ARC-SORT: Adaptive, Robust, Compensated Multi-Object Tracking with Interacting Multiple Models

Girish Kumar Adari

cs.CVOct 15, 2026

ARC-SORT: Adaptive, Robust, Compensated Multi-Object Tracking with Interacting Multiple Models

Authors: Girish Kumar Adari

Organizations: Athlitix Inc.

Abstract

Multi-object tracking in sports broadcasts is hard for two reasons: athletes move fast and change direction abruptly, and the camera pans and zooms at the same time. We present ARC-SORT (Adaptive, Robust, Compensated SORT), a tracking-by-detection tracker that trains nothing and runs in real time. Its design rule: the tracker measures its own uncertainty, and that measurement decides how it acts. An Interacting Multiple Model (IMM) Kalman filter gives every track a predictive covariance, and we turn it into a per-track search region: the less reliable the motion prediction, the larger the region in which a detection may match (uncertainty-adaptive Expansion IoU, EIoU). On identical inputs, this continuous coupling is what matters: the same covariance used as a hard Mahalanobis gate hurts at every threshold we swept, and richer covariance summaries add nothing measurable. When two tracks cross, the assignment cost itself reports whether the decision was close. ARC-SORT keeps such pairs under review and applies at most one corrective identity swap once post-crossing appearance evidence shows the first decision was wrong; this adds +1.0 HOTA (higher order tracking accuracy) on both the SportsMOT and DanceTrack test sets at under a millisecond per frame. For players who leave the frame for seconds, motion extrapolation carries no information, so our offline re-identification drops the usual motion gate and links tracklets by appearance alone. ARC-SORT reaches 81.82 HOTA on the SportsMOT test set with the fewest identity switches of the 69 leaderboard entries as of August 2026 and tied-third multi-object tracking accuracy (MOTA, 97.09), and 62.1 HOTA on DanceTrack test with the same association settings. On Deep-EIoU's public detections and embeddings it reaches 80.31 HOTA, +3.1 over Deep-EIoU. It runs at about 50 frames per second on a laptop RTX 3060 GPU. Code will be released.

Figures & tables

Panel of Figure 1 — Figure 1: ARC-SORT: one principle, twice instantiated, once taken to its limit. Online (top): the IMM predictive covariance, warped by camera-motion compensation, sets a per-track expansion $e_{i}$ dilating the matching region exactly when the prediction is unreliable; near-ambiguous match pairs (low permutation margin) enter a deferred two-hypothesis state, settled by at most one corrective identity swap per episode (labels only, filter states untouched; orange). Offline (bottom): across multi-second gaps tracklets link by appearance prototypes alone, motion veto removed (orange). OCM, OCR, ORU: observation-centric momentum, recovery, and re-update (Secs. 3.3 , 3.5 ).

Figure 2: The motion gate is the binding constraint (SportsMOT val, pooled HOTA). (a) At

g{=}300

, relaxing the gate tolerance is monotone: default 79.19

\to

removed 84.03. (b) Gate on, sweeping gap cap and threshold is a flatline at

{\approx}79.2

; gate off, a plateau (83.9–84.2), a robust optimum rather than a spike. Star: selected configuration (best worst-sport HOTA).

Method	Tr.	Off.	HOTA $\uparrow$	DetA $\uparrow$	AssA $\uparrow$	MOTA $\uparrow$	IDF1 $\uparrow$	IDs $\downarrow$
DiffMOT [ 27 ]	✓	–	76.2	89.3	65.1	97.1	76.1	–
McByte [ 37 ]	–	–	76.9	–	–	97.2	77.5	–
Deep-EIoU [ 21 ]	–	–	77.2	88.2	67.7	96.3	79.8	2659
SportMamba [ 22 ]	✓	–	77.3	89.5	66.8	96.9	77.7	–
DeepMoveSORT [ 1 ]	✓	–	78.7	88.1	70.3	96.5	81.7	–
Deep HM-SORT [ 16 ]	–	–	80.1	88.3	72.7	96.6	85.2	1703

Table 1: Comparison on the SportsMOT test set. Competitor rows are from each paper’s own results table (“–”: not reported); ARC-SORT rows from the evaluation server. “Tr.”: trains an association, motion, or appearance component. “Off.”: offline tracklet-level (re-)association.

