Add files using upload-large-folder tool

LICENSE

@@ -0,0 +1,2 @@
+Creative Commons Attribution-NonCommercial 4.0 International License
+https://creativecommons.org/licenses/by-nc/4.0/

README.md

@@ -0,0 +1,157 @@
+---
+version: 1.0.0
+license: cc-by-nc-4.0
+task_categories:
+  - image-classification
+  - video-classification
+language:
+  - en
+tags:
+  - camera-detection
+  - shot-scale
+  - camera-angle
+  - cinematography
+  - broadcast
+  - sports
+  - soccer
+  - computer-vision
+  - infactory
+annotations_creators:
+  - machine-generated
+  - expert-generated
+pretty_name: Soccer Camera Detection Dataset
+size_categories:
+  - 1K<n<10K
+---
+
+# Soccer Camera Detection Dataset
+
+A curated dataset for classifying camera characteristics in broadcast soccer video, annotated with **9-class shot scale** (CineScale-aligned) and **5-class camera angle** (CineScale2-aligned).
+
+## Dataset Description
+
+This public sample consists of **650 frames** extracted from **10 video clips** of professional soccer broadcasts spanning 2013–2025. Each frame is annotated across two orthogonal dimensions:
+
+- **Shot Scale** — how much of the subject is visible (9 classes from Extreme Close-Up to Extreme Long Shot)
+- **Camera Angle** — the vertical orientation of the camera relative to the subject (5 classes from Overhead to Low Angle)
+
+This is a representative subset of a larger internal dataset, selected to cover diverse match conditions across different eras (2013-2025), teams, and broadcast styles.
+
+### Shot Scale Distribution
+
+| Class | Frames | % |
+|-------|--------|---|
+| MLS | 167 | 25.7% |
+| ELS | 159 | 24.5% |
+| LS | 118 | 18.2% |
+| MS | 85 | 13.1% |
+| MCU | 50 | 7.7% |
+| FS | 43 | 6.6% |
+| IS | 14 | 2.2% |
+| CU | 8 | 1.2% |
+| ECU | 6 | 0.9% |
+
+### Camera Angle Distribution
+
+| Class | Frames | % |
+|-------|--------|---|
+| NEU | 369 | 56.8% |
+| HIGH | 252 | 38.8% |
+| LOW | 29 | 4.5% |
+
+### Source Data
+
+- **Domain**: Professional Soccer Broadcasts (Serie A via Infront Italy)
+- **Resolution**: 1280×720 (720p)
+- **Annotation Style**: Per-frame JSON
+- **Labeling Method**: Proprietary model + human verification (100% verified)
+
+
+## Dataset Structure
+
+```
+infactory-ai/camera-detection/
+├── README.md
+├── metadata.csv
+├── dataset_info.json
+├── taxonomy/
+│   ├── shot_scale.json
+│   ├── camera_angle.json
+│   └── mappings.json
+├── schema/
+│   └── annotation_schema.json
+└── data/
+    └── frames/
+        └── {uuid}_{frame_number}.jpg
+```
+
+### Metadata Fields (`metadata.csv`)
+
+| Field | Type | Description |
+|-------|------|-------------|
+| `frame_id` | string | Unique frame identifier (UUID) |
+| `source_video_id` | string | UUID of the source video clip |
+| `frame_number` | int | Frame index in the source clip |
+| `timestamp_sec` | float | Timestamp in seconds |
+| `shot_scale` | string | Shot scale label (ECU, CU, MCU, MS, MLS, LS, ELS, FS, IS) |
+| `shot_scale_code` | int | Numeric code (0–8) |
+| `camera_angle` | string | Camera angle label (OVH, HIGH, NEU, LOW, DUTCH) |
+| `camera_angle_additional` | string | Additional angle labels if multi-label (pipe-separated) |
+| `subject_type` | string | Primary subject type |
+| `camera_position` | string | Physical camera position |
+| `confidence` | float | Model confidence before human verification |
+| `human_verified` | bool | All frames are human-verified |
+
+## Usage
+
+### Loading with Hugging Face Datasets
+
+```python
+from datasets import load_dataset
+
+dataset = load_dataset("infactory-ai/camera-detection")
+```
+
+### Loading with pandas
+
+```python
+import pandas as pd
+
+df = pd.read_csv("metadata.csv")
+print(df["shot_scale"].value_counts())
+print(df["camera_angle"].value_counts())
+```
+
+## Team
+
+| Name | Role |
+|------|------|
+| **Valentino Constantinou** | Head of Infrastructure |
+| **John Kanalakis** | Chief Technology Officer |
+
+## License
+
+This dataset is released under the [Creative Commons Attribution-NonCommercial 4.0 International License (CC BY-NC 4.0)](https://creativecommons.org/licenses/by-nc/4.0/).
+
+**You are free to:**
+- **Share** — copy and redistribute the material in any medium or format
+- **Adapt** — remix, transform, and build upon the material
+
+**Under the following terms:**
+- **Attribution** — You must give appropriate credit to Infactory, provide a link to the license, and indicate if changes were made.
+- **Non-Commercial** — You may not use the material for commercial purposes without a separate commercial license from Infactory.
+
+**Commercial licensing:** For commercial use, contact [hello@infactory.ai](mailto:hello@infactory.ai).
+
+## Citation
+
+```bibtex
+@dataset{camera_detection_2026,
+  title={Soccer Camera Detection Dataset},
+  author={Constantinou, Valentino and Kanalakis, John},
+  year={2026},
+  publisher={Infactory},
+  url={https://huggingface.co/datasets/infactory-ai/camera-detection},
+  note={Per-frame camera characterization: 9-class shot scale (CineScale-aligned), 5-class camera angle (CineScale2-aligned)}
+}
+```

