Shot Transition Detection on Synthetic Data benchmark leaderboard · SOTA2 Research

Shot Transition Detection on Synthetic Data

92.4Segment PrecisionPySceneDetect (Adaptive)

Evaluation Results

Method	Links
PySceneDