Tracker comparison · FastTrack vs Deep OC-SORT

FastTrack vs Deep OC-SORT

A head-to-head of FastTrack and Deep OC-SORT for multi-object tracking behind YOLO: how fast each runs across 10 NVIDIA GPUs and how steadily it holds object IDs, measured on the same clip and the same detector.

The verdict: On an NVIDIA H100, FastTrack is the quicker of the two at 104 FPS versus 101 for Deep OC-SORT. Rankings shift on other hardware, though. Deep OC-SORT posts the top FPS on 8 of the 10 tiers. FastTrack is the steadier tracker: 24 ID fragmentations against 42 for Deep OC-SORT (43% fewer), across 12 unique IDs to 12 on the same clip. That hands FastTrack both the speed and the cleaner IDs on this hardware, making it the safer default: step up to Deep OC-SORT for crowded or occluded scenes where identity persistence matters.

Faster overall
Deep OC-SORT

Higher FPS on 8 of the 10 GPU tiers tested.

Steadier IDs
FastTrack

Only 24 ID fragmentations, versus 42 for Deep OC-SORT.

ID-stability: what a fragmentation actually looks like

How well a tracker holds a single, consistent ID on each object is a property of the algorithm, not the GPU, so these numbers barely move across hardware; we report them once (measured on the NVIDIA H100). This clip has no MOT ground truth, so instead of MOTA/IDF1 we report the raw stability signals: how many distinct IDs the tracker created, how many times a track was broken (fragmentations), and the average track length. Fewer IDs and fewer fragmentations mean steadier identities.

TrackerApproachUnique IDsFragmentationsAvg track length
FastTrackMotion + Re-ID122483.8
Deep OC-SORTMotion + Re-ID124280.4

Speed by GPU

FastTrack and Deep OC-SORT, ranked fastest-first. Pick any GPU to see which one leads on that hardware. End-to-end detection + tracking throughput (frames per second, higher is better) for FastTrack and Deep OC-SORT across all 10 GPU tiers. GPUs are ordered flagship-first.

Show GPUs
Frames per second for 2 trackers across 10 GPU tiers. FastTrack ranges 61 to 147 FPS; Deep OC-SORT ranges 60 to 155 FPS.04080120160FPSB200H200H100RTX PRO 6000A100 80GBA100 40GBL40SA10L4T4
  • FastTrack
  • Deep OC-SORT

FastTrack vs Deep OC-SORT: the four questions that decide it

Every figure below is computed from the same run as the tables above: throughput, per-frame latency, cost and track continuity for this pair specifically, rather than a general ranking.

Real-time headroom

Across the 10 tiers, FastTrack clears 30 FPS on 10 and 60 FPS on 10; Deep OC-SORT clears 30 FPS on 10 and 60 FPS on 10. Both hold real time on every GPU tested, so this pair is decided on ID behaviour and cost rather than speed. Per frame on an NVIDIA H100 that is 9.6 ms for FastTrack and 9.9 ms for Deep OC-SORT, detection included.

Does the ranking hold across GPUs?

No, and this is the part a single-GPU benchmark hides. FastTrack leads on the NVIDIA H100, but Deep OC-SORT takes the lead on NVIDIA B200, NVIDIA H200, NVIDIA RTX PRO 6000, NVIDIA A100 80GB, NVIDIA A100 40GB, NVIDIA L40S, NVIDIA A10 and NVIDIA L4. If you are sizing hardware, pick the tracker on the tier you will actually deploy to, not on the flagship the benchmark headline was measured on.

How long an identity survives

Average track length was 83.8 frames for FastTrack and 80.4 for Deep OC-SORT on the 200-frame clip, so FastTrack held each object for longer before losing or re-numbering it. Read alongside the fragmentation counts, that is what an ID switch feels like downstream: a counting line double-counts the same person, or a dwell-time average collapses because one visit was recorded as three. Neither number is MOTA, because this clip has no ground truth, but both come from the same run and point the same way.

