,512)    │
│                                            │                          │
│                                        d_img [B,512]                 │
│                                            │                          │
│  L1 (overview)    ──► DGTRS-CLIP ──► z₁ [B,768] ─┐                   │
│  L2 (full desc)   ──► DGTRS-CLIP ──► z₂ [B,768] ─┼─ cat ──► [B,2304]│
│  L3 (fingerprint) ──► DGTRS-CLIP ──► z₃ [B,768] ─┘    │             │
│      (248 tokens, KPS pos. emb.)              MLP(2304→768→512)      │
│                                                    │                  │
│                                                d_txt [B,512]         │
│                                                    │                  │
│           q = σ(α)·d_img + (1−σ(α))·d_txt   GatedFusion             │
│                          │                                            │
│                    q̂ = q / ‖q‖₂  ──► query [B,512]                   │
└───────────────────────────────────────────────────────────────────────┘

┌──────────────────────────── GALLERY BRANCH ──────────────────────────┐
│                                                                       │
│  sat_img ──► DINOv3 ViT-L/16 SAT ──► CLS token                      │
│  [B,3,256,256]  (frozen, 303M)        [B,1024]                       │
│                                          │                            │
│                                    proj_sat: Linear(1024,512)        │
│                                          │                            │
│                                   ĝ = g / ‖g‖₂  ──► gallery [B,512] │
└───────────────────────────────────────────────────────────────────────┘

BASELINE: σ(α) = 1.0  →  q = d_img  (text branch disabled, DGTRS not loaded)

Text hierarchy (L1 / L2 / L3)

Each drone image has a VLM-generated caption (Qwen3-VL) split into 3 levels:

Level	Name	Content	Typical length
L1	Overview	First sentence of P1: land-cover summary with class percentages	15–30 tokens
L2	Full description	Complete P1 (inventory) + P2 (spatial layout with 5+ zones)	100–200 tokens
L3	Fingerprint	P3: unique landmarks and spatial signature for matching	20–50 tokens

All three levels are encoded by a single DGTRS-CLIP ViT-L-14 text encoder (248-token context via KPS positional embedding, 768-dim output).

Text fusion:

z_text = MLP( [z₁ ; z₂ ; z₃] )

where [z₁ ; z₂ ; z₃] ∈ ℝ^(B×2304)  — concatenation of three 768-dim embeddings
      MLP: Linear(2304, 768) → GELU → Linear(768, 512)
      z_text ∈ ℝ^(B×512)

Gated fusion:

q = σ(α) · d_img + (1 − σ(α)) · d_txt

where α — learnable scalar in logit-space (init: σ(α) ≈ 0.7)
      σ — sigmoid function
      d_img — projected drone image embedding [B, 512]
      d_txt — fused text embedding [B, 512]

Loss function

Symmetric InfoNCE with learnable temperature (CLIP-style logit_scale):

L = w_q2g · L_q→g + w_g2q · L_g→q

L_q→g = CrossEntropy( q̂ · ĝᵀ / τ,  targets )
L_g→q = CrossEntropy( ĝ · q̂ᵀ / τ,  targets )

where τ = 1 / exp(logit_scale),  logit_scale — learnable scalar
      τ ∈ [0.01, 0.5]  (clamped)
      τ_init = 0.07
      w_q2g = 0.6,  w_g2q = 0.4
      targets = [0, 1, 2, ..., B−1]  (positives on diagonal)
      label_smoothing = 0.1

Loss is computed in fp32 (outside AMP autocast) to prevent gradient overflow in the learnable temperature.

Metrics

Metric	Formula	Direction
R@K (Recall at K)	fraction of queries where correct gallery is in top-K	drone → satellite (primary)
Delta R@1	R@1(with_text) − R@1(baseline)	higher = text helps

Reported: R@1, R@5, R@10 for both q→g and g→q directions.

Optimizer & scheduler

Optimizer: AdamW
  - Projection heads (proj_drone, proj_sat, TextFusionMLP, gate α, logit_scale):
      lr = 1e-4,  weight_decay = 1e-4
  - DGTRS text encoder (last resblock + ln_final + text_projection):
      lr = 1e-5  (10× lower, --text-lr-factor 0.1)

Scheduler: Linear warmup (2 epochs) + cosine annealing
  - Per-step (not per-epoch)
  - warmup: lr linearly ramps from 0 to lr_max over warmup_steps
  - cosine: lr decays from lr_max to 0 over remaining steps

Gradient clipping: max_norm = 1.0
Mixed precision: AMP fp16 for model forward, fp32 for loss

Augmentations

Transform	Drone (train)	Satellite (train)	Eval
RandomResizedCrop(256, scale=0.7–1.0)	✓	✓	—
Resize(256) + CenterCrop(256)	—	—	✓
RandomHorizontalFlip(0.5)	✓	✓	—
RandomRotation(15°)	✓	—	—
ColorJitter(0.3, 0.3, 0.2, 0.1)	✓	✓	—
RandomGrayscale(0.05)	✓	✓	—
GaussianBlur(k=3, σ=0.1–2.0)	✓	—	—
ImageNet Normalize	✓	✓	✓