dataset_info.json

@@ -0,0 +1,25 @@
+{
+  "dataset_name": "Soccer Camera Detection",
+  "version": "1.0.0",
+  "description": "Per-frame shot scale and camera angle annotations for broadcast soccer.",
+  "statistics": {
+    "total_frames": 650,
+    "total_clips": 10,
+    "shot_scale_distribution": {
+      "CU": 8,
+      "ECU": 6,
+      "ELS": 159,
+      "FS": 43,
+      "IS": 14,
+      "LS": 118,
+      "MCU": 50,
+      "MLS": 167,
+      "MS": 85
+    },
+    "camera_angle_distribution": {
+      "HIGH": 252,
+      "LOW": 29,
+      "NEU": 369
+    }
+  }
+}

schema/annotation_schema.json

@@ -0,0 +1,286 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "$id": "https://infactory.ai/schemas/camera-detection/annotation/v1",
+  "title": "Camera Detection Frame Annotation",
+  "description": "Schema for per-frame annotations in the Infactory Camera Detection Dataset. Each record describes the camera characteristics of a single video frame across three orthogonal dimensions: shot scale, camera angle, and camera movement.",
+  "type": "object",
+  "required": [
+    "frame_id",
+    "source_video",
+    "frame_number",
+    "timestamp_sec",
+    "shot_scale",
+    "shot_scale_code",
+    "camera_angle",
+    "camera_movement",
+    "annotator",
+    "annotation_version"
+  ],
+  "properties": {
+    "frame_id": {
+      "type": "string",
+      "format": "uuid",
+      "description": "Unique identifier for this annotated frame (UUID v4)."
+    },
+    "source_video": {
+      "type": "string",
+      "description": "Filename of the source video from which this frame was extracted.",
+      "pattern": "^.+\\.(mp4|avi|mov|mkv|webm)$"
+    },
+    "source_video_id": {
+      "type": "string",
+      "description": "Unique identifier for the source video, if available (e.g., UUID or dataset-specific ID)."
+    },
+    "frame_number": {
+      "type": "integer",
+      "minimum": 0,
+      "description": "Zero-indexed frame number within the source video."
+    },
+    "timestamp_sec": {
+      "type": "number",
+      "minimum": 0,
+      "description": "Timestamp of the frame in seconds from the start of the source video."
+    },
+
+    "shot_scale": {
+      "type": "string",
+      "enum": ["ECU", "CU", "MCU", "MS", "MLS", "LS", "ELS", "FS", "IS"],
+      "description": "Primary shot scale classification. See DEFINITIONS.md Section 1 for full definitions."
+    },
+    "shot_scale_code": {
+      "type": "integer",
+      "enum": [0, 1, 2, 3, 4, 5, 6, 7, 8],
+      "description": "Numeric code for shot scale. Codes 1-7 are ordinal (closest to farthest). 0=FS, 8=IS are non-ordinal."
+    },
+    "shot_scale_underlying": {
+      "type": ["string", "null"],
+      "enum": ["ECU", "CU", "MCU", "MS", "MLS", "LS", "ELS", null],
+      "description": "When shot_scale is 'FS' (Foreground Shot), this field captures the underlying shot scale of the primary subject behind the foreground element. Null for all other shot_scale values."
+    },
+
+    "camera_angle": {
+      "type": "string",
+      "enum": ["OVH", "HIGH", "NEU", "LOW", "DUTCH"],
+      "description": "Camera angle classification. See DEFINITIONS.md Section 2 for full definitions."
+    },
+    "camera_angle_vertical_tendency": {
+      "type": ["string", "null"],
+      "enum": ["OVH", "HIGH", "NEU", "LOW", null],
+      "description": "When camera_angle is 'DUTCH', this optional field captures the underlying vertical tendency. Null when camera_angle is not DUTCH."
+    },
+
+    "camera_movement": {
+      "type": "string",
+      "enum": ["STATIC", "PAN", "TILT", "TRACK", "ZOOM_IN", "ZOOM_OUT", "HANDHELD", "CRANE", "AERIAL"],
+      "description": "Camera movement classification based on temporal context around this frame. See DEFINITIONS.md Section 3 for full definitions."
+    },
+
+    "subject_type": {
+      "type": "string",
+      "enum": ["person", "group", "field_of_play", "object", "graphic", "crowd", "venue", "mixed"],
+      "description": "What the primary subject of the frame is. See DEFINITIONS.md Section 4.2."
+    },
+    "subject_count": {
+      "type": ["integer", "null"],
+      "minimum": 0,
+      "description": "Number of distinct subjects visible when subject_type is 'person' or 'group'. Null when not applicable."
+    },
+
+    "camera_position": {
+      "type": ["string", "null"],
+      "enum": [
+        "main_center", "main_left", "main_right",
+        "behind_goal", "goal_line_tech", "spider_cam",
+        "sideline", "corner", "inside_goal",
+        "aerial", "tunnel", "studio", "other", null
+      ],
+      "description": "Physical camera position in the venue. Primarily for sports broadcast content. Null when unknown or not applicable. See DEFINITIONS.md Section 4.1."
+    },
+
+    "transition_context": {
+      "type": "string",
+      "enum": ["none", "hard_cut", "dissolve", "wipe", "fade_black", "logo_transition"],
+      "default": "none",
+      "description": "How this frame relates to surrounding editorial transitions. See DEFINITIONS.md Section 4.3."
+    },
+
+    "safe_area": {
+      "type": ["object", "null"],
+      "description": "SMPTE ST 2046-1 safe area compliance. Null when not evaluated.",
+      "properties": {
+        "action_safe": {
+          "type": "boolean",
+          "description": "Subject's key features fall within the Action-Safe zone (inner 93% of frame per SMPTE ST 2046-1)."
+        },
+        "title_safe": {
+          "type": "boolean",
+          "description": "Subject's key features fall within the Title-Safe zone (inner 90% of frame per SMPTE ST 2046-1)."
