The full dataset viewer is not available (click to read why). Only showing a preview of the rows.
Error code: DatasetGenerationError
Exception: ArrowNotImplementedError
Message: Cannot write struct type 'prediction' with no child field to Parquet. Consider adding a dummy child field.
Traceback: Traceback (most recent call last):
File "/usr/local/lib/python3.12/site-packages/datasets/builder.py", line 1831, in _prepare_split_single
writer.write_table(table)
File "/usr/local/lib/python3.12/site-packages/datasets/arrow_writer.py", line 712, in write_table
self._build_writer(inferred_schema=pa_table.schema)
File "/usr/local/lib/python3.12/site-packages/datasets/arrow_writer.py", line 757, in _build_writer
self.pa_writer = pq.ParquetWriter(
^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/site-packages/pyarrow/parquet/core.py", line 1070, in __init__
self.writer = _parquet.ParquetWriter(
^^^^^^^^^^^^^^^^^^^^^^^
File "pyarrow/_parquet.pyx", line 2363, in pyarrow._parquet.ParquetWriter.__cinit__
File "pyarrow/error.pxi", line 155, in pyarrow.lib.pyarrow_internal_check_status
File "pyarrow/error.pxi", line 92, in pyarrow.lib.check_status
pyarrow.lib.ArrowNotImplementedError: Cannot write struct type 'prediction' with no child field to Parquet. Consider adding a dummy child field.
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/usr/local/lib/python3.12/site-packages/datasets/builder.py", line 1847, in _prepare_split_single
num_examples, num_bytes = writer.finalize()
^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/site-packages/datasets/arrow_writer.py", line 731, in finalize
self._build_writer(self.schema)
File "/usr/local/lib/python3.12/site-packages/datasets/arrow_writer.py", line 757, in _build_writer
self.pa_writer = pq.ParquetWriter(
^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/site-packages/pyarrow/parquet/core.py", line 1070, in __init__
self.writer = _parquet.ParquetWriter(
^^^^^^^^^^^^^^^^^^^^^^^
File "pyarrow/_parquet.pyx", line 2363, in pyarrow._parquet.ParquetWriter.__cinit__
File "pyarrow/error.pxi", line 155, in pyarrow.lib.pyarrow_internal_check_status
File "pyarrow/error.pxi", line 92, in pyarrow.lib.check_status
pyarrow.lib.ArrowNotImplementedError: Cannot write struct type 'prediction' with no child field to Parquet. Consider adding a dummy child field.
The above exception was the direct cause of the following exception:
Traceback (most recent call last):
File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 1339, in compute_config_parquet_and_info_response
parquet_operations = convert_to_parquet(builder)
^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 972, in convert_to_parquet
builder.download_and_prepare(
File "/usr/local/lib/python3.12/site-packages/datasets/builder.py", line 894, in download_and_prepare
self._download_and_prepare(
File "/usr/local/lib/python3.12/site-packages/datasets/builder.py", line 970, in _download_and_prepare
self._prepare_split(split_generator, **prepare_split_kwargs)
File "/usr/local/lib/python3.12/site-packages/datasets/builder.py", line 1702, in _prepare_split
for job_id, done, content in self._prepare_split_single(
^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/site-packages/datasets/builder.py", line 1858, in _prepare_split_single
raise DatasetGenerationError("An error occurred while generating the dataset") from e
datasets.exceptions.DatasetGenerationError: An error occurred while generating the datasetNeed help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
id int64 | annotations list | file_upload string | drafts list | predictions list | data dict | meta dict | created_at timestamp[us, tz=UTC] | updated_at timestamp[us, tz=UTC] | allow_skip bool | inner_id int64 | total_annotations int64 | cancelled_annotations int64 | total_predictions int64 | comment_count int64 | unresolved_comment_count int64 | last_comment_updated_at null | project int64 | updated_by int64 | comment_authors list |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
263 | [
{
"bulk_created": false,
"completed_by": 1,
"created_at": "2026-01-03T14:05:30.855478Z",
"draft_created_at": "2026-01-03T14:04:43.739843Z",
"ground_truth": false,
"id": 25,
"import_id": null,
"last_action": null,
"last_created_by": null,
"lead_time": 6075.672999999999,
"p... | 69520220-clip2_caos_READY.mp4 | [] | [] | {
"video": "/data/upload/5/69520220-clip2_caos_READY.mp4"
} | {} | 2026-01-03T14:04:07.504000 | 2026-01-03T15:46:12.204000 | true | 1 | 1 | 0 | 0 | 0 | 0 | null | 5 | 1 | [] |
🚀 compaTAI: Mexico City Chaos Dataset (Video Alpha Sample)
Training AI to survive where the rules are suggestions.
compaTAI presents a high-entropy Video Tracking Dataset captured in the chaotic heart of Mexico City (CDMX). While standard datasets (Waymo, nuScenes) are built on predictable environments, compaTAI focuses on "The Chaos Edge Case": the informal and non-linear traffic behaviors unique to Latin American megacities.
⬇️ Raw videos available for download — annotate them yourself and train your own models.
Annotations are community-driven. Buy a pack, label it your way, contribute back.
