SAVRN
Search Contact SAVRN

Dataset · Object detection

cati-singapore-dataset

by Suhas Reddy SuhxsReddy/cati-singapore-dataset

Real-time vehicle detection data collected from Singapore's 90 LTA traffic cameras using CATI (Context-Aware Traffic Intelligence) — a novel FiLM-conditioned YOLOv11 detector that adapts to environmental conditions in real time.

Rows
Configurations
Size143.1 MB
Licensemit
AccessPublicly accessible
Monthly Downloads3.7k

Dataset Card

By Suhas Reddy, published under mit, revision 6b1e7befd203.

CATI Singapore Expressway Traffic Dataset

Real-time vehicle detection data collected from Singapore's 90 LTA traffic cameras using CATI (Context-Aware Traffic Intelligence) — a novel FiLM-conditioned YOLOv11 detector that adapts to environmental conditions in real time.

Dataset Description

This dataset contains per-camera vehicle detection results collected continuously from Singapore's Land Transport Authority (LTA) expressway camera network. Each record captures a full detection sweep of a single camera including vehicle counts, class breakdown, directional split, and environmental context.

Coverage

Expressway Cameras
CTE (Central Expressway) ~16
PIE (Pan-Island Expressway) ~20
AYE (Ayer Rajah Expressway) ~8
ECP (East Coast Parkway) ~10
MCE (Marina Coastal Expressway) ~4
TPE (Tampines Expressway) ~8
BKE (Bukit Timah Expressway) ~6
KJE (Kranji Expressway) ~6
SLE (Seletar Expressway) ~6

Data Schema

Column Type Description
timestamp string Detection timestamp (SGT, UTC+8)
camera_id int LTA camera ID
road string Expressway code (CTE, PIE, AYE, etc.)
lat float Camera latitude (WGS84)
lon float Camera longitude (WGS84)
weather string Central Singapore weather at time of capture
total_vehicles int Total vehicles detected
dir_a int Vehicles moving in direction A (tracked via 2-frame IoU)
dir_b int Vehicles moving in direction B
car int Cars detected
motorcycle int Motorcycles detected
bus int Buses detected
truck int Trucks detected
van int Vans detected
lorry int Lorries detected
conf_threshold float Detection confidence threshold used
iou_threshold float NMS IoU threshold used
imgsz int Inference image size (px)
model_version string Model checkpoint identifier
dir_a_label string Human-readable direction A label (e.g. "towards Woodlands")
dir_b_label string Human-readable direction B label (e.g. "towards City")
is_ramp bool True if camera is likely on an entry/exit ramp (proximity heuristic)

Versioning note: Filter by conf_threshold, iou_threshold, and imgsz to isolate consistent collection windows. Exclude is_ramp=True rows for mainline-only throughput analysis.

Model

Detections are produced by CATI — a novel architecture that injects FiLM (Feature-wise Linear Modulation) layers into YOLOv11s, conditioning the backbone on real-time environmental metadata:

  • Weather condition and temperature (data.gov.sg)
  • Time of day (cyclical encoding)
  • Camera GPS position (sinusoidal positional encoding)
  • PM2.5 air quality
  • Camera resolution class

Model weights: SuhxsReddy/cati-singapore

How to Load

from datasets import load_dataset
import pandas as pd

ds = load_dataset("SuhxsReddy/cati-singapore-dataset")
df = ds["train"].to_pandas()

# Vehicles by road
print(df.groupby("road")["total_vehicles"].mean())

# Directional flow on CTE
cte = df[df["road"] == "CTE"]
print(cte[["timestamp", "camera_id", "dir_a", "dir_b", "total_vehicles"]])

Collection

  • Sweep interval: ~90 seconds (full network scan)
  • Detection threshold: confidence ≥ 0.08, IoU ≤ 0.25
  • Image source: LTA Datamall Traffic Images API
  • Infrastructure: HuggingFace Spaces (CPU), continuous 24/7 collection

License

MIT — data sourced from Singapore's open data APIs under the Singapore Open Data Licence.