+        }
+      },
+      "required": ["action_safe", "title_safe"]
+    },
+
+    "technical": {
+      "type": "object",
+      "description": "Technical metadata about the source frame.",
+      "properties": {
+        "resolution_width": {
+          "type": "integer",
+          "minimum": 1,
+          "description": "Frame width in pixels."
+        },
+        "resolution_height": {
+          "type": "integer",
+          "minimum": 1,
+          "description": "Frame height in pixels."
+        },
+        "frame_rate": {
+          "type": "number",
+          "minimum": 0,
+          "description": "Source video frame rate in frames per second."
+        },
+        "aspect_ratio": {
+          "type": "string",
+          "description": "Display aspect ratio (e.g., '16:9', '4:3', '21:9')."
+        },
+        "codec": {
+          "type": ["string", "null"],
+          "description": "Video codec (e.g., 'H.264', 'H.265', 'VP9')."
+        },
+        "color_space": {
+          "type": ["string", "null"],
+          "description": "Color sampling format (e.g., '4:2:0', '4:2:2', '4:4:4')."
+        },
+        "bit_depth": {
+          "type": ["integer", "null"],
+          "description": "Bit depth per channel (e.g., 8, 10, 12)."
+        }
+      },
+      "required": ["resolution_width", "resolution_height", "frame_rate"]
+    },
+
+    "annotator": {
+      "type": "string",
+      "description": "Identifier of the annotator or model that produced this annotation (e.g., 'human_annotator_01', 'model_v1.0', 'gemini_2.0_flash')."
+    },
+    "annotation_version": {
+      "type": "string",
+      "description": "Version of the annotation schema used (e.g., '1.0.0').",
+      "pattern": "^\\d+\\.\\d+\\.\\d+$"
+    },
+    "confidence": {
+      "type": ["number", "null"],
+      "minimum": 0.0,
+      "maximum": 1.0,
+      "description": "Model confidence score for the annotation. Null for human annotations. Range [0.0, 1.0]."
+    },
+    "human_verified": {
+      "type": "boolean",
+      "default": false,
+      "description": "Whether this annotation has been verified by a human reviewer."
+    },
+    "verification_annotator": {
+      "type": ["string", "null"],
+      "description": "Identifier of the human who verified this annotation. Null if not yet verified."
+    },
+
+    "notes": {
+      "type": ["string", "null"],
+      "description": "Free-text annotator notes for edge cases or ambiguous frames."
+    },
+
+    "domain": {
+      "type": "string",
+      "enum": ["sports_broadcast", "news_broadcast", "cinema", "documentary", "user_generated", "other"],
+      "description": "Content domain of the source material."
+    },
+    "sport_type": {
+      "type": ["string", "null"],
+      "description": "Specific sport, when domain is 'sports_broadcast' (e.g., 'soccer', 'basketball', 'tennis'). Null otherwise."
+    }
+  },
+  "additionalProperties": false,
+
+  "if": {
+    "properties": { "shot_scale": { "const": "FS" } }
+  },
+  "then": {
+    "required": ["shot_scale_underlying"]
+  },
+
+  "examples": [
+    {
+      "frame_id": "a1b2c3d4-e5f6-7890-abcd-ef1234567890",
+      "source_video": "GA_ITS_MSA_20241201_JUV_NAP_SIS2.mp4",
+      "source_video_id": "b0558318-3605-4d54-9930-41c77fd4b20a",
+      "frame_number": 1234,
+      "timestamp_sec": 41.13,
+      "shot_scale": "MCU",
+      "shot_scale_code": 3,
+      "shot_scale_underlying": null,
+      "camera_angle": "NEU",
+      "camera_angle_vertical_tendency": null,
+      "camera_movement": "STATIC",
+      "subject_type": "person",
+      "subject_count": 1,
+      "camera_position": "sideline",
+      "transition_context": "none",
+      "safe_area": {
+        "action_safe": true,
+        "title_safe": true
+      },
+      "technical": {
+        "resolution_width": 1920,
+        "resolution_height": 1080,
+        "frame_rate": 25.0,
+        "aspect_ratio": "16:9",
+        "codec": "H.264",
+        "color_space": "4:2:0",
+        "bit_depth": 8
+      },
+      "annotator": "model_v1.0",
+      "annotation_version": "1.0.0",
+      "confidence": 0.94,
+      "human_verified": false,
+      "verification_annotator": null,
+      "notes": null,
+      "domain": "sports_broadcast",
+      "sport_type": "soccer"
+    },
+    {
+      "frame_id": "f9e8d7c6-b5a4-3210-fedc-ba9876543210",
+      "source_video": "GA_ITS_MSA_20241201_JUV_NAP_SIS2.mp4",
+      "source_video_id": "b0558318-3605-4d54-9930-41c77fd4b20a",
+      "frame_number": 5678,
+      "timestamp_sec": 189.27,
+      "shot_scale": "ELS",
+      "shot_scale_code": 7,
+      "shot_scale_underlying": null,
+      "camera_angle": "HIGH",
+      "camera_angle_vertical_tendency": null,
+      "camera_movement": "PAN",
+      "subject_type": "field_of_play",
+      "subject_count": null,
+      "camera_position": "main_center",
+      "transition_context": "hard_cut",
+      "safe_area": null,
+      "technical": {
+        "resolution_width": 1920,
+        "resolution_height": 1080,
+        "frame_rate": 25.0,
+        "aspect_ratio": "16:9",
+        "codec": "H.264",
+        "color_space": "4:2:0",
+        "bit_depth": 8
+      },
+      "annotator": "model_v1.0",
+      "annotation_version": "1.0.0",
+      "confidence": 0.87,
+      "human_verified": true,
+      "verification_annotator": "human_annotator_01",
+      "notes": "Main broadcast camera following open play left to right.",
+      "domain": "sports_broadcast",
+      "sport_type": "soccer"
+    }
+  ]
+}