Model summary

Component	Params	Trainable	Notes
DINOv3 ViT-L/16 LVD (drone)	303M	0	frozen
DINOv3 ViT-L/16 SAT (satellite)	303M	0	frozen
DGTRS-CLIP ViT-L-14 (text)	124M	~7.6M	last block + ln_final + text_projection
proj_drone	524K	524K	Linear(1024, 512)
proj_sat	524K	524K	Linear(1024, 512)
TextFusionMLP	2.2M	2.2M	Linear(2304,768) + GELU + Linear(768,512)
GatedFusion α	1	1	scalar
logit_scale	1	1	learnable temperature
Total	733M	10.9M (1.49%)

Experiments

V3 — GTA-UAV + DINOv3 + DGTRS-CLIP (active)

Dataset: GTA-UAV-LR (33K drone + 14K satellite, GTA V synthetic)

• RGB: /home/servml/Документы/datasets/GTA-UAV-LR/
• Captions: /home/servml/Документы/datasets/GTA-UAV-LR-captions/ (40K JSON, 3-paragraph VLM)
• Segmentation: /home/servml/Документы/datasets/GTA-UAV-LR-aug/ (17 classes)
• Seg filter: 37,498 passed / 10,905 excluded (>=90% background+water)
• Split: 80/20 random (26,966 train / 6,742 test → 24,891/6,252 after seg filter)

V2 — UAV-GeoLoc + GeoRSCLIP (legacy)

Single backbone with template captions.

Query:   drone_img + caption -> GeoRSCLIP -> GatedFusion -> query
Gallery: sat_img -> GeoRSCLIP -> gallery

Dataset: UAV-GeoLoc Terrain split (206K train queries)

Structure

caption-test/
├── conf/                         # Gin configs (v2)
│   ├── balanced.gin
│   ├── baseline_no_text.gin
│   └── text_heavy.gin
├── nn_models/                    # Pre-trained checkpoints (v3, gitignored)
│   ├── DINO_WEB/                 # DINOv3 ViT-L/16 LVD-1689M (.pth)
│   ├── DINO_SAT/                 # DINOv3 ViT-L/16 SAT-493M (.safetensors)
│   └── LRSCLIP/                  # DGTRS-CLIP ViT-L-14 (.pt)
├── meta/                         # Generated metadata
│   ├── train_80.json             # 80% train split (26,966 pairs)
│   ├── test_20.json              # 20% test split (6,742 pairs)
│   └── seg_filter.json           # Segmentation filter results
├── scripts/
│   ├── make_split.py             # Create 80/20 train/test split
│   ├── filter_segmentation.py    # Exclude 90%+ background/water images
│   ├── compare_runs.py           # Delta R@1 comparison report
│   └── generate_captions.py      # Offline caption generation
├── src/
│   ├── datasets/
│   │   ├── gtauav_dataset.py     # GTA-UAV-LR loader + L1/L2/L3 parsing (v3)
│   │   └── visloc_with_captions.py  # UAV-GeoLoc loader (v2)
│   ├── models/
│   │   ├── dgtrs/                # Official DGTRS-CLIP text encoder (Apache-2.0)
│   │   │   ├── model.py          # DGTRSTextEncoder, build_model, tokenize
│   │   │   ├── simple_tokenizer.py  # BPE tokenizer (248 tokens)
│   │   │   └── bpe_simple_vocab_16e6.txt.gz
│   │   ├── asymmetric_encoder.py # DINOv3 + DGTRS + GatedFusion (v3)
│   │   └── dual_encoder.py       # GeoRSCLIP + GatedFusion (v2)
│   ├── losses/
│   │   └── multi_infonce.py      # InfoNCE with learnable temperature
│   ├── training/
│   │   ├── train_gtauav.py       # Training loop GTA-UAV (v3)
│   │   └── train.py              # Training loop UAV-GeoLoc (v2)
│   └── eval/
│       └── evaluate.py           # R@K metrics, Delta R@1
└── checkpoints/                  # GeoRSCLIP RS5M_ViT-B-32.pt (v2)

Prerequisites

torch>=2.0
safetensors
coloredlogs
tqdm
ftfy
regex
gin-config
Pillow
numpy

Workflow (V3 — GTA-UAV)

1. Create 80/20 split and filter segmentation

python -m scripts.make_split --output-dir meta
python -m scripts.filter_segmentation --output meta/seg_filter.json

2. Train baseline (no text)

python -m src.training.train_gtauav --baseline --filter-meta meta/seg_filter.json

3. Train with captions (L1/L2/L3)

python -m src.training.train_gtauav --filter-meta meta/seg_filter.json

4. Resume from checkpoint

python -m src.training.train_gtauav --resume out/gtauav/with_text/ckpt_epoch004.pt \
    --filter-meta meta/seg_filter.json

5. Compare and get verdict

python -m scripts.compare_runs \
    --baseline_report out/gtauav/baseline/eval_report.json \
    --full_report out/gtauav/with_text/eval_report.json \
    --output out/gtauav/comparison.md

Decision rule

Delta R@1 (drone→satellite)	Verdict
>= +3%	PASS — captions informative, proceed to NADEZHDA teacher
+1% to +3%	MARGINAL — add VLM refinement, re-run
0 to +1%	WEAK — redesign caption pipeline
< 0	HARMFUL — critical bug

Code style

• from __future__ import annotations everywhere
• Type hints on all signatures
• Google-style docstrings
• English-only comments