Method	FPS	Scope	Hardware / benchmark (as reported)
McByte [ 37 ]	3–5	full	A100
TDLP [ 3 ]	2.0 (11.4 assoc.)	full	consumer GPU, SportsMOT
NOOUGAT (online) [ 30 ]	12	assoc.	V100, DanceTrack val
CAMELTrack [ 36 ]	13	full	MOT17
Deep HM-SORT [ 16 ]	${\approx}13$	full	L40 (16 GB VM)
Deep-EIoU [ 21 ]	${\approx}14.6$	full	RTX 4080

Table 2: Reported throughput in frames per second (FPS). Competitor numbers are each paper’s own claim on its own hardware and benchmark, so the column is indicative, not a controlled comparison; “assoc.” excludes detection and embedding extraction, “full” includes them. ARC-SORT is measured on a 6 GB laptop GPU.

Method	Tr.	Off.	HOTA $\uparrow$	DetA $\uparrow$	AssA $\uparrow$	MOTA $\uparrow$	IDF1 $\uparrow$
SORT [ 7 ] †	–	–	47.9	72.0	31.2	91.8	50.8
DeepSORT [ 42 ] †	–	–	45.6	71.0	29.7	87.8	47.9
ByteTrack [ 47 ] †	–	–	47.3	71.6	31.4	89.5	52.5
OC-SORT [ 9 ] †	–	✓	55.1	80.4	40.4	92.2	54.9
Deep OC-SORT [ 28 ] †	–	–	61.3	82.2	45.8	92.3	61.5
DiffMOT [ 27 ] †	✓	–	62.3	82.5	47.2	92.8	63.0

Table 3: Comparison on the DanceTrack test set. Competitor rows from each paper’s own table (“–”: not reported); “Tr.”/“Off.” as in Table 1 . † : public DanceTrack YOLOX-X detections (ByteTrack release), the same as ours; ∗ : own detector with extra data. ARC-SORT uses those detections and Deep OC-SORT’s SBS-S50 embeddings.

Mechanism	HOTA $\uparrow$	AssA $\uparrow$	IDF1 $\uparrow$	IDSW $\downarrow$
Mahalanobis veto (best $\tau_{M}$ ) ¶	76.53	67.34	79.88	547
Fixed EIoU ( $e{=}0.7$ )	77.04	68.20	80.52	512
Velocity switch (best of 3) ∗	77.24	68.54	80.73	506
Pure $\sigma$ -DIoU ( $\beta{=}0$ , best $\tau$ ) §	77.46	68.84	80.70	530
Covariance-adaptive (ours)	77.97	69.78	81.48	514
Anisotropic per-axis ( $s{=}2$ )	77.97	69.78	81.49	513

Table 4: Association-mechanism ladder on SportsMOT val (45 sequences, shared cached detections and embeddings): the same per-track covariance offered to every mechanism. The veto collapses association, fixed and switched expansion under-use the signal, continuous adaptivity recovers the full gap.

	SportsMOT			DanceTrack
HOTA	online	chain	test	online	chain	test
without	77.97	84.12	80.82	56.86	58.73	61.1
with	78.80	84.96	81.82	58.36	60.62	62.1
$\Delta$	+0.83	+0.84	+1.00	+1.50	+1.89	+1.00

Table 5: Deferred identity commitment , identical thresholds on both benchmarks. “Online”: tracker with interpolation; “chain”:

+

appearance-only merge; test rows from the evaluation servers.

Appendix figures & tables9 assets

Supplementary material from the paper’s appendix.