Cost to run

At NVIDIA H100 rates, FastTrack costs $0.0106 per 1,000 frames against $0.0108 for Deep OC-SORT. Over a single 30 FPS camera running for an hour that is roughly $1.14 versus $1.17. That is small per camera, and a real multiplier across a wall of them. Cost here is pure occupancy: a slower tracker holds the GPU for longer, so throughput and spend are the same fact in two units.

Choosing between them

Choose FastTrack if…

Teams that want high FPS without ByteTrack's occasional ID churn.

Strengths

  • Top-tier FPS on most GPUs
  • Very stable IDs (few fragmentations)
  • Strong all-round default

Trade-offs

  • Newer, smaller community than ByteTrack/BoT-SORT
  • Appearance step adds a little overhead vs pure motion

Motion + Re-ID, 2024. Enable it with tracker="fasttrack.yaml".

Choose Deep OC-SORT if…

Crowded or occluded scenes where identity persistence matters.

Strengths

  • Strong ID stability from appearance Re-ID
  • Fast on Hopper/Blackwell GPUs
  • Good occlusion recovery

Trade-offs

  • Re-ID model adds compute and memory
  • Slower than pure-motion trackers on budget GPUs

Motion + Re-ID, 2023. Enable it with tracker="deepocsort.yaml".

Other head-to-head comparisons

More matchups involving FastTrack and Deep OC-SORT, each with the same speed, cost and ID-stability breakdown.

See the full 6-tracker benchmark across 10 GPUs →

How these numbers were measured

FastTrack and Deep OC-SORT ran on the identical 200-frame clip with the same yolo26n.pt detector on each of the 10 GPU tiers, so the only variable is the tracker. There is no MOT ground truth on this clip, so no MOTA or IDF1 is claimed; the stability figures are raw counts. Full methodology and the benchmark script live on the hub, alongside all six trackers.

Blog

From the blog

Tutorials, code, and notes on computer vision, deep learning, and applied AI.

View all posts →

Frequently asked questions

Is FastTrack faster than Deep OC-SORT?
On an NVIDIA H100 with yolo26n.pt, FastTrack ran at 104 FPS and Deep OC-SORT at 101 FPS end to end, detection included. Across all 10 GPU tiers FastTrack was ahead on 2 and Deep OC-SORT on 8. The ranking is not constant across the range, so check the tier you will deploy on.
Which holds object IDs better, FastTrack or Deep OC-SORT?
FastTrack. On the same clip it recorded 24 ID fragmentations against 42, across 12 unique IDs for FastTrack and 12 for Deep OC-SORT, with average track lengths of 83.8 and 80.4 frames. ID stability is a property of the algorithm rather than the GPU, so this holds on any hardware.
Is FastTrack or Deep OC-SORT cheaper to run?
FastTrack, at $0.0106 per 1,000 frames on an NVIDIA H100 versus $0.0108. The difference is occupancy: the slower tracker holds the GPU longer per frame, so the cost gap tracks the speed gap.
Can FastTrack and Deep OC-SORT run in real time?
FastTrack sustained 30 FPS or better on 10 of the 10 GPU tiers and Deep OC-SORT on 10. Per frame on an NVIDIA H100 that is 9.6 ms and 9.9 ms respectively. Anything under 33 ms per frame keeps up with a 30 FPS camera.
Should I choose FastTrack or Deep OC-SORT?
Choose FastTrack for teams that want high FPS without ByteTrack's occasional ID churn. Choose Deep OC-SORT for crowded or occluded scenes where identity persistence matters. Near-ByteTrack speed with much steadier IDs. OC-SORT plus a Re-ID model for stable identities.
Does a more expensive GPU make tracking more accurate?
No. A faster GPU only makes tracking run faster; it does not change how accurately the tracker follows objects. Accuracy and ID-stability depend on the tracking algorithm and your detector, not the hardware. That is why this page reports speed per GPU, but ID-stability only once.
How was this YOLO tracker benchmark run?
Every tracker ran on the same 200-frame clip with the same yolo26n.pt detector, on each of the 10 GPU tiers. FPS is the end-to-end detection-plus-tracking rate. All six trackers are built into Ultralytics, so results are reproducible with a single script.