Citation

@dataset{cati_singapore_2026,
  author    = {SuhxsReddy},
  title     = {CATI Singapore Expressway Traffic Dataset},
  year      = {2026},
  publisher = {HuggingFace},
  url       = {https://huggingface.co/datasets/SuhxsReddy/cati-singapore-dataset}
}

Details

Repository
SuhxsReddy/cati-singapore-dataset
Publisher
Suhas Reddy
Task category
Object detection
Tags
traffic, singapore, expressway
Size category
10K<n<100K
Languages
en
Revision
6b1e7befd2037f848b4434a0168c405a28703c63
Last updated
2026-09-18

Files

128 files, 143.1 MB in total.

Data3 files · 125.1 MB
Documentation1 file · 4.3 KB
Other123 files · 18.0 MB
Repository1 file · 2.6 KB
Every file
FileTypeSizeSHA-256
camera_network.jsonData109.5 KB
cati_detections.csvData8.6 MB
cati_detections_v2.csvData116.5 MB6a771c3a1bf2
README.mdDocumentation4.3 KB
annotated/cam_1001_KPE.jpgOther16.1 KBad3bc7d0b030
annotated/cam_1002_CTE.jpgOther15.4 KB9818951d46e5
annotated/cam_1002_KPE.jpgOther19.3 KBb0dedbde362e
annotated/cam_1003_CTE.jpgOther20.3 KB83bef3659e39
annotated/cam_1003_PIE.jpgOther23.6 KBb71aa3254348
annotated/cam_1004_CTE.jpgOther15.8 KB547eeb6370ee
annotated/cam_1004_PIE.jpgOther13.8 KB0b96774901d0
annotated/cam_1005_CTE.jpgOther15.4 KBf3e73359b9a8
annotated/cam_1005_KPE.jpgOther16.5 KB9a0d9a6addf4
annotated/cam_1006_CTE.jpgOther22.6 KB7e48c0278eb9
annotated/cam_1006_KPE.jpgOther20.5 KB0015791ab17f
annotated/cam_1111_CTE.jpgOther181.5 KB587a4930b12c
annotated/cam_1111_LOYANG AVE.jpgOther252.4 KBa20f5e5880e4
annotated/cam_1112_CTE.jpgOther240.9 KBa467601794b8
annotated/cam_1112_LOYANG AVE.jpgOther279.9 KBa8fd8310fcf7
annotated/cam_1113_CTE.jpgOther141.4 KB2dc8dc0561c8
annotated/cam_1113_TANAH MERAH COAST ROAD.jpgOther221.0 KB8c8fff64ed3b
annotated/cam_1501_CTE.jpgOther18.2 KB37b963d9559c
annotated/cam_1501_MCE.jpgOther19.0 KBd0fb8be8cc36
annotated/cam_1503_CTE.jpgOther13.4 KB58f49c931bdb
annotated/cam_1503_MCE.jpgOther12.2 KBac0a29128615
annotated/cam_1504_CTE.jpgOther13.0 KB22060522a024
annotated/cam_1505_MCE.jpgOther7.3 KB083b84d02f6d
annotated/cam_1701_CTE.jpgOther175.1 KB25f5c589dac3
annotated/cam_1702_CTE.jpgOther109.0 KB57e417d6d95d
annotated/cam_1703_CTE.jpgOther170.1 KB864e2b6c3084
annotated/cam_1704_CTE.jpgOther113.0 KB1951376d2c3b
annotated/cam_1705_CTE.jpgOther136.3 KB7ff446aacc7b
annotated/cam_1706_CTE.jpgOther175.2 KBa50d4a79c86e
annotated/cam_1707_CTE.jpgOther226.1 KBda272a56267d
annotated/cam_1709_CTE.jpgOther106.9 KB015d423b233c
annotated/cam_1711_CTE.jpgOther148.6 KB9a16992a4e0f
