File size: 4,426 Bytes
a6e2ff7
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
c44229f
a6e2ff7
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
f1e24de
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
---
pretty_name: Diogenes Fallout 4 Depth and Pose Archive v02
license: other
license_name: original-capture-archive-terms
license_link: https://huggingface.co/datasets/DiogenesLab/Diogenes_Fallout4_Depth_Archive_v02/blob/main/NOTICE.md
tags:
- video
- camera-pose
- archival
- non-trainable
size_categories:
- n<1K
---

# Diogenes Fallout 4 Depth and Pose Archive v02

This revision preserves two historical 120-second Commonwealth depth/camera capture packages and their spatial visualizations. Explicit publication copies mask a player-entered name during a brief menu animation. Unchanged captures, raw depth, action-camera arrays and historical QC records retain their original bytes and eligibility limits. **All 40 objects / 10,160,351,015 publication bytes are archived and verified across five complete TAR files.** [ARCHIVE_INDEX.json](ARCHIVE_INDEX.json) and [MANIFEST.jsonl](MANIFEST.jsonl) record current hashes, sizes and fixed revisions. Five original objects have explicit publication copies; two publication records were added. The first capture and its original ZIP remain byte-for-byte unchanged.

## Scientific content and conventions

Each original capture has 3,600 1920×1080 RGB video frames at 30 fps, with 1,800 1280×720 single-channel float16 EXR depth frames at 15 fps. The dataset contains video/audio, original 15/30 fps action-camera arrays, system information and source-frame/QC evidence. Outer ZIP bundles duplicate their extracted package representations; they are not additional recordings.

Depth is linear view-space Z in metres, with NaN invalid/sky pixels. Intrinsics calibrated for 1920×1080 must be scaled by 2/3 for depth rasters. Stored quaternions are world-to-camera, so back-projection uses their transpose/conjugate. First-person viewmodel depths are roughly 0.21–1.0 m; FP16 precision does not promise centimetre resolution beyond about 16 m.

The packages were reconstructed from longer recordings and include historical temporal jumps. Original source-frame selections and differing accounting descriptions remain unchanged. Publication does not replace them with an invented unified count or upgrade failed, missing, non-trainable or certification flags.

## Publication copies

The `_02` RGB and overlay videos have small opaque rectangles following the identified player-name slot during the opening menu animation. Every frame and presentation timestamp remains. All decoded YUV pixels outside those masks match the original exactly, and existing audio packet payloads/timing are retained. The corresponding outer ZIP replaces those two videos and adds `PUBLICATION.json` and `MANIFEST.public.json` while retaining all other original members byte-for-byte. Historical `sha256.json` and QC records continue to describe original artifacts; use the new publication manifest for current bytes. Do not treat modified media as retaining an original signature or hash attestation.

Embedded viewer and overview media are reviewed on the same source-frame basis. Original geometric outputs are point clouds and incomplete corridor TSDF meshes, not exported game assets. First-person geometry was filtered from these derivatives; original reconstruction limits and dynamic-object ghost trails remain.

## Rights and limitations

Underlying Fallout 4 content belongs to Bethesda and its licensors. The project owner authorized this closeout archive; no new commercial, game-content or model-training rights are granted. Viewer software includes Three.js r147 under MIT. See [NOTICE.md](NOTICE.md).

Review combines full-frame OCR for the identified name, direct visual inspection of its menu animation, complete frame/pixel/timestamp preservation checks and representative surrounding scenes. It is not universal identity detection, exhaustive semantic audio review or a new quality certification. Historical non-trainable and eligibility states remain unchanged. Source/schema code: [Diogenes-Archive](https://github.com/CuddlerAI/Diogenes-Archive).

The repaired spatial overview and embedded viewer preserve their complete presentation timing and audio packet bytes. Point-cloud/mesh content remains unchanged: the identified menu-name pixels lie in near-plane viewmodel depths excluded from these reconstructions; geometric inspection and source-depth checks support that scope-specific finding. No claim is made that a color-key heuristic alone excluded every menu frame.