source_mapping.json

@@ -0,0 +1,12 @@
+{
+  "37a98e4f-7018-49f2-b321-2ced60453e20": "0c2857fc-f97e-4301-80d7-63507a1303a7",
+  "0c984110-de7f-4e71-97c5-1f9aaf062e7e": "d485c6ac-dd01-455b-b6de-744550c4c216",
+  "20d8344f-84a8-4564-b575-3ed913cf7346": "c17e1a0a-dc98-435e-a195-e5176f14ae31",
+  "4e0be653-33bd-4bb1-b29a-372832586c88": "21f4f832-1d99-4647-8973-1eaec52f8353",
+  "8c169226-aa86-47b8-a900-c75944fe3ef6": "cc89d977-fd4f-4d0a-9a1e-d010a8f5b579",
+  "4d24650f-c7d0-47b0-a8bc-de20d2d0c146": "bdfb7106-c2ba-4c16-a2e9-88ceff6ed3da",
+  "9168f102-40b9-41a2-8e9c-982de6099ee6": "a1884fe4-fb6e-439f-8b04-381f12ecd884",
+  "604cb77f-7183-4950-88b3-52fe74b9108a": "44eb7456-67b1-44d6-bed2-c9ea8626b4f8",
+  "4fad4868-934e-473b-a74a-ad0191a6777f": "de7a01c2-c383-4ae3-a9c1-5d385ed33f60",
+  "6708badd-c675-475d-a05a-ca69925e5073": "623686ca-50c6-4c6b-8801-8e484ff033d3"
+}

taxonomy/camera_angle.json

@@ -0,0 +1,58 @@
+{
+  "$schema": "https://infactory.ai/schemas/taxonomy/v1",
+  "taxonomy_name": "camera_angle",
+  "taxonomy_version": "1.0.0",
+  "description": "Camera angle classification based on the vertical orientation of the camera's optical axis relative to the subject. Determines whether the camera looks down, across, or up at the subject.",
+  "primary_source": "CineScale2 (Savardi et al., 2023)",
+  "num_classes": 5,
+  "classes": [
+    {
+      "code": "OVH",
+      "label": "overhead",
+      "display_name": "Overhead",
+      "definition": "Camera looks nearly straight down on the subject, approaching perpendicular to the ground plane. Angle relative to ground exceeds approximately 70 degrees.",
+      "psychological_effect": "Omniscience, detachment, tactical overview, vulnerability of subject.",
+      "broadcast_frequency": "rare",
+      "typical_platforms": ["spider_cam", "drone", "stadium_roof_cam"]
+    },
+    {
+      "code": "HIGH",
+      "label": "high_angle",
+      "display_name": "High Angle",
+      "definition": "Camera is above subject's eye level and angled downward. Subject is recognizable from a front or side perspective (not purely top-down). Horizon line typically in the upper half of frame.",
+      "psychological_effect": "Diminishment, vulnerability, overview, context.",
+      "broadcast_frequency": "very_common",
+      "typical_platforms": ["main_broadcast_camera", "gantry_camera", "upper_tier"]
+    },
+    {
+      "code": "NEU",
+      "label": "neutral",
+      "display_name": "Neutral / Eye Level",
+      "definition": "Camera at approximately the same height as the subject's eyes or center of mass. Optical axis roughly horizontal (within ±10 degrees). The psychologically 'transparent' default angle.",
+      "psychological_effect": "Equality, naturalism, objectivity, conversational rapport.",
+      "broadcast_frequency": "very_common",
+      "typical_platforms": ["pitch_side_camera", "interview_camera", "tunnel_camera"]
+    },
+    {
+      "code": "LOW",
+      "label": "low_angle",
+      "display_name": "Low Angle",
+      "definition": "Camera is below subject's eye level and angled upward. Subject framed against sky, ceiling, or elevated background. Horizon line typically in the lower third of frame.",
+      "psychological_effect": "Power, dominance, heroism, dramatic emphasis.",
+      "broadcast_frequency": "common",
+      "typical_platforms": ["pitch_level_camera", "ground_camera", "behind_goal_low"]
+    },
+    {
+      "code": "DUTCH",
+      "label": "dutch_angle",
+      "display_name": "Dutch Angle",
+      "definition": "Camera rotated along its longitudinal (roll) axis, creating a visibly tilted horizon. A deliberate compositional choice for disorientation or dynamism, not the result of an unsteady camera.",
+      "psychological_effect": "Unease, disorientation, tension, stylistic emphasis.",
+      "broadcast_frequency": "very_rare",
+      "typical_platforms": ["specialty_camera", "replay_camera"]
+    }
+  ],
+  "default_class": "NEU",
+  "ordering": "OVH → HIGH → NEU → LOW form a vertical axis. DUTCH is orthogonal (roll axis) and can co-occur with any vertical angle.",
+  "notes": "Minor deviations from horizontal (±10 degrees) should be classified as NEU. Only label HIGH or LOW when the departure from horizontal is noticeable and intentional."
+}