🌟 Why Video Tracking?
Unlike static image datasets, our video sequence provides temporal consistency. This allows models to train for Multi-Object Tracking (MOT) and Behavior Prediction, essential for navigating environments where lane markings don't exist and movement is erratic.
🏭 How the Data is Produced
Each dataset batch originates from a real 12–16 hour taxi shift in Mexico City, generating approximately 53 raw dashcam videos (~1 GB each). These are processed through the compaTAI pipeline on AWS EC2 + Amazon Bedrock:
- Frame extraction — 30 keyframes sampled per video
- AI scene analysis — Bedrock Vision classifies behaviors per frame (confidence ≥ 0.82)
Clip segmentation — 8–10 representative 15-second clips selected per video
Pack assembly — 7–8 thematic packs of 10–30 clips per batch, organized by category
This means new data drops continuously — every shift adds fresh real-world chaos footage.
📊 Alpha Sample Specifications
- | Field | Value |
- |---|---|
- | Total Duration | 52.06 seconds (alpha clip) |
- | Total Frames | 1,562 (at 30 FPS) |
- | Resolution | HD (Fast-Start Streaming) |
- | Annotation Format | Label Studio JSONL (Temporal Bounding Box) |
- | Location 1 | Mexico City — Critical Junctions |
- | Location 2 | State of Mexico — Critical Junctions |
- | Type | 2D Video Rectangle Tracking + Interpolated Keyframes |
🏷️ The "Chaos" Taxonomy
Our labels capture the specific Mexican Edge Cases that standard sensors often misinterpret:
Label Description Why it matters Pedestrian_IrregularPedestrians crossing mid-avenue or between cars Predicts non-linear human trajectory Street_VendorMobile vendors navigating active traffic lanes Unique obstacle detection for informal economies Infrastructure_DeficitMissing signals, potholes, zero lane markings Trains defensive driving and path planning Moto_FilteringMotorcycles weaving between lanes at high speed High-frequency proximity detection Microbus_StopPublic transport stopping mid-road Predicts sudden traffic flow interruptions Double_ParkingVehicles blocking lanes informally Urban navigation in constrained spaces Intersection_NegotiationUncontrolled intersection crossing logic Decision-making without traffic signals Cultural_JaywalkingPedestrian patterns unique to CDMX culture Behavior prediction in high-density zones
🛒 Buy Raw Video Packs — Make Your Own Annotations
This alpha sample is free. Full packs with raw
.mp4clips (15s each, YOLO-ready) are available at store.compatai.mx.No annotations included — you get the raw clips to label with your own tools and taxonomy. Perfect for teams building custom models.
🚗 Informal Parking —
cat_parking
Pack Clips Price File Starter 10 clips × 15s $29 cat_parking_pack1.zipPro 20 clips × 15s $49 cat_parking_pack2.zipFull 30 clips × 15s $79 cat_parking_pack3.zip🏍️ Motorcycle Weaving —
cat_motos
Pack Clips Price File Starter 10 clips × 15s $29 cat_motos_pack1.zipPro 20 clips × 15s $49 cat_motos_pack2.zipFull 30 clips × 15s $79 cat_motos_pack3.zip🚦 Intersection Chaos —
cat_intersections
Pack Clips Price File Starter 10 clips × 15s $29 cat_intersections_pack1.zipPro 20 clips × 15s $49 cat_intersections_pack2.zipFull 30 clips × 15s $79 cat_intersections_pack3.zip🚌 Microbus Behavior —
cat_microbus
Pack Clips Price File Starter 10 clips × 15s $29 cat_microbus_pack1.zipPro 20 clips × 15s $49 cat_microbus_pack2.zipFull 30 clips × 15s $79 cat_microbus_pack3.zip🚶 Jaywalking + 🛒 Vendors + 🕳️ Infrastructure —
3 categories bundle
Pack Clips Price File Starter 10 clips × 15s $29 cat_bundle3_pack1.zipPro 20 clips × 15s $49 cat_bundle3_pack2.zipFull 30 clips × 15s $79 cat_bundle3_pack3.zip
🛠️ Data Structure
The
.jsonlfile contains the temporal sequence of each object. Each annotation includes:
frame— exact frame number
x, y, width, height— normalized coordinates (0–100) for resolution-independent training
time— exact timestamp within the clip
🚀 How to Access
Download the raw video and JSONL metadata directly from this repository:
from datasets import load_dataset dataset = load_dataset("manuelvaraletai/compaTAI-CDMX-Alpha")⚠️ A compaTAI visualization utility script is coming soon.
💰 Enterprise & Custom Services
This repository is a technical sample. compaTAI offers specialized data for enterprise-grade autonomous systems:
- Full Video Datasets — 100+ hours of labeled CDMX traffic footage
- Specific Edge Cases — Custom recordings: "Rainy Nights", "Peak Hour at Indios Verdes", etc.
- Custom Labeling — We use the compaTAI pipeline to label your raw data with our specialized taxonomy
Enterprise Licensing & Custom Data Requests → manuel.vargas@compatai.mx
🔗 Visit our store: store.compatai.mx
- Downloads last month
- 74