Appendix

Figure 3: The mechanism live (SportsMOT val basketball, one track; recorder byte-identical to the ablation anchor). Top:

e_{i}

(Eq. ( 2 )) hugs

e_{\min}

under confident tracking, bumps at maneuver onsets, grows toward

e_{\max}

across a 12-frame miss (gray), and snaps tight on re-capture. Bottom: IMM model probabilities; constant-acceleration takes over during the frame-590 burst: disagreement widens

S

, and the net widens exactly then.

Figure 4: A camera exit, survived (SportsMOT val football). The target (solid blue) leaves the frame at $t{=}230$ and re-enters $293$ frames ( $11.7$ s) later. The appearance-only merge holds one identity across the gap (GT-verified); with the motion veto active (dashed), the terminal-velocity extrapolation lands nowhere near the re-entry and the identity splits $3\to 36$ (amber $\to$ purple).

Rung	HOTA $\uparrow$	AssA $\uparrow$	IDF1 $\uparrow$	IDSW $\downarrow$
Mode ladder ( ablation_eiou_modes )
00_repro_adaptive_s2	77.97	69.78	81.48	514
01_fixed_e07	77.04	68.20	80.52	512
02_vswitch_memo_defaults	77.24	68.54	80.73	506
03_vswitch_strong_fast	77.08	68.28	80.51	515
04_vswitch_low_thresh	76.86	67.95	80.35	529

Appendix

Table 6: Every measured rung of the mechanism campaigns on SportsMOT val (shared cached inputs; each campaign’s rung 00 reproduces the 77.97 anchor exactly). Velocity-switch calibration: MeMoSORT defaults

\theta{=}0.1172

p{=}0.5/0.6

; “strong fast”

p_{\text{fast}}{=}1.0

; “low thresh”

\theta{=}0.05

p_{\text{fast}}{=}1.0

Run	tol	HOTA $\uparrow$	AssA $\uparrow$	IDF1 $\uparrow$	IDSW $\downarrow$
merge_g100_t06	2.0	79.16	72.04	83.60	434
merge_g100_t07	2.0	79.20	72.11	83.65	436
merge_g200_mtoff_t08	off	83.93	80.96	91.49	258
merge_g200_t06	2.0	79.14	72.00	83.59	434
merge_g200_t07	2.0	79.19	72.09	83.64	436
merge_g300_mt1000000_t07	$10^{6}$	84.03	81.23	91.83	248

Appendix

Table 7: The full offline merge grid on SportsMOT val (pooled). “tol”: motion-gate tolerance in box diagonals (2.0 = customary default; off = removed). The pooled argmax is g450/off/0.80 (84.19); we select g300/off/0.85 by best worst-sport HOTA.

Configuration	HOTA $\uparrow$	AssA $\uparrow$	IDF1 $\uparrow$	IDSW $\downarrow$	Bask.	Foot.	Voll.
	Overall (pooled)				HOTA by sport
No merge (online tracker)	77.97	69.78	81.48	514	79.28	75.41	81.59
Merge, motion gate on ( $g{=}100$ , $\tau{=}0.70$ )	79.20	72.11	83.65	436	80.55	76.67	82.38
Merge, gate off ( $g{=}300$ , $\tau{=}0.85$ )	84.12	81.37	91.72	273	85.57	83.03	83.86

Appendix

Table 8: Offline tracklet merge, per-sport breakdown on SportsMOT val (pooled TrackEval columns as in Sec. 4.5 ; per-sport HOTA is the unweighted sequence mean). Bold: selected configuration (best worst-sport HOTA).

	HOTA $\uparrow$			AssA $\uparrow$
Benchmark (val)	online	$+$ merge	$\Delta$	online	$+$ merge	$\Delta$
SportsMOT	77.97	84.12	$+$ 6.15	69.78	81.37	$+$ 11.59
DanceTrack	56.86	58.73	$+$ 1.87	41.65	44.70	$+$ 3.04

Appendix

Table 9: Merge transfer across benchmarks. The same offline merge (

\tau{=}0.85

g{=}300

, motion gate off) applied to each benchmark’s online val output, with different embedding models (sports OSNet vs. FastReID SBS-S50): appearance-only re-identification with the motion veto disabled is benchmark-general. All cells are pooled TrackEval scores, matching the merge table in the main paper.