taxonomy/mappings.json

@@ -0,0 +1,119 @@
+{
+  "$schema": "https://infactory.ai/schemas/taxonomy/v1",
+  "mapping_version": "1.0.0",
+  "description": "Cross-dataset label mappings between the Infactory Camera Detection taxonomy and major published datasets. Enables downstream researchers to re-bin labels for direct comparison with existing benchmarks.",
+  "shot_scale_mappings": {
+    "to_movieshots_5class": {
+      "target_dataset": "MovieShots (Rao et al., ECCV 2020)",
+      "target_classes": 5,
+      "citation": "rao2020unified",
+      "mapping": {
+        "ECU": {"target_label": "ECS", "target_code": 5, "target_display": "Extreme Close-up Shot"},
+        "CU":  {"target_label": "CS",  "target_code": 4, "target_display": "Close-up Shot"},
+        "MCU": {"target_label": "CS",  "target_code": 4, "target_display": "Close-up Shot", "note": "MovieShots collapses MCU into CS. Information loss: MCU (the BBC interview standard) is not distinguishable from CU in the 5-class system."},
+        "MS":  {"target_label": "MS",  "target_code": 3, "target_display": "Medium Shot"},
+        "MLS": {"target_label": "MS",  "target_code": 3, "target_display": "Medium Shot", "note": "MovieShots collapses MLS into MS. Information loss: the 'cowboy shot' is not distinguishable from waist-up framing."},
+        "LS":  {"target_label": "FS",  "target_code": 2, "target_display": "Full Shot"},
+        "ELS": {"target_label": "LS",  "target_code": 1, "target_display": "Long Shot", "note": "Terminology shift: MovieShots 'LS' corresponds to our 'ELS'. MovieShots 'FS' corresponds to our 'LS'."},
+        "FS":  {"target_label": null,  "target_code": null, "target_display": null, "note": "Foreground Shot has no equivalent in MovieShots. Map by underlying_scale if available."},
+        "IS":  {"target_label": null,  "target_code": null, "target_display": null, "note": "Insert Shot has no equivalent in MovieShots. Exclude from MovieShots evaluation."}
+      },
+      "reverse_mapping_notes": "MovieShots → Infactory is ambiguous for CS (could be CU or MCU) and MS (could be MS or MLS). When reverse-mapping, default to the finer-grained class closer to the median of the range: CS → CU, MS → MS."
+    },
+    "to_cinescale_9class": {
+      "target_dataset": "CineScale (Savardi et al., 2018; Benini et al., 2021)",
+      "target_classes": 9,
+      "citation": "SSM18",
+      "mapping": {
+        "ECU": {"target_label": "ECU", "target_code": 1, "target_display": "Extreme Close Up"},
+        "CU":  {"target_label": "CU",  "target_code": 2, "target_display": "Close Up"},
+        "MCU": {"target_label": "MCU", "target_code": 3, "target_display": "Medium Close Up"},
+        "MS":  {"target_label": "MS",  "target_code": 4, "target_display": "Medium Shot"},
+        "MLS": {"target_label": "MLS", "target_code": 5, "target_display": "Medium Long Shot"},
+        "LS":  {"target_label": "LS",  "target_code": 6, "target_display": "Long Shot"},
+        "ELS": {"target_label": "ELS", "target_code": 7, "target_display": "Extreme Long Shot"},
+        "FS":  {"target_label": "FS",  "target_code": 0, "target_display": "Foreground Shot"},
+        "IS":  {"target_label": "IS",  "target_code": 8, "target_display": "Insert Shots"}
+      },
+      "reverse_mapping_notes": "1:1 mapping. CineScale and Infactory taxonomies are directly compatible."
+    },
+    "to_bbc_basic_3class": {
+      "target_dataset": "BBC Basic Shot Types (BBC Bitesize / BBC Academy)",
+      "target_classes": 3,
+      "citation": "bbc_bitesize",
+      "mapping": {
+        "ECU": {"target_label": "close_up",    "target_display": "Close-Up"},
+        "CU":  {"target_label": "close_up",    "target_display": "Close-Up"},
+        "MCU": {"target_label": "close_up",    "target_display": "Close-Up"},
+        "MS":  {"target_label": "medium_shot",  "target_display": "Medium Shot"},
+        "MLS": {"target_label": "medium_shot",  "target_display": "Medium Shot"},
+        "LS":  {"target_label": "long_shot",    "target_display": "Long Shot / Wide Shot"},
