Upload folder using huggingface_hub (part 70)

LICENSE

@@ -0,0 +1,97 @@
+Community Data License Agreement - Sharing - Version 1.0
+
+SPDX-License-Identifier: CDLA-Sharing-1.0
+
+This is the Community Data License Agreement – Sharing, Version 1.0 ("Agreement"). Data is provided to You under this Agreement by each of the Data Providers. Your exercise of any of the rights and permissions granted below constitutes Your acceptance and agreement to be bound by the terms and conditions of this Agreement.
+
+The benefits that each Data Provider receives from making Data available and that You receive from Data or otherwise under these terms and conditions shall be deemed sufficient consideration for the formation of this Agreement. Accordingly, Data Provider(s) and You (the "Parties") agree as follows:
+
+Section 1. Definitions
+
+1.1 "Add" means to supplement Data with Your own or someone else's Data, resulting in Your "Additions." Additions do not include Results.
+
+1.2 "Computational Use" means Your analysis (through the use of computational devices or otherwise) or other interpretation of Data. By way of example and not limitation, "Computational Use" includes the application of any computational analytical technique, the purpose of which is the analysis of any Data in digital form to generate information about Data such as patterns, trends, correlations, inferences, insights and attributes.
+
+1.3 "Data" means the information (including copyrightable information, such as images or text), collectively or individually, whether created or gathered by a Data Provider or an Entity acting on its behalf, to which rights are granted under this Agreement.
+
+1.4 "Data Provider" means any Entity (including any employee or contractor of such Entity authorized to Publish Data on behalf of such Entity) that Publishes Data under this Agreement prior to Your Receiving it.
+
+1.5 "Enhanced Data" means the subset of Data that You Publish and that is composed of (a) Your Additions and/or (b) Modifications to Data You have received under this Agreement.
+
+1.6 "Entity" means any natural person or organization that exists under the laws of the jurisdiction in which it is organized, together with all other entities that control, are controlled by, or are under common control with that entity.
+
+For the purposes of this definition, "control" means (a) the power, directly or indirectly, to cause the direction or management of such entity, whether by contract or otherwise, (b) the ownership of more than fifty percent (50%) of the outstanding shares or securities, (c) the beneficial ownership of such entity or, (d) the ability to appoint, whether by agreement or right, the majority of directors of an Entity.
+
+1.7 "Ledger" means a digital record of Data or grants of rights in Data governed by this Agreement, using any technology having functionality to record and store Data or grants, contributions, or licenses to Data governed by this Agreement.
+
+1.8 "Modify" means to delete, erase, correct or re-arrange Data, resulting in "Modifications." Modifications do not include Results.
+
+1.9 "Publish" means to make all or a subset of Data (including Your Enhanced Data) available in any manner which enables its Use, including by providing a copy on physical media or remote access. For any form of Entity, that is to make the Data available to any individual who is not employed by that Entity or engaged as a contractor or agent to perform work on that Entity's behalf. A "Publication" occurs each time You Publish Data.
+
+1.10 "Receive" or "Receives" means to have been given access to Data, locally or remotely.
+
+1.11 "Results" means the outcomes or outputs that You obtain from Your Computational Use of Data. Results shall not include more than a de minimis portion of the Data on which the Computational Use is based.
+
+1.12 "Sui Generis Database Rights" means rights, other than copyright, resulting from Directive 96/9/EC of the European Parliament and of the Council of 11 March 1996 on the legal protection of databases, as amended and/or succeeded, as well as other equivalent rights anywhere in the world.
+
+1.13 "Use" means using Data (including accessing, copying, studying, reviewing, adapting, analyzing, evaluating, or making Computational Use of it), either by machines or humans, or a combination of both.
+
+1.14 "You" or "Your" means any Entity that Receives Data under this Agreement.
+
+Section 2. Right and License to Use and to Publish
+
+2.1 Subject to the conditions set forth in Section 3 of this Agreement, Data Provider(s) hereby grant(s) to You a worldwide, non-exclusive, irrevocable (except as provided in Section 5) right to: (a) Use Data; and (b) Publish Data.
+
+2.2 To the extent that the Data or the coordination, selection or arrangement of Data is protected or protectable under copyright, Sui Generis Database Rights, or other law, Data Provider(s) further agree(s) that such Data or coordination, selection or arrangement is hereby licensed to You and to anyone else who Receives Data under this Agreement for Use and Publication, subject to the conditions set forth in Section 3 of this Agreement.