Pipeline stage	FPS $\uparrow$
Association only (cached detections + CMC, no ReID)	204.8
Tracker + live OSNet re-identification (ReID), excl. image decode	64.8
Full pipeline (incl. JPEG decode)	50.7

Appendix

Table 10: Runtime on a single RTX 3060 Laptop GPU (SportsMOT val), all mechanisms active. Deferred commitment costs

{\approx}0.6

ms/frame, visible only in the network-free microbenchmark (base configuration 234.1 FPS); with any per-frame network the cost disappears into noise. Every stage exceeds the 25 FPS broadcast rate.

Tracker	HOTA $\uparrow$	DetA $\uparrow$	AssA $\uparrow$	MOTA $\uparrow$	IDF1 $\uparrow$	IDs $\downarrow$
ARC-SORT (ours)	77.71	86.94	69.50	95.39	81.35	552
BoT-SORT [ 4 ]	76.49	88.54	66.12	95.68	77.43	731
ByteTrack [ 47 ]	75.01	87.73	64.17	95.04	75.76	812
Deep OC-SORT [ 28 ]	74.99	88.05	63.91	95.07	75.99	787
StrongSORT [ 12 ]	73.43	87.28	61.82	94.17	73.80	1854
OC-SORT [ 9 ]	72.51	86.77	60.63	93.64	73.26	1208

Appendix

Table 11: Shared-detection study on SportsMOT val: seven SORT-family trackers run on identical cached YOLOX-X detections (raw tracker outputs, no interpolation; pooled TrackEval). With the detector held fixed, the spread is association alone.

Sequence	Sport	HOTA	AssA	IDF1	Sequence	Sport	HOTA	AssA	IDF1
v_00HRwkvvjtQ_c001	basketball	86.95	84.66	94.79	v_9MHDmAMxO5I_c006	volleyball	83.72	82.31	93.36
v_00HRwkvvjtQ_c003	basketball	91.60	91.72	99.03	v_9MHDmAMxO5I_c009	volleyball	81.92	77.01	89.73
v_00HRwkvvjtQ_c005	basketball	89.01	89.12	98.88	v_BgwzTUxJaeU_c008	basketball	83.78	78.84	93.04
v_00HRwkvvjtQ_c007	basketball	83.48	79.89	94.80	v_BgwzTUxJaeU_c012	basketball	81.58	76.65	91.76
v_00HRwkvvjtQ_c008	basketball	89.81	90.10	99.32	v_BgwzTUxJaeU_c014	basketball	84.48	79.28	92.70
v_00HRwkvvjtQ_c011	basketball	86.75	84.68	94.76	v_G-vNjfx1GGc_c004	football	83.35	79.93	89.52

Appendix

Table 12: Per-sequence results of the selected configuration (online

+

interpolation

+

merge,

\tau{=}0.85

g{=}300

, gate off) on SportsMOT val.

Multi-Object Tracking

Multi-object tracking (MOT) links detections across video frames to maintain object identities through motion, occlusion, and changing viewpoints.

