fuse_proj: Initial operational package for 3 researchers (Pavlenko/Blizno/Moroz)

Multimodal fusion research on StripNet+GTA-UAV proxy:
- 3 independent fusion tracks: condition-aware (A), token/bottleneck (B), role-aware (C)
- Shared interfaces, protocol, dataset audit, baseline benchmarks
- Canonical version-chain references to vault (SPEC, ANALYSIS, TRIAGE)
- Personalized task plans and decision tables for each researcher
- 3 generated DOCX task assignment files with milestones and DoD checklist
- Full modality dropout diagnostics and missing-modality robustness requirements
- Data contract, benchmark registry, experiment tracking infrastructure

Operational documents:
- docs/00_project/: MERIDIAN context, protocol, repository reuse guide, experiment specification
- docs/01_tasks/: Master assignment + 3 individual researcher tracks + joint integration
- docs/02_references/: Core literature, version-chain bases, code maps
- docs/03_codebase_guides/: Existing code snapshots from vault
- scripts/: gen_task_plans.js (DOCX generation), placeholder infrastructure
- vendor_reference/: Snapshots of caption_test, depth_edges_annotate, existing SOFIA/SegModel code
- reports/, results/, experiments/: Shared output structure for all 3 researchers

3 DOCX files generated from gen_task_plans.js (Times New Roman 14pt, GOST format):
- План_заданий_Павленко_БВ.docx (Condition-Aware track, fusion API owner)
- План_заданий_Близно_МВ.docx (Token/Bottleneck track, benchmark owner)
- План_заданий_Мороз_ЕС.docx (Role-Aware track, data contract owner)

Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>

in/config_files/README.md

@@ -0,0 +1,19 @@
+# Конфигурации
+
+Перед первым запуском заполнить:
+
+- `data.gin`: GTA-UAV RGB, captions, annotations и split JSON.
+- `model.gin`: StripNet checkpoint и fusion variant.
+- `training.gin`: одинаковые training settings для всех primary variants.
+- `hardware.gin`: hardware profile.
+- `external_projects.gin`: пути к двум исходным проектам.
+
+Персональные fusion configs добавляются отдельными файлами:
+
+```text
+fusion_condition_aware.gin
+fusion_token_bottleneck.gin
+fusion_role_aware.gin
+```
+
+Общие файлы не дублировать в персональных директориях. Каждый run сохраняет их копии.

in/config_files/data.gin

@@ -0,0 +1,9 @@
+# GTA-UAV paths. Fill these values on the target experiment machine.
+DataConfig.dataset_root = ''
+DataConfig.annotation_root = ''
+DataConfig.caption_root = ''
+DataConfig.train_json = ''
+DataConfig.test_json = ''
+DataConfig.segmentation_filter = ''
+DataConfig.image_size = 256
+DataConfig.num_segmentation_classes = 17

in/config_files/external_projects.gin

@@ -0,0 +1,3 @@
+# Reusable local source projects.
+ExternalProjectsConfig.caption_test_root = 'C:/Users/Lisadminipc/Documents/code/caption-test'
+ExternalProjectsConfig.annotation_project_root = 'C:/Users/Lisadminipc/Documents/code/depth_edges_annotate_worlduav'

in/config_files/hardware.gin

@@ -0,0 +1,6 @@
+# RTX 4090 reference profile.
+HardwareConfig.device = 'cuda'
+HardwareConfig.use_amp = True
+HardwareConfig.amp_dtype = 'float16'
+HardwareConfig.total_vram_gb = 24.0
+HardwareConfig.reserve_vram_gb = 2.0

in/config_files/model.gin

@@ -0,0 +1,7 @@
+# Shared model contract for fair fusion comparison.
+ModelConfig.stripnet_checkpoint = ''
+ModelConfig.descriptor_dim = 1024
+ModelConfig.fusion_variant = 'rgb_only'
+ModelConfig.text_encoder = 'dgtrs'
+ModelConfig.use_conv_mona = True
+ModelConfig.conv_mona_last_n_stages = 2

in/config_files/training.gin

@@ -0,0 +1,13 @@
+# Common optimization settings.
+TrainingConfig.output_root = 'out'
+TrainingConfig.seed = 42
+TrainingConfig.epochs = 10
+TrainingConfig.batch_size = 8
+TrainingConfig.num_workers = 4
+TrainingConfig.learning_rate = 0.0001
+TrainingConfig.weight_decay = 0.0001
+TrainingConfig.gradient_clip = 1.0
+TrainingConfig.tau_init = 0.07
+TrainingConfig.label_smoothing = 0.1
+TrainingConfig.weight_q2g = 0.6
+TrainingConfig.weight_g2q = 0.4