+
+2.3 Except for these rights and licenses expressly granted, no other intellectual property rights are granted or should be implied.
+
+Section 3. Conditions on Rights Granted
+
+3.1 If You Publish Data You Receive or Enhanced Data:
+
+(a) The Data (including the Enhanced Data) must be Published under this Agreement in accordance with this Section 3; and
+
+(b) You must cause any Data files containing Enhanced Data to carry prominent notices that You have changed those files; and
+
+(c) If You Publish Data You Receive, You must preserve all credit or attribution to the Data Provider(s). Such retained credit or attribution includes any of the following to the extent they exist in Data as You have Received it: legal notices or metadata; identification of the Data Provider(s); or hyperlinks to Data to the extent it is practical to do so.
+
+3.2 You may not restrict or deter the ability of anyone who Receives the Data (a) to Publish the Data in a publicly-accessible manner or (b) if the project has designated a Ledger for recording Data or grants of rights in Data for purposes of this Agreement, to record the Data or grants of rights in Data in the Ledger.
+
+3.3 If You Publish Data You Receive, You must do so under an unmodified form of this Agreement and include the text of this Agreement, the name of this Agreement and/or a hyperlink or other method reasonably likely to provide a copy of the text of this Agreement.
+
+You may not modify this Agreement or impose any further restrictions on the exercise of the rights granted under this Agreement, including by adding any restriction on commercial or non-commercial Use of Data (including Your Enhanced Data) or by limiting permitted Use of such Data to any particular platform, technology or field of endeavor. Notices that purport to modify this Agreement shall be of no effect.
+
+3.4 You and each Data Provider agree that Enhanced Data shall not be considered a work of joint authorship by virtue of its relationship to Data licensed under this Agreement and shall not require either any obligation of accounting to or the consent of any Data Provider.
+
+3.5 This Agreement imposes no obligations or restrictions on Your Use or Publication of Results.
+
+Section 4. Data Provider(s)' Representations
+
+4.1 Each Data Provider represents that the Data Provider has exercised reasonable care, to assure that: (a) the Data it Publishes was created or generated by it or was obtained from others with the right to Publish the Data under this Agreement; and (b) Publication of such Data does not violate any privacy or confidentiality obligation undertaken by the Data Provider.
+
+Section 5. Termination
+
+5.1 All of Your rights under this Agreement will terminate, and Your right to Receive, Use or Publish the Data will be revoked or modified if You materially fail to comply with the terms and conditions of this Agreement and You do not cure such failure in a reasonable period of time after becoming aware of such noncompliance. If Your rights under this Agreement terminate, You agree to cease Receipt, Use and Publication of Data.
+
+However, Your obligations and any rights and permissions granted by You under this Agreement relating to Data that You Published prior to such termination will continue and survive.
+
+5.2 If You institute litigation against a Data Provider or anyone else who Receives the Data (including a cross-claim in a lawsuit) based on the Data, other than a claim asserting breach of this Agreement, then any rights previously granted to You to Receive, Use or Publish Data under this Agreement will terminate as of the date such litigation is filed.
+
+Section 6. Disclaimer of Warranties and Limitation of Liability
+
+6.1 EXCEPT AS EXPRESSLY SET FORTH IN THIS AGREEMENT, THE DATA (INCLUDING ENHANCED DATA) IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, EITHER EXPRESS OR IMPLIED INCLUDING, WITHOUT LIMITATION, ANY WARRANTIES OR CONDITIONS OF TITLE, NON-INFRINGEMENT, MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.
+
+6.2 NEITHER YOU NOR ANY DATA PROVIDERS SHALL HAVE ANY LIABILITY FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING WITHOUT LIMITATION LOST PROFITS), HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OR DISTRIBUTION OF THE DATA OR THE EXERCISE OF ANY RIGHTS GRANTED HEREUNDER, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGES.
+
+Section 7. Miscellaneous
+
+7.1 You agree that it is solely Your responsibility to comply with all applicable laws with regard to Your Use or Publication of Data, including any applicable privacy, data protection, security and export laws. You agree to take reasonable steps to assist a Data Provider fulfilling responsibilities to comply with applicable laws with regard to Use or Publication of Data Received hereunder.