annotated/cam_2701_CTE.jpgOther175.8 KB7dc60f6cb56a
annotated/cam_2701_WOODLANDS CAUSEWAY.jpgOther165.3 KB75d619d7832a
annotated/cam_2702_CTE.jpgOther141.9 KBf942277209c9
annotated/cam_2702_SLE.jpgOther238.1 KBa7dbfcabb5d1
annotated/cam_2703_CTE.jpgOther193.7 KBe1039521773d
annotated/cam_2703_SLE.jpgOther196.1 KB158d6d4608be
annotated/cam_2704_CTE.jpgOther174.8 KB79691d24732f
annotated/cam_2704_WOODLANDS AVE.jpgOther259.3 KBad0c2cb84040
annotated/cam_2705_CTE.jpgOther118.3 KB60fb3718b003
annotated/cam_2705_SLE.jpgOther164.6 KBb040a51f3c19
annotated/cam_2706_CTE.jpgOther75.3 KB5d2026fe93b3
annotated/cam_2706_SLE.jpgOther117.1 KB564d277d381a
annotated/cam_2707_CTE.jpgOther113.4 KBc84e9bf787e9
annotated/cam_2707_SLE.jpgOther112.4 KB4a68bd50a904
annotated/cam_2708_CTE.jpgOther160.7 KB237ad5bfd472
annotated/cam_2708_SLE.jpgOther170.3 KB90b464b49b80
annotated/cam_3702_ECP.jpgOther209.2 KB28e8a5e0cd92
annotated/cam_3704_ECP.jpgOther189.5 KB7b7ca961e956
annotated/cam_3705_ECP.jpgOther219.2 KB65954cdf2382
annotated/cam_3793_ECP.jpgOther209.5 KB3b68596e8c7a
annotated/cam_3795_ECP.jpgOther216.9 KBf3afac71adf8
annotated/cam_3796_ECP.jpgOther173.7 KBce14da838e6a
annotated/cam_3797_ECP.jpgOther187.6 KB04fbd96f0a8d
annotated/cam_3798_ECP.jpgOther193.9 KBb9047ee61abf
annotated/cam_4701_PIE.jpgOther133.4 KBcd5dfb474a80
annotated/cam_4702_PIE.jpgOther83.6 KBb3e88d8e780f
annotated/cam_4703_PIE.jpgOther237.6 KB0d9cfa4094a4
annotated/cam_4703_TUAS SECOND LINK.jpgOther255.1 KBb51c2376ea0b
annotated/cam_4704_PIE.jpgOther189.5 KB668832d0a83a
annotated/cam_4705_PIE.jpgOther137.3 KBcbd1fc603605
annotated/cam_4706_PIE.jpgOther112.8 KBb096fbf9112c
annotated/cam_4707_PIE.jpgOther169.0 KBa40b39007b7e
annotated/cam_4708_PIE.jpgOther133.3 KB92e9573a0786
annotated/cam_4709_PIE.jpgOther130.5 KB426465511d9c
annotated/cam_4710_PIE.jpgOther171.0 KB62cecde5398a
annotated/cam_4712_PIE.jpgOther255.4 KBb1875c57408f
annotated/cam_4713_PIE.jpgOther205.5 KB0d4c51d8dc90
annotated/cam_4714_PIE.jpgOther351.0 KBa4b3bea448f6
annotated/cam_4716_PIE.jpgOther248.5 KB82a920710502
annotated/cam_4798_AYE.jpgOther194.3 KB2b091145cd74
annotated/cam_4798_PIE.jpgOther213.3 KB38796fbe2743
annotated/cam_4799_AYE.jpgOther182.6 KBdf799ed95930
annotated/cam_4799_PIE.jpgOther175.7 KBef7131b3ab88
annotated/cam_5794_AYE.jpgOther205.2 KBbfeaa19de863
annotated/cam_5795_AYE.jpgOther164.4 KBee3ef9f138b5
annotated/cam_5797_AYE.jpgOther123.1 KB8cc25aa34ee3
annotated/cam_5798_AYE.jpgOther158.6 KBc6f230fc2dd8