+        "ELS": {"target_label": "long_shot",    "target_display": "Long Shot / Wide Shot"},
+        "FS":  {"target_label": null,           "target_display": null, "note": "Map by underlying_scale."},
+        "IS":  {"target_label": null,           "target_display": null, "note": "Not applicable in BBC basic taxonomy."}
+      },
+      "reverse_mapping_notes": "BBC basic → Infactory is highly ambiguous. close_up could be ECU, CU, or MCU. medium_shot could be MS or MLS. long_shot could be LS or ELS."
+    },
+    "to_soccernet_13class": {
+      "target_dataset": "SoccerNet-v2 Camera Shot Segmentation (Deliège et al., CVPR 2021)",
+      "target_classes": 13,
+      "citation": "soccernet_v2",
+      "mapping_type": "partial",
+      "notes": "SoccerNet classifies by CAMERA POSITION (where the camera is physically located), not by SHOT SCALE (how much of the subject is visible). These are orthogonal dimensions. Direct label mapping is not meaningful. Instead, SoccerNet camera positions can be used as the 'camera_position' metadata field in our schema.",
+      "position_to_our_metadata": {
+        "Main camera center":                 "main_center",
+        "Close-up player or field referee":   "sideline",
+        "Main camera left":                   "main_left",
+        "Main camera right":                  "main_right",
+        "Goal line technology camera":         "goal_line_tech",
+        "Main behind the goal":               "behind_goal",
+        "Spider camera":                       "spider_cam",
+        "Close-up side staff":                "sideline",
+        "Close-up corner":                     "corner",
+        "Close-up behind the goal":           "behind_goal",
+        "Inside the goal":                     "inside_goal",
+        "Public":                              "other",
+        "other":                               "other"
+      },
+      "typical_scale_correlations": {
+        "Main camera center": ["ELS", "LS"],
+        "Close-up player or field referee": ["CU", "MCU", "MS"],
+        "Spider camera": ["ELS", "LS"],
+        "Goal line technology camera": ["MS", "LS"],
+        "Inside the goal": ["LS", "ELS"]
+      }
+    }
+  },
+  "camera_movement_mappings": {
+    "to_movieshots_4class": {
+      "target_dataset": "MovieShots (Rao et al., ECCV 2020)",
+      "target_classes": 4,
+      "mapping": {
+        "STATIC":   {"target_label": "Static",  "target_display": "Static Shot"},
+        "PAN":      {"target_label": "Motion",  "target_display": "Motion Shot"},
+        "TILT":     {"target_label": "Motion",  "target_display": "Motion Shot"},
+        "TRACK":    {"target_label": "Motion",  "target_display": "Motion Shot"},
+        "ZOOM_IN":  {"target_label": "Push",    "target_display": "Push Shot"},
+        "ZOOM_OUT": {"target_label": "Pull",    "target_display": "Pull Shot"},
+        "HANDHELD": {"target_label": "Motion",  "target_display": "Motion Shot"},
+        "CRANE":    {"target_label": "Motion",  "target_display": "Motion Shot"},
+        "AERIAL":   {"target_label": "Motion",  "target_display": "Motion Shot"}
+      },
+      "reverse_mapping_notes": "MovieShots 'Motion' is highly ambiguous in reverse: could be PAN, TILT, TRACK, HANDHELD, CRANE, or AERIAL."
+    }
+  },
+  "camera_angle_mappings": {
+    "to_cinescale2_5class": {
+      "target_dataset": "CineScale2 (Savardi et al., 2023)",
+      "target_classes": 5,
+      "mapping": {
+        "OVH":   {"target_label": "Overhead",  "target_display": "Overhead"},
+        "HIGH":  {"target_label": "High",      "target_display": "High"},
+        "NEU":   {"target_label": "Neutral",   "target_display": "Neutral"},
+        "LOW":   {"target_label": "Low",       "target_display": "Low"},
+        "DUTCH": {"target_label": "Dutch",     "target_display": "Dutch"}
+      },
+      "reverse_mapping_notes": "1:1 mapping. CineScale2 and Infactory angle taxonomies are directly compatible."
+    }
+  }
+}