+
+7.2 You and Data Provider(s), collectively and individually, waive and/or agree not to assert, to the extent permitted by law, any moral rights You or they hold in Data.
+
+7.3 This Agreement confers no rights or remedies upon any person or entity other than the Parties and their respective heirs, executors, successors and assigns.
+
+7.4 The Data Provider(s) reserve no right or expectation of privacy, data protection or confidentiality in any Data that they Publish under this Agreement. If You choose to Publish Data under this Agreement, You similarly do so with no reservation or expectation of any rights of privacy or confidentiality in that Data.
+
+7.5 The Community Data License Agreement workgroup under The Linux Foundation is the steward of this Agreement ("Steward"). No one other than the Steward has the right to modify or publish new versions of this Agreement. Each version will be given a distinguishing version number. You may Use and Publish Data Received hereunder under the terms of the version of the Agreement under which You originally Received the Data, or under the terms of any subsequent version published by the Steward.

LICENSE_SCOPE.md

@@ -0,0 +1,63 @@
+# CraftSight Data License Scope
+
+**Dataset version:** 1.0.0  
+**Primary data license:** Community Data License Agreement – Sharing, Version 1.0 (`CDLA-Sharing-1.0`)
+
+This file identifies which parts of the CraftSight release are provided as **Data** under the unmodified agreement in `LICENSE`. It does not alter, supplement, or restrict that agreement.
+
+## Covered Data
+
+To the extent that the dataset maintainers own or control the necessary rights, the following maintainer-created materials are provided under `CDLA-Sharing-1.0`:
+
+- annotation values and annotation tables;
+- maintainer-created metadata fields and normalized metadata tables;
+- label schemas, class lists, and label definitions;
+- train, validation, and test split assignments and manifests;
+- audit outputs and quality-control tables;
+- the selection, coordination, and arrangement of the covered metadata and annotations;
+- `metadata.csv`, `label_schema.json`, `label_definitions.md`, `classes_core.txt`, `classes_full.txt`, `all_images.txt`, `unlabeled.csv`, and the CSV/JSON files under `splits/`, except for any embedded third-party material identified below.
+
+The initial Data Provider is Konstantin Shaga.
+
+## Enhanced Data
+
+When publishing modified, corrected, rearranged, supplemented, or extended versions of the Covered Data:
+
+1. publish the Covered Data and Enhanced Data under the unmodified `CDLA-Sharing-1.0`;
+2. prominently identify changed data files;
+3. preserve existing credits, source identifiers, legal notices, and practical source hyperlinks;
+4. include the unmodified agreement or a reasonably accessible copy or link.
+
+## Results
+
+The CDLA does not impose obligations or restrictions on Results produced through Computational Use, provided those Results contain no more than a de minimis portion of the Covered Data.
+
+This statement concerns only the CDLA-covered layer. It does not override licenses or other terms governing source images, gameplay recordings, third-party metadata, trademarks, or Minecraft content.
+
+## Excluded Materials
+
+The following are not licensed under `CDLA-Sharing-1.0` by this repository:
+
+- Minecraft screenshots and gameplay frames;
+- source videos and recordings;
+- Minecraft textures, models, interfaces, trademarks, names, or other game content;
+- third-party metadata copied from an upstream dataset;
+- upstream dataset files governed by their own licenses;
+- software or utility scripts, unless a separate code license explicitly applies.
+
+These materials remain subject to their original licenses, notices, and other applicable rights. See `THIRD_PARTY_NOTICES.md`.
+
+## No Relicensing of Source Images
+
+A row pairing a CDLA-covered annotation with a separately licensed image does not relicense that image. Recipients must satisfy both:
+
+- `CDLA-Sharing-1.0` for the covered annotation and metadata layer; and
+- the applicable source-specific terms for the corresponding image.
+
+Where the source-specific terms are more restrictive, those restrictions govern use of the image, not the independently licensed annotation layer.
+
+## No Additional Restrictions on Covered Data
+
+The source-specific image notices do not add restrictions to Data actually covered by `CDLA-Sharing-1.0`. They identify materials outside the CDLA grant.
+
+If this scope notice conflicts with the unmodified agreement regarding Covered Data, the agreement in `LICENSE` controls.