annotated/cam_5799_AYE.jpgOther153.8 KBde3d74b1ecdc
annotated/cam_6701_ECP.jpgOther146.4 KB5d12e95e17df
annotated/cam_6701_PIE.jpgOther132.9 KB6d7788402ce3
annotated/cam_6703_MCE.jpgOther131.0 KB40243711c184
annotated/cam_6703_PIE.jpgOther129.2 KBe55d8e0f097b
annotated/cam_6704_MCE.jpgOther180.9 KBa84cbd88a9d6
annotated/cam_6704_PIE.jpgOther170.1 KBa19ae0e3447e
annotated/cam_6705_MCE.jpgOther95.4 KB57017fa96846
annotated/cam_6705_PIE.jpgOther96.1 KB5510356e13be
annotated/cam_6706_ECP.jpgOther127.9 KB0de6bd25ed38
annotated/cam_6706_PIE.jpgOther105.2 KB8501e49ec6fb
annotated/cam_6708_ECP.jpgOther133.8 KB815da6afbab3
annotated/cam_6708_PIE.jpgOther143.5 KB65fa5f4cba65
annotated/cam_6710_ECP.jpgOther169.1 KB2b02298dbd21
annotated/cam_6710_PIE.jpgOther168.2 KB123151da8bd1
annotated/cam_6711_ECP.jpgOther114.9 KBa5a2f6809b32
annotated/cam_6711_PIE.jpgOther112.3 KBee5c4842444c
annotated/cam_6712_ECP.jpgOther197.5 KB54a331367c1a
annotated/cam_6712_PIE.jpgOther210.8 KB99ad5facaa4a
annotated/cam_6713_ECP.jpgOther149.7 KB81c7760f9f20
annotated/cam_6713_PIE.jpgOther94.0 KBb76f30fb5032
annotated/cam_6714_ECP.jpgOther155.5 KB37648dea23a0
annotated/cam_6714_PIE.jpgOther162.0 KBb2a2889ecce0
annotated/cam_6715_ECP.jpgOther162.0 KB1178cca3c8d0
annotated/cam_6715_PIE.jpgOther144.7 KB3e43f1dca95a
annotated/cam_6716_ECP.jpgOther175.6 KB5dc79ab9a0f0
annotated/cam_6716_PIE.jpgOther167.9 KB4e98c4d21ca1
annotated/cam_7791_TPE.jpgOther149.2 KB568c12d8e140
annotated/cam_7793_TPE.jpgOther196.0 KBa0d6f63a50d9
annotated/cam_7794_TPE.jpgOther138.3 KBbaf5c13752ea
annotated/cam_7795_TPE.jpgOther107.3 KB16cc0b9fd36b
annotated/cam_7796_TPE.jpgOther182.4 KBb84875f5b2ae
annotated/cam_7797_TPE.jpgOther179.0 KB3603753ffb87
annotated/cam_7798_TPE.jpgOther163.0 KB38680a837f6f
annotated/cam_8701_KJE.jpgOther158.9 KBb58ceb99e32b
annotated/cam_8702_KJE.jpgOther146.4 KB7b7dfc4a906c
annotated/cam_8704_KJE.jpgOther170.2 KB82e0f3aefc46
annotated/cam_8706_KJE.jpgOther160.3 KB5d5c98573b57
annotated/cam_9701_BKE.jpgOther132.9 KB96935e12c650
annotated/cam_9702_BKE.jpgOther133.9 KBf58e9edee9e0
annotated/cam_9703_BKE.jpgOther209.7 KB3defd4517b94
annotated/cam_9704_BKE.jpgOther174.8 KBc899d5b69fce
annotated/cam_9705_BKE.jpgOther164.7 KB7d1b2f83c90b
annotated/cam_9706_BKE.jpgOther112.2 KB1976f42d1266
.gitattributesRepository2.6 KB

License and Download

License
mit
Access
No access gate
Download from Suhas Reddy

Released by Suhas Reddy through its official repository on Hugging Face. Read the license.