taxonomy/shot_scale.json

@@ -0,0 +1,123 @@
+{
+  "$schema": "https://infactory.ai/schemas/taxonomy/v1",
+  "taxonomy_name": "shot_scale",
+  "taxonomy_version": "1.0.0",
+  "description": "Shot scale classification based on apparent distance between camera and primary subject, measured by visible anatomical landmarks when a human subject is present.",
+  "primary_source": "CineScale (Savardi et al., 2018; Benini et al., 2021)",
+  "num_classes": 9,
+  "num_primary_classes": 7,
+  "num_special_classes": 2,
+  "classes": [
+    {
+      "code": 0,
+      "abbreviation": "FS",
+      "label": "foreground_shot",
+      "display_name": "Foreground Shot",
+      "category": "special",
+      "definition": "Primary subject is partially occluded by a deliberate foreground element (e.g., over-the-shoulder framing, through-object composition).",
+      "body_anchor": null,
+      "body_anchor_description": "Not applicable — determined by the foreground compositional element, not body position.",
+      "requires_underlying_scale": true,
+      "has_human_subject": "optional"
+    },
+    {
+      "code": 1,
+      "abbreviation": "ECU",
+      "label": "extreme_close_up",
+      "display_name": "Extreme Close-Up",
+      "category": "primary",
+      "definition": "A single anatomical feature or small object fills the frame: an eye, mouth, hand, or comparable isolated region. Less than the full face is visible.",
+      "body_anchor": "isolated_feature",
+      "body_anchor_description": "Single feature only: eye, ear, mouth, hand. The full face (hairline to chin) is NOT visible.",
+      "requires_underlying_scale": false,
+      "has_human_subject": "required_for_body_anchor"
+    },
+    {
+      "code": 2,
+      "abbreviation": "CU",
+      "label": "close_up",
+      "display_name": "Close-Up",
+      "category": "primary",
+      "definition": "Head and face fill the frame. Full facial structure visible from hairline to chin. Shoulders barely visible or not visible.",
+      "body_anchor": "neck",
+      "body_anchor_description": "Frame spans from top of head to base of neck. Shoulders may be marginally visible but do not constitute a meaningful portion of the frame.",
+      "requires_underlying_scale": false,
+      "has_human_subject": "required_for_body_anchor"
+    },
+    {
+      "code": 3,
+      "abbreviation": "MCU",
+      "label": "medium_close_up",
+      "display_name": "Medium Close-Up",
+      "category": "primary",
+      "definition": "Head to mid-chest. The standard broadcast interview frame. Both shoulders clearly visible. Frame cuts above the nipple line.",
+      "body_anchor": "mid_chest",
+      "body_anchor_description": "Lower frame boundary at mid-chest, above the nipple line and below the collarbone. Both shoulders clearly visible. The 'BBC talking head' frame.",
+      "requires_underlying_scale": false,
+      "has_human_subject": "required_for_body_anchor"
+    },
+    {
+      "code": 4,
+      "abbreviation": "MS",
+      "label": "medium_shot",
+      "display_name": "Medium Shot",
+      "category": "primary",
+      "definition": "Head to waist. Conversational distance. Facial expression, body language, and hand gestures all visible. Frame cuts at the belt line.",
+      "body_anchor": "waist",
+      "body_anchor_description": "Lower frame boundary at or near the waist (belt line). Arms visible to at least the elbow. Hands typically visible if at sides or gesturing.",
+      "requires_underlying_scale": false,
+      "has_human_subject": "required_for_body_anchor"
+    },
+    {
+      "code": 5,
+      "abbreviation": "MLS",
+      "label": "medium_long_shot",
+      "display_name": "Medium Long Shot",
+      "category": "primary",
+      "definition": "Head to knees. Also called 'cowboy shot' or 'American shot'. Full torso and thighs visible, feet cropped.",
+      "body_anchor": "knees",
+      "body_anchor_description": "Lower frame boundary at or near the knees. Full torso and thighs visible. Feet and lower legs are cropped.",
+      "requires_underlying_scale": false,
+      "has_human_subject": "required_for_body_anchor"
+    },
+    {
+      "code": 6,
+      "abbreviation": "LS",
+      "label": "long_shot",
+      "display_name": "Long Shot",
+      "category": "primary",
+      "definition": "Full body visible from head to feet. Subject is the clear focal point but setting provides significant context. Subject occupies roughly 1/3 to 2/3 of frame height.",
+      "body_anchor": "feet",
+      "body_anchor_description": "Entire body visible: head, torso, legs, and feet. Subject typically occupies between 1/3 and 2/3 of the frame height.",
+      "requires_underlying_scale": false,
+      "has_human_subject": "required_for_body_anchor"
+    },
+    {
+      "code": 7,
+      "abbreviation": "ELS",
+      "label": "extreme_long_shot",
+      "display_name": "Extreme Long Shot",
+      "category": "primary",
+      "definition": "Environment dominates the frame. Subjects, if present, are small figures (typically less than 1/4 of frame height). The classic 'establishing shot'.",
+      "body_anchor": null,
+      "body_anchor_description": "Human subjects, if present, occupy a small fraction of frame height (typically <1/4). Individual facial features not discernible at native resolution.",
+      "requires_underlying_scale": false,
+      "has_human_subject": "optional"
+    },
+    {
+      "code": 8,
+      "abbreviation": "IS",
+      "label": "insert_shot",
+      "display_name": "Insert Shot",
+      "category": "special",
+      "definition": "A non-human subject fills the frame: scoreboard, ball, graphic, tactical diagram, clock, trophy, replay indicator, or other object/information display.",
+      "body_anchor": null,
+      "body_anchor_description": "Not applicable — subject is non-human. If the subject is a human body part (hand, eye, foot), use ECU instead.",
+      "requires_underlying_scale": false,
+      "has_human_subject": "excluded"
+    }
+  ],
+  "ordering": "Classes 1-7 form an ordinal scale from closest to farthest. Classes 0 (FS) and 8 (IS) are non-ordinal special classes.",
+  "default_class": null,
+  "unknown_class": null
+}