2026-10-06T15:02:40Z https://api.figshare.com/v2/oai

oai:figshare.com:article/33944731 2026-09-22T12:45:22Z category_28831 category_28843 category_28885 item_type_3 month_year_09_2026

A multi-pitch multi-speed dataset for embodied spatial cognition evaluation in a virtual urban environment Chen Zugang (17475180) Autonomous agents and multiagent systems Knowledge representation and reasoning Image processing Visual Place Recognition Spatial Cognition Embodied AI Minecraft Project Malmo Synthetic Dataset Trajectory Data Navigation CityVPR is a synthetic dataset collected in a custom-built Minecraft virtual city for evaluating and training spatial cognitive abilities in embodied agents. It provides first-person RGB images from 11 distinct navigation routes, captured at 4 different pitch angles and 3 different speed modes, with precise pose information for each frame.## Dataset Statistics- Total frames: 57,349- Navigation tasks: 132 (11 routes x 4 pitch angles x 3 speed modes)- Total size: ~12.25 GB- Image resolution: 1280 x 720 (PNG)- Frame rate: 10 fps- City size: 100 x 100 - Buildings: 38- Waypoints: 32 ## Speed Modes- sm1 (speedMode_1): Constant speed of 4.32 between waypoints- sm2 (speedMode_2): Constant speed of 2.59 between waypoints- sm3 (speedMode_3): Accelerating from 2.59 to 5.18 between waypoints ## File StructureThis dataset is split into 34 zip archives. Download all files and extract them into the same directory to reconstruct the original folder structure. File naming:- city_vpr_sm1_route1.zip ~ city_vpr_sm1_route11.zip (speedMode_1)- city_vpr_sm2_route1.zip ~ city_vpr_sm2_route11.zip (speedMode_2)- city_vpr_sm3_route1.zip ~ city_vpr_sm3_route11.zip (speedMode_3)- city_vpr_metadata.zip (auxiliary files) ## Directory Structure ### city_vpr_metadata.zipcity_vpr_metadata/├── route_maps/ # 11 route map SVGs (top-down view)├── buildings.json # Building positions and dimensions├── city_map.svg # City road map (top-down view)├── data_quality_report.json # Data quality check report├── dataset_metadata.json # Structured metadata├── README.md # Dataset documentation├── routes.json # Waypoint sequence for each of the 11 routes└── waypoints.json # 32 waypoint coordinates ### city_vpr_smi_routej.zip (e.g., city_vpr_sm1_route1.zip = speedMode_1, route 1)city_vpr_smi_routej/├── pitch_0/ # Forward horizontal (0 degrees)│ ├── r1_pitch0_1.png # Image frames│ ├── r1_pitch0_2.png│ ├── ...│ ├── record.txt # Raw record: timestamp, XYZ, facing, pitch, speed│ └── trajectory.csv # Structured: timestamp, XYZ, facing, pitch, speed├── pitch_30/ # 30 degrees downward│ ├── (same structure)├── pitch_60/ # 60 degrees downward│ ├── (same structure)└── pitch_90/ # Straight down (90 degrees) └── (same structure) where i = speed mode (1, 2, or 3), j = route number (1 to 11) ## Data Collection PlatformCollected using Project Malmo (Microsoft) in a Minecraft 1.11.2 virtual city environment. ## Trajectory Data Format ### record.txtEach line records one frame:frame_name: X: Y: Z: Facing: Pitch: Speed: Time: ### trajectory.csv| Column | Description ||-------------|--------------------------------------------------|| frame_name | Image filename || time | Timestamp relative to task start || x, y, z | Agent position in world coordinates || facing | Yaw angle || pitch | Pitch angle || speed | Walking speed |Source code for data collection and analysis: <a href="https://github.com/sdakjds/cityvpr" target="_blank">sdakjds/cityvpr: CityVPR Dataset Tools - A Multi-Pitch Multi-Speed VPR Dataset in Minecraft</a> 2026-09-22T12:45:22Z Dataset Dataset 10.6084/m9.figshare.33944731.v1 https://figshare.com/articles/dataset/A_multi-pitch_multi-speed_dataset_for_embodied_spatial_cognition_evaluation_in_a_virtual_urban_environment/33944731 CC BY 4.0