README.md

@@ -1,3 +1,570 @@
----
-license: cdla-sharing-1.0
----
+---
+pretty_name: "CraftSight: A Multi-label Perception Dataset for Minecraft Agents"
+license: other
+license_name: "cdla-sharing-1.0-mixed"
+license_link: https://cdla.dev/sharing-1-0/
+task_categories:
+  - image-classification
+  - image-feature-extraction
+annotations_creators:
+  - expert-generated
+  - machine-generated
+size_categories:
+  - 10K<n<100K
+tags:
+  - computer-vision
+  - multi-label-classification
+  - image-feature-extraction
+  - representation-learning
+  - embodied-ai
+  - agent-perception
+  - minecraft
+  - reinforcement-learning
+  - gameplay
+configs:
+  - config_name: core
+    default: true
+    data_files:
+      - split: train
+        path: splits/core/train.csv
+      - split: validation
+        path: splits/core/val.csv
+      - split: test
+        path: splits/core/test.csv
+  - config_name: full
+    data_files:
+      - split: train
+        path: splits/full/train.csv
+      - split: validation
+        path: splits/full/val.csv
+      - split: test
+        path: splits/full/test.csv
+  - config_name: unlabeled
+    data_files:
+      - split: train
+        path: unlabeled.csv
+---
+
+# CraftSight
+
+**Frame-level multi-label visual annotations for Minecraft agents.**
+
+CraftSight provides Minecraft gameplay frames annotated with terrain, hazards, blocks, interfaces, structures, mobs, and scene-level context. It is intended for multi-label perception, representation learning, active learning, and embodied-agent research.
+
+> **Not an official Minecraft product. Not approved by or associated with Mojang or Microsoft.**
+
+## Dataset overview
+
+| Item | Count |
+|---|---:|
+| Indexed images | 64,279 |
+| Annotated images | 3,058 |
+| Active annotated images | 2,823 |
+| Ignored annotated images | 235 |
+| Unlabeled images | 61,221 |
+| Benchmark trajectories | 571 |
+| Core labels | 19 |
+| Full labels | 48 |
+| Rows with optional game state | 50 |
+
+CraftSight exposes three configurations:
+
+| Configuration | Rows | Targets | Recommended use |
+|---|---:|---:|---|
+| **`core`** | 2,823 | 19 labels | Standard training and comparable evaluation |
+| **`full`** | 2,823 | 48 labels | Long-tail, few-shot, active-learning, and exploratory work |
+| **`unlabeled`** | 61,221 | none | Self-supervised learning, embeddings, clustering, retrieval, and annotation expansion |
+
+`core` and `full` contain exactly the same active frames and use the same trajectory-level train/validation/test assignment. They differ only in their `label_*` columns.
+
+## Quick start
+
+```python
+from datasets import load_dataset
+
+dataset = load_dataset("Krows7/CraftSight", "core")
+
+print(dataset)
+print(dataset["train"][0])
+```
+
+`core` is the default configuration:
+
+```python
+dataset = load_dataset("Krows7/CraftSight")
+```
+
+Load all 48 labels:
+
+```python
+full = load_dataset("Krows7/CraftSight", "full")
+```
+
+Load the unlabeled image index:
+
+```python
+unlabeled = load_dataset(
+    "Krows7/CraftSight",
+    "unlabeled",
+    split="train",
+)
+```
+
+## Working with labels
+
+Every target is stored in a separate `label_<name>` column.
+
+| Value | Meaning |
+|---:|---|
+| `1` | The label is present |
+| `0` | The label was reviewed and is absent |
+| `-1` | The label is unknown or was not reviewed |
+
+**Never treat `-1` as a negative label.**
+
+```python
+label_columns = [
+    column
+    for column in dataset["train"].column_names
+    if column.startswith("label_")
+]
+```
+
+Example masked binary cross-entropy:
+
+```python
+import torch
+import torch.nn.functional as F
+
+# logits and targets have shape [batch_size, number_of_labels]
+known = targets != -1
+safe_targets = targets.clamp_min(0).float()
+
+loss_per_target = F.binary_cross_entropy_with_logits(
+    logits,
+    safe_targets,
+    reduction="none",
+)
+
+loss = loss_per_target[known].mean()
+```
+
+The canonical label order and numeric semantics are defined in [`label_schema.json`](label_schema.json). Class boundaries and edge cases are documented in [`label_definitions.md`](label_definitions.md).
