Available for delivery at short notice.

Playcraft ATCC

Automatic traffic counting and classification from footage you already have: CCTV, drone or a phone on a pole. No roadside hardware, no per-kilometre fee, and the footage never leaves your building.

Divided carriageway, both directions

26 JUL 2025 · 06:15
POS00:00:00 / 00:00:10
A five-axle truck, a two-axle truck and two motorcycles tracked at once across a divided carriageway, in opposing directions on a single counting line. The panel increments per class as each crosses.
Runs on
Windows desktop · offline
Input
CCTV, drone or phone-on-pole video
Primary output
Classified volume count with PCU, per-vehicle log and verification video
Licence
Machine-bound, verified offline
Send us 15 minutes of footageSee it run on a call

Back within a working day: the classified count, the annotated verification video, and where your camera position is costing you. You keep the output.

Seen running

What it looks like on real roads

Clips and frames from production surveys. Not staged, not rendered for this page. Play them and watch the counters move as each vehicle crosses.

Five-axle articulated truck

10 JUL 2025 · 09:43
POS00:00:00 / 00:00:05
Axle-count classes are where generic detectors fail — a five-axle truck and a multi-axle vehicle look alike to a model trained on car, truck and bus. Here it lands in the right column, and the tally shows it. Bystanders at the roadside are blurred; we do not publish identifiable people.

The class that is easiest to get wrong

Five seconds at a junction. The class that is easiest to get wrong lands in the right column, and the tally shows it.

Running tally — 5,150 counted27 JUL 2026 · 09:26
A frame taken about an hour and three quarters into a survey. The panel reads 3,099 up, 2,051 down, 5,150 total, with a running split per category: 1,928 up and 1,088 down for two-wheelers, 495 and 245 for three-wheelers, 555 and 468 for cars, 33 and 21 for buses, and further rows for light commercial vehicles, trucks, tractors and trailers. Eight vehicles are boxed and labelled in the same frame, including a bus, three motorcycles, two auto rickshaws, a car and a cycle.
Thousands of vehicles, still per class. Every one of the 5,150 is attributed to a class and a direction as it crosses — no aggregate estimate at the end of the run. Eight vehicles are being tracked in this single frame.
Single frame — four simultaneous tracks26 JUL 2025 · 06:15
A frame of annotated survey video on a divided carriageway. A counting line spans both directions. Four vehicles are boxed and labelled with class, track identifier and confidence: a five-axle truck at 53 per cent, a two-axle truck at 82 per cent and two motorcycles at 68 and 70 per cent. A panel shows two up, one down, three total, broken down per class.
Every vehicle carries its own audit trail. Class, track identifier and confidence on every vehicle, with running totals per class. Any fifteen-minute figure in the report reconciles back to a timestamp in this footage.

Specification

What it does, how it arrives, what you get back

Capabilities

  • 18 vehicle categories plus an Other bucket, following IRC/MoRTH survey convention; the current model has no class for articulated trucks or cycle rickshaws
  • PCU conversion per IRC:64-1990; another schedule, such as IRC:106-1990 for urban roads, is set up on request
  • Directional counts in 15-minute intervals; an hour is four rows added together
  • Per-vehicle log — any interval reconciles back to the source frame
  • Counting line placed visually on a reference frame, in two clicks

Delivery

  • Offline desktop application for Windows
  • Machine-bound licence, verified without network access
  • One-click install; no portal, no token refresh, no outbound traffic

Outputs

  • Classified volume count survey output, both directions and directional split
  • PCU summary per IRC:64-1990
  • Per-vehicle classified log with timestamps and confidence
  • Annotated verification video for manual audit
  • CSV files, which open directly in Excel

What you actually receive

The deliverable, from a real survey

Files from the production run behind the published accuracy figures, not illustrative numbers. The total of 4,595 is the same one that appears as Site A in the accuracy table.

Classified volume count

combined.csv
Class073907450745080008000814Total
Bus8302361
Car/Van/Jeep162369393924
2-Wheeler4019771,0732,451
3-Wheeler128353319800
Mini LCV345064148
2-Axle Truck3191739
Cycle205047117
Total7691,8691,9574,595

Recording began at 07:39, so the first window is partial (six minutes). The count file labels windows by their actual span; the verification index below aligns the same three windows to the quarter-hour.

PCU summary per IRC:64-1990

pcu_summary.csv
CategoryCountFactorPCU
Bus613.0183
Mini Bus61.59
Car / Van / Jeep9241.0924
2-Wheeler2,4510.51,225.5
3-Wheeler8001.0800
Mini LCV1481.0148
LCV11.51.5
2-Axle Truck393.0117
3-Axle Truck103.030
MAV164.572
Tractor61.59
Tractor Trailer164.572
Cycle1170.558.5
Total4,5953,649.5

Tractor recomputed at 1.5 (IRC:64-1990 Table 1, row 3). The file as delivered applied 4.5, the tractor-trailer factor: 27.0 PCU, 3,667.5 in total.

Count and PCU on one scale

pcu_summary.csv

Pick a category to follow it across both bars.

Six categories recorded zero on this survey and are omitted here; the delivered file lists all 19.

Verification index

VERIFICATION.txt
Clock windowVideo positionUpDownTotal
07:30 – 07:4500:00:00 – 00:06:00448321769
07:45 – 08:0000:06:00 – 00:21:001,2096601,869
08:00 – 08:1500:21:00 – 00:36:001,2996581,957
Total2,9561,6394,595

Seven rows of the per-vehicle log, replayed at ¼ speed

crossings_audited.csv
UP
4
DOWN
3
TOTAL
7
ClassUpDown
Car↑ 2↓ 1
2-Wheeler↑ 2↓ 1
3-Wheeler↑ 0↓ 1

LAST ROW

07:39:01 00:00:01 track 1 car 0.851 DOWN

Every row in the delivered log carries clock time, video position, track, class, confidence and direction. This replays seven of them.

Per-vehicle classified log

crossings_audited.csv
ClockVideo pos.TrackClassConf.Dir.
07:39:0000:00:002car0.822UP
07:39:0000:00:004two_wheeler0.627UP
07:39:0000:00:0013two_wheeler0.573UP
07:39:0000:00:0010three_wheeler0.291DOWN
07:39:0000:00:009two_wheeler0.599DOWN
07:39:0100:00:015car0.694UP
07:39:0100:00:011car0.851DOWN

Conf. is the model’s confidence in the class it assigned. Low-confidence rows stay in the file rather than being dropped, so a reviewer can scrub to exactly those crossings.

The verification index is the point. Every clock window maps to a position in the annotated video, so a reviewer can scrub to it, count what crosses, and compare against our figure. The per-vehicle log carries the same pairing on every row. A number you can check is worth more than a number you are asked to trust.

Measured, not claimed

Accuracy, with its method

Every figure here is stated against independent hand counts, and the harder of the two measures is the one we lead with.

Read the full method

Three sites, Indian arterial roads, live uncontrolled traffic

11 August 2026
SiteVehiclesAccuracy
A4,579 manual4,595 counted93.0% weighted99.7% volume
B5,824 manual5,375 counted91.2% weighted92.3% volume
C6,157 manual5,759 counted92.3% weighted93.5% volume
All16,560 manual15,729 counted92.1% weighted95.0% volume

Manual — hand-counted by an independent survey firm Counted — by the software

We lead with 92.1% weighted, not 95.0% volume. Weighted accuracy asks whether every vehicle landed in the right column. Volume accuracy lets over- and under-counts cancel.

Before you ask

Video ATCC for Indian survey work: what engineers ask first

What footage does it need?

Recorded video at 1080p or above — from CCTV, a drone or a phone on a pole — with a bitrate high enough that small, fast vehicles are not smeared by compression. The camera has to see the full width being counted, ideally from a high mounting with a steep view, with the counting line in the nearer half of the frame.

Site requirements, with the evidence behind them

Can it count from CCTV we already have?

Often, but not always. What decides it is how many pixels a two-wheeler occupies where it crosses the counting line, and most corridor CCTV was not mounted for counting. Fifteen minutes of footage from that camera answers the question.

Counting from existing CCTV

Which vehicle classes does it count?

The output has 19 class columns — 18 named vehicle categories plus Other — in 15-minute intervals, by direction, with PCU applied per IRC:64-1990. The current model fills 17 of them: it has no class for articulated trucks or cycle rickshaws, so those two columns read zero. Another schedule, such as IRC:106-1990 for urban roads, is set up on request.

Every class with its PCU value

How accurate is it?

95.0% by volume and 92.1% by weighted class, measured against 16,560 vehicles counted by hand by an independent survey firm. We lead with the lower figure, because volume accuracy lets over- and under-counts cancel.

The full method and per-site results

Does our footage leave the building?

No. It is an offline Windows application with a machine-bound licence that is verified without network access: no portal, no token refresh, no outbound traffic.

What does it cost?

We do not publish prices, because scope decides the number. What moves it is the licence term (lifetime or annual), how many machines it runs on, and what the deployment needs. There is no per-kilometre fee.

How licensing works

Is there a trial licence?

No. Instead, send us 15 minutes of footage. Back within a working day: the classified count, the annotated verification video, and where your camera position is costing you. You keep the output. Or watch it run on footage of your choice on a call.

Reference

Further reading

Also available

Other tools

Next step

Send us 15 minutes of your footage

  • Classified count, per class and per direction
  • Annotated verification video
  • A note on what your camera position is costing you

No obligation. You keep the output. Back within a working day.

Send us footageSee it run on a call