Model reference — Video
This service counts how many people entered and left through a doorway or across a line, and reports occupancy over time — footfall by hour, peak load, dwell in a zone — without identifying anybody.
Counting requires tracking, not just detection. A person standing in a doorway appears in hundreds of frames, and to count them once the system must follow them as one continuous track and record which way they crossed the line. That is why the tracker matters more than the detector here: most counting errors come from a track being lost and restarted, which counts one person twice. Camera placement is decisive — an overhead or steeply angled view counts far more accurately than a view along a corridor where people occlude each other.
Models in this group take a recorded video file as input and are applied frame by frame. Each specification table gives three hardware tiers — Minimum, the smallest setup on which the model runs correctly; Medium, the usual production configuration; and High, a configuration sized for peak volume. The rate is the number of sample inputs processed per hour on that hardware. Use these rates for initial sizing. Before production, benchmark your own data to validate accuracy, latency, throughput and cost. The sample input here is one minute of 1080p video at 25 frames per second.
Vendor: Ultralytics
What it does: detects people on every frame quickly and accurately, in a range of sizes so the same family serves a camera-side device and a central server. The usual detector for counting.
| Requirement | Minimum | Medium | High |
|---|---|---|---|
| GPU type | RTX 3060 | RTX 4090 | A100 80 GB |
| VRAM | 6 GB | 24 GB | 80 GB |
| vCPUs | 4 | 8 | 24 |
| RAM | 16 GB | 32 GB | 64 GB |
| Server | 1× RTX 3060 12 GB | 1× RTX 4090 24 GB | 1× A100 SXM 80 GB |
| Rate (video minutes processed per hour) | ≈ 1,600 | ≈ 6,000 | ≈ 16,000 |
Vendor: Baidu (PaddlePaddle)
What it does: a detector with cleaner, more consistent output frame to frame, which reduces the flickering detections that cause a tracker to lose someone.
| Requirement | Minimum | Medium | High |
|---|---|---|---|
| GPU type | RTX 3060 | RTX 4090 | A100 80 GB |
| VRAM | 8 GB | 24 GB | 80 GB |
| vCPUs | 6 | 12 | 24 |
| RAM | 16 GB | 32 GB | 64 GB |
| Server | 1× RTX 3060 12 GB | 1× RTX 4090 24 GB | 1× A100 SXM 80 GB |
| Rate (video minutes processed per hour) | ≈ 700 | ≈ 2,600 | ≈ 7,000 |
Vendor: Huazhong University of Science and Technology
What it does: follows each person as one track through the scene, which is what turns detections into counts. Fast, accurate and the usual default.
| Requirement | Minimum | Medium | High |
|---|---|---|---|
| GPU type | RTX 3060 | RTX 4090 | A100 80 GB |
| VRAM | 6 GB | 24 GB | 80 GB |
| vCPUs | 4 | 8 | 24 |
| RAM | 16 GB | 32 GB | 64 GB |
| Server | 1× RTX 3060 12 GB | 1× RTX 4090 24 GB | 1× A100 SXM 80 GB |
| Rate (video minutes processed per hour) | ≈ 1,200 | ≈ 4,000 | ≈ 10,000 |
Vendor: Tel Aviv University
What it does: recovers a track after a person is briefly hidden behind another, which prevents one crossing being counted as two. Worth its extra cost in busy doorways.
| Requirement | Minimum | Medium | High |
|---|---|---|---|
| GPU type | RTX 3060 | RTX 4090 | A100 80 GB |
| VRAM | 6 GB | 24 GB | 80 GB |
| vCPUs | 4 | 8 | 24 |
| RAM | 16 GB | 32 GB | 64 GB |
| Server | 1× RTX 3060 12 GB | 1× RTX 4090 24 GB | 1× A100 SXM 80 GB |
| Rate (video minutes processed per hour) | ≈ 800 | ≈ 2,800 | ≈ 7,000 |
Vendor: community-maintained open-source project
What it does: the long-established tracker, lighter and well understood, adequate for quiet entrances with clear separation between people.
| Requirement | Minimum | Medium | High |
|---|---|---|---|
| GPU type | RTX 3060 | RTX 4090 | A100 80 GB |
| VRAM | 6 GB | 24 GB | 80 GB |
| vCPUs | 4 | 8 | 24 |
| RAM | 16 GB | 32 GB | 64 GB |
| Server | 1× RTX 3060 12 GB | 1× RTX 4090 24 GB | 1× A100 SXM 80 GB |
| Rate (video minutes processed per hour) | ≈ 1,400 | ≈ 5,000 | ≈ 13,000 |
Vendor: Tencent Youtu Lab
What it does: counts people directly in crowded frames where individual tracking breaks down, used to cross-check occupancy figures at peak.
| Requirement | Minimum | Medium | High |
|---|---|---|---|
| GPU type | RTX 3060 | RTX 4090 | A100 80 GB |
| VRAM | 8 GB | 24 GB | 80 GB |
| vCPUs | 6 | 12 | 24 |
| RAM | 16 GB | 32 GB | 64 GB |
| Server | 1× RTX 3060 12 GB | 1× RTX 4090 24 GB | 1× A100 SXM 80 GB |
| Rate (video minutes processed per hour) | ≈ 240 | ≈ 900 | ≈ 2,400 |
Accuracy depends on camera angle, footfall density and how people move through the space. Our consultants review your camera views and count a sample against manual counts, then recommend the detector, tracker and line placement, and report the accuracy achieved.
At the start of a project we may run a short proof of concept on a sample of your own data, measuring the accuracy and the throughput the model actually achieves on your material. That replaces the estimates on this page with real figures, so the cost and the schedule for the full engagement are known before it is committed.
Share a representative sample, expected volume, latency target and deployment location for People entrance / exit counting. We benchmark the shortlisted models, recommend the lowest-cost GPU configuration that meets the target, and scale it from pilot capacity to a dedicated production cluster.