+
+## Images
+
+`image_path` is relative to one shared image store. Images are not duplicated between:
+
+- `core` and `full`;
+- train, validation, and test;
+- labeled and unlabeled manifests.
+
+A local layout may look like:
+
+```text
+dataset-root/
+├── images/
+│   ├── basalt/
+│   ├── mobs/
+│   └── custom/
+├── metadata.csv
+├── unlabeled.csv
+└── splits/
+```
+
+Load a local image:
+
+```python
+from pathlib import Path
+from PIL import Image
+
+root = Path("/path/to/dataset-root")
+example = dataset["train"][0]
+
+image = Image.open(
+    root / "images" / example["image_path"]
+).convert("RGB")
+```
+
+The metadata package may be distributed separately from the image payload. Image provenance and permitted uses differ by `data_source`; review [`THIRD_PARTY_NOTICES.md`](THIRD_PARTY_NOTICES.md) before downloading, redistributing, or using images commercially.
+
+## Choosing a configuration
+
+### `core`
+
+Use `core` for:
+
+- headline benchmark results;
+- model comparison;
+- standard multi-label training;
+- threshold selection on validation data;
+- experiments requiring positive support for every target in every split.
+
+Core labels:
+
+```text
+water
+drop_off
+dark_cave
+sand_or_redsand
+tree_log
+leaves
+stone
+inventory_open
+hostile_near
+wood_planks
+cobblestone
+torch
+water_bucket_in_hand
+building
+hostile_present
+sky_visible
+open_surface
+mountains
+glass_pane_or_window
+```
+
+### `full`
+
+Use `full` for:
+
+- rare-label and long-tail studies;
+- few-shot experiments;
+- active learning;
+- label completion;
+- schema extension;
+- partially labeled learning.
+
+The full configuration preserves all 48 schema labels. Some classes have limited support, and `furnace` and `ladder` currently have no positive active examples.
+
+<details>
+<summary><strong>Show all 48 labels</strong></summary>
+
+```text
+water
+lava
+fire
+drop_off
+dark_cave
+web_cobweb
+ice
+sand_or_redsand
+gravel
+tree_log
+leaves
+stone
+crafting_table
+furnace
+chest
+bed
+inventory_open
+underwater
+hostile_near
+wood_planks
+coal_ore
+iron_ore
+cobblestone
+door
+ladder
+torch
+water_bucket_in_hand
+building
+player_damage
+hostile_present
+creeper
+zombie
+skeleton
+spider
+enderman
+witch
+cow
+sheep
+pig
+chicken
+villager
+sky_visible
+open_surface
+farmland
+crop
+ocean
+mountains
+glass_pane_or_window
+```
+
+</details>
+
+### `unlabeled`
+
+Use `unlabeled` for:
+
+- self-supervised pretraining;
+- image feature extraction;
+- embedding generation;
+- clustering and similarity search;
+- trajectory representation learning;
+- active-learning candidate selection.
+
+It contains only path and trajectory metadata:
+
+```text
+image_path
+data_source
+agent
+task
+seed
+episode
+frame_index
+trajectory_id
+```
+
+## Splits
+
+The supervised configurations use a leakage-resistant group split:
+
+| Split | Images | Trajectories |
+|---|---:|---:|
+| Train | 1,961 | 499 |
+| Validation | 426 | 38 |
+| Test | 436 | 34 |
+
+All frames from one `trajectory_id` remain in one split. No trajectory appears in more than one split.
+
+The canonical assignment is stored in:
+
+```text
+splits/core/splits_by_trajectory.csv
+```
+
+The `full` configuration reuses the same manifest. Do not create a new random frame-level split, because adjacent frames may be nearly identical.
+
+## Dataset structure
+
+### Repository layout
+
+```text
+.
+├── README.md
+├── LICENSE
+├── LICENSE_SCOPE.md
+├── THIRD_PARTY_NOTICES.md
+├── label_schema.json
+├── state_schema.json
+├── label_definitions.md
+├── classes_core.txt
+├── classes_full.txt
+├── metadata.csv
+├── all_images.txt
+├── unlabeled.csv
+└── splits/
+    ├── core/
+    │   ├── train.csv
+    │   ├── val.csv
+    │   ├── test.csv
+    │   ├── splits.csv
+    │   ├── splits_by_trajectory.csv
+    │   └── split_report.json
+    └── full/
+        ├── train.csv
+        ├── val.csv
+        ├── test.csv
+        ├── splits.csv
+        ├── splits_by_trajectory.csv
+        └── split_report.json
+```
+
+`metadata.csv` is the canonical annotation table and includes both active and ignored rows. Standard train/validation/test CSVs contain only active rows with `ignore = 0`.
+
+`all_images.txt` is the canonical list of all 64,279 image paths. `unlabeled.csv` contains exactly the 61,221 paths absent from `metadata.csv`.
+
+### Annotated row fields
+
+#### Identity and trajectory metadata
+
+| Column | Type | Description |
+|---|---|---|
+| `image_path` | string | Relative path in the shared image store |
+| `data_source` | string | `basalt`, `mobs`, or `custom` |
+| `agent` | nullable string | Agent or participant identifier |
+| `task` | nullable string | Task name |
+| `seed` | nullable integer | Trajectory seed |
+| `episode` | nullable string | Episode identifier |
+| `frame_index` | nullable integer | Frame index parsed from the source filename when available |
+| `trajectory_id` | string | Group used for leakage-safe splitting |
+
+#### Annotation metadata
+
+| Column | Type | Description |
+|---|---|---|
+| `annotation_method` | string | `manual`, `model_assisted`, `model_accept`, or `model_ignore` |
+| `ignore` | integer | `1` excludes the row from standard supervised splits |
+| `thr_mode` | nullable string | Model-assist threshold mode |
+| `thr_global` | nullable float | Global model-assist threshold |
+
+#### Optional game state
+
+| Column | Type | Description |
+|---|---|---|
+| `has_state` | integer | Whether state metadata is available |
+| `state_hp` | nullable integer | Player health |
+| `state_armor` | nullable integer | Armor points |
+| `state_hunger` | nullable integer | Hunger level |
+| `state_biome` | nullable string | Recorded biome |
+| `state_selected_slot` | nullable integer | Selected hotbar slot |
+| `state_held_item` | nullable string | Recorded held item |
+| `state_time_of_day` | nullable string | Coarse time-of-day category |
+
+The state schema is defined in [`state_schema.json`](state_schema.json). State is a separate modality and must not silently replace visual evidence.
+
+## Supported tasks
+
+CraftSight supports:
+
+- multi-label image classification;
+- image feature extraction;
+- visual representation learning;
+- scene and hazard recognition;
+- perception modules for embodied and reinforcement-learning agents;
+- active-learning and partially labeled learning research.
+
+CraftSight is **not currently an object-detection dataset**. It does not provide bounding boxes, instance masks, or object coordinates.
+
+## Evaluation
+
+Use `core` for comparable benchmark results.
+
+Recommended metrics:
+
+- macro F1;
+- micro F1;
+- mean average precision;
+- per-class average precision;
+- per-class precision and recall;
+- known-target and positive-example counts for each class.
+
+Select thresholds only on validation data. Mask all `-1` targets during training and evaluation.
+
+Do not report a class metric for a split with no known positive examples for that class.
+
+## Dataset creation
+
+### Image sources
+
+| Namespace | Source | Image terms |
+|---|---|---|
+| `mobs` | Minecraft Screenshots Dataset with Features by `sqdartemy` | CC BY-NC 4.0 |
+| `basalt` | BASALT Benchmark Evaluation Dataset, MineRL BASALT team, Zenodo record `8021960` | MIT according to the upstream record |
+| `custom` | Gameplay recordings captured by the dataset maintainer | Minecraft and other applicable terms |
+
+Image counts by namespace:
+
+| Namespace | Indexed | Annotated | Unlabeled |
+|---|---:|---:|---:|
+| `basalt` | 60,677 | 2,725 | 57,952 |
+| `mobs` | 3,545 | 276 | 3,269 |
+| `custom` | 57 | 57 | 0 |
+
+See [`THIRD_PARTY_NOTICES.md`](THIRD_PARTY_NOTICES.md) for source links, attribution requirements, and source-specific conditions.