See all 186 papers

Mar 6, 2026

Occlusion-Aware SORT: Observing Occlusion for Robust Multi-Object Tracking

Multi-object tracking (MOT) involves analyzing object trajectories and counting the number of objects in video sequences. However, 2D MOT faces challenges due to positional cost confusion arising from partial occlusion. To address this issue, we present the novel Occlusion-Aware SORT (OA-SORT) framework, a plug-and-play and training-free framework that includes the Occlusion-Aware Module (OAM), the Occlusion-Aware Offset (OAO), and the Bias-Aware Momentum (BAM). Specifically, OAM analyzes the occlusion status of objects, where a Gaussian Map (GM) is introduced to reduce background influence. In contrast, OAO and BAM leverage the OAM-described occlusion status to mitigate cost confusion and suppress estimation instability. Comprehensive evaluations on the DanceTrack, SportsMOT, and MOT17 datasets demonstrate the importance of occlusion handling in MOT. On the DanceTrack test set, OA-SORT achieves 63.1% and 64.2% in HOTA and IDF1, respectively. Furthermore, integrating the Occlusion-Aware framework into the four additional trackers improves HOTA and IDF1 by an average of 2.08% and 3.05%, demonstrating the reusability of the occlusion awareness.

Chunjiang Li, Jianbo Ma, Li Shen +2

College of Computer Science, Sichuan University, Chengdu, China · Institute of Optics and Electronics, CAS, Chengdu, China

Multi-Object Tracking Occlusion Reasoning

Ask PDF

Aug 16, 2026

Training-Free Long-Term Multi-Object Tracking for Sports Video Analytics

Long-term multi-object tracking in sports remains challenging due to frequent occlusions, rapid camera motion, and repeated player reappearances. We introduce McByte++, a training-free tracking-by-detection framework that integrates lightweight mask propagation, conditional camera motion compensation, and online re-identification within a unified pipeline. Compared to its predecessor, McByte++ substantially improves runtime efficiency while enhancing identity preservation. On SoccerNet-tracking and SportsMOT benchmarks, McByte++ achieves up to +3.0 HOTA and +6.1 IDF1 improvements over the original McByte in the online setting, with further gains when combined with offline global association. Replacing heavy segmentation components and optimizing motion modeling yields up to an order-of-magnitude speed increase. All results are obtained without detector retraining or dataset-specific tuning. Code will be made available at https://github.com/tstanczyk95/McBytePlusPlus.

Tomasz Stanczyk, Seongro Yoon, Francois Bremond

Inria, France · Université Côte d’Azur, France

Multi-Object Tracking Sports Video Analysis

Ask PDF

Sep 19, 2026

DOA-SORT: Directional Occlusion-Aware Multi-Object Tracking with Distributional Observations

Identity association in multi-object tracking (MOT) is vulnerable to partial occlusion, truncated detections, and fluctuating confidence scores. Existing motion-dominant trackers commonly represent occlusion as a scalar penalty. This treatment misses the directional observation bias caused by occlusion: left, right, top, and bottom occlusions distort the location and shape of a detection in different ways. We propose \ours{} (Directional Occlusion-Aware SORT), an online and training-free tracker that models these biases explicitly. First, it infers a soft front--back ordering from box overlap and relative bottom positions, and estimates directional occlusion coverage and depth. It then constructs a mixture of one clean and four directional occlusion observation components. The model uses a five-dimensional observation comprising box center, area, confidence, and aspect ratio, and adapts observation noise to predicted occlusion and detection confidence. The directional mixture likelihood is used in high-confidence association, low-confidence association, and track recovery; ambiguity penalties and local order-consistency swaps further reduce identity errors among nearby objects. On the DanceTrack validation split, \ours{} improves HOTA from 63.00 to 66.34, AssA from 45.10 to 49.57, and IDF1 from 62.19 to 65.28 over OA-SORT with the same detector and evaluation protocol. The gains are concentrated in association quality while detection accuracy remains stable. Additional local evaluations on MOT17 and MOT20 train splits characterize cross-dataset behavior under the same no-ReID tracking protocol.

Hao Wang

BDNRC

Multi-Object Tracking Data Association

Ask PDF

Abstract

Multi-Object Tracking

Explore similar work

Occlusion-Aware SORT: Observing Occlusion for Robust Multi-Object Tracking

Training-Free Long-Term Multi-Object Tracking for Sports Video Analytics

DOA-SORT: Directional Occlusion-Aware Multi-Object Tracking with Distributional Observations