+
+### Annotation process
+
+Active annotations were produced through:
+
+| Method | Active rows |
+|---|---:|
+| Manual | 2,742 |
+| Model-assisted | 54 |
+| Accepted model suggestion | 27 |
+
+Annotations are image-level and multi-label. Multiple labels may be positive in one frame. Annotation rules prioritize visible evidence from the current frame rather than inference from adjacent frames, task names, or hidden game state.
+
+### Quality control
+
+The tabular release was checked for:
+
+- schema and column-order consistency;
+- valid target values;
+- duplicate image paths;
+- path-derived metadata consistency;
+- state-field consistency;
+- split coverage;
+- trajectory leakage;
+- alignment between `core` and `full`;
+- positive-label support across core splits;
+- exact partitioning of annotated and unlabeled image paths.
+
+The current release has:
+
+- no duplicate annotated paths;
+- no overlap between `metadata.csv` and `unlabeled.csv`;
+- complete coverage of `all_images.txt`;
+- no trajectory leakage;
+- exact agreement between core and full split assignments.
+
+## Limitations
+
+- Labels are highly imbalanced.
+- `furnace` and `ladder` have no positive active examples.
+- Several full labels are rare or absent from individual validation/test splits.
+- Only 50 annotated rows contain game-state metadata.
+- Frames within a trajectory remain temporally correlated.
+- Sources, tasks, and agents are not uniformly represented.
+- Model-assisted annotations may inherit model errors.
+- Some semantic label relationships are not mechanically enforced in every row.
+- `frame_index` is unavailable for some source filenames.
+- Performance may change with game version, edition, texture pack, shaders, field of view, resolution, and UI scale.
+
+Users should report per-class support and inspect failure cases before drawing broad conclusions.
+
+## Licensing
+
+CraftSight uses a layered licensing model.
+
+### Annotation and metadata layer
+
+Maintainer-created annotations, normalized metadata, schemas, class lists, definitions, split manifests, and covered data organization are provided under **CDLA-Sharing-1.0**.
+
+- The unmodified agreement is in [`LICENSE`](LICENSE).
+- The covered repository layer is defined in [`LICENSE_SCOPE.md`](LICENSE_SCOPE.md).
+- Published modified or extended covered data must remain under the unmodified CDLA-Sharing-1.0.
+- Changed data files must be identified.
+- Existing attribution and practical source links must be preserved.
+
+The CDLA does not impose restrictions on qualifying computational Results, subject to the agreement's definitions and terms.
+
+### Image layer
+
+Images are not relicensed under CraftSight's CDLA grant:
+
+- `mobs` images are subject to CC BY-NC 4.0;
+- `basalt` images follow the terms stated by the upstream Zenodo record;
+- `custom` images remain subject to Minecraft and other applicable rights.
+
+See [`THIRD_PARTY_NOTICES.md`](THIRD_PARTY_NOTICES.md) before redistribution or commercial use.
+
+## Minecraft disclaimer
+
+**NOT AN OFFICIAL MINECRAFT DATASET. NOT APPROVED BY OR ASSOCIATED WITH MOJANG OR MICROSOFT.**
+
+Minecraft is a trademark of the Microsoft group of companies. Minecraft-related rights remain with Mojang, Microsoft, and their licensors.
+
+## Citation
+
+Until a paper citation is available, cite the repository and exact release or commit used:
+
+```bibtex
+@dataset{craftsight_2026,
+  author    = {Shaga, Konstantin},
+  title     = {CraftSight: A Multi-label Perception Dataset for Minecraft Agents},
+  year      = {2026},
+  publisher = {Hugging Face},
+  version   = {1.0.0},
+  url       = {https://huggingface.co/datasets/Krows7/CraftSight}
+}
+```
+
+## Maintenance
+
+- Dataset version: `1.0.0`
+- Data provider: Konstantin Shaga
+
+Use the repository Discussions tab for annotation corrections, provenance updates, licensing questions, and schema proposals.

THIRD_PARTY_NOTICES.md

@@ -0,0 +1,75 @@
+# Third-Party Notices and Image Provenance
+
+CraftSight is a mixed-provenance dataset. Its annotation and metadata layer is licensed separately from its image payload.
+
+## Source mapping
+
+| `data_source` namespace | Image provenance | Upstream terms |
+|---|---|---|
+| `mobs` | **Minecraft Screenshots Dataset with Features**, published by `sqdartemy` on Kaggle | Creative Commons Attribution-NonCommercial 4.0 International (`CC BY-NC 4.0`) |
+| `basalt` | **BASALT Benchmark Evaluation Dataset**, MineRL BASALT team, version v2, Zenodo DOI `10.5281/zenodo.8021960` | MIT License according to the Zenodo record |
+| `custom` | Gameplay recordings captured by the CraftSight dataset maintainer(s) | Not covered by the CraftSight CDLA grant; Minecraft and other applicable rights remain relevant |
+
+## 1. Kaggle-derived images
+
+**Source:** Minecraft Screenshots Dataset with Features  
+**Creator:** `sqdartemy`  
+**Source location:** https://www.kaggle.com/datasets/sqdartemy/minecraft-screenshots-dataset-with-features  
+**Upstream license:** Creative Commons Attribution-NonCommercial 4.0 International  
+**License text:** https://creativecommons.org/licenses/by-nc/4.0/legalcode
+
+The Kaggle-derived images are not relicensed under `CDLA-Sharing-1.0`.
+
+The `NonCommercial` condition applies to use of those images. In particular, the CDLA's permission for commercial Use of the independently covered annotation layer does not remove the upstream noncommercial restriction from the image layer.
+
+When sharing or using these images, preserve source attribution, identify adaptations when required, provide the CC BY-NC 4.0 license reference, and do not imply endorsement by the upstream creator.
+
+This notice assumes CraftSight used the source screenshots but did not copy the upstream feature CSV into the CDLA-covered metadata. If upstream features or annotations were copied or adapted, those fields must also be identified as CC BY-NC 4.0 material rather than CDLA-covered Data.
+
+## 2. BASALT-derived images
+
+**Source:** BASALT Benchmark Evaluation Dataset  
+**Creator:** MineRL BASALT team  
+**Version:** v2, published June 14, 2023  
+**DOI:** https://doi.org/10.5281/zenodo.8021960  
+**Upstream record:** https://zenodo.org/records/8021960  
+**Upstream license shown by the record:** MIT License
+
+The Zenodo record identifies the MineRL BASALT team as creator and marks the record as MIT-licensed. Preserve all copyright and license notices included in the original BASALT distribution.
+
+The MIT License permits use, copying, modification, merging, publication, distribution, sublicensing, and sale, subject to preservation of its copyright and permission notices. The source record and original distribution remain the authoritative source for those notices.
+
+The BASALT-derived image files are not declared to be CDLA-covered by CraftSight. Their inclusion also does not alter rights in Minecraft content depicted in the recordings.
+
+## 3. Maintainer-captured images
+
+Frames under `custom` were captured from gameplay recordings made by the CraftSight dataset maintainer(s).
+
+They are excluded from the CraftSight CDLA grant because they depict third-party Minecraft content. Distribution and use remain subject to applicable law, Minecraft terms, and the Minecraft Usage Guidelines.
+
+## 4. Minecraft rights and disclaimer
+
+Minecraft-related imagery, interfaces, textures, models, names, and trademarks are owned or controlled by Mojang and/or Microsoft or their licensors. No CraftSight license transfers those rights.
+
+**NOT AN OFFICIAL MINECRAFT DATASET. NOT APPROVED BY OR ASSOCIATED WITH MOJANG OR MICROSOFT.**
+
+Minecraft Usage Guidelines:  
+https://www.minecraft.net/en-us/usage-guidelines
+
+Minecraft is a trademark of the Microsoft group of companies.
+
+## 5. Combined use
+
+Recipients using an image together with its CraftSight annotation must comply with every applicable layer:
+
+- CDLA-Sharing-1.0 for the CraftSight annotation and metadata layer;
+- CC BY-NC 4.0 for Kaggle-derived images;
+- MIT and preserved upstream notices for BASALT-derived materials, according to the source record;
+- applicable terms and rights for maintainer-captured images;
+- Minecraft/Mojang/Microsoft rights and applicable terms.
+
+The CDLA's treatment of trained models and other Results applies only to the CDLA-covered layer. It does not determine whether commercial training or distribution is permitted under an upstream image license, especially CC BY-NC 4.0.
+
+## 6. Corrections
+
+Provenance and licensing corrections should be reported through the dataset repository's Discussions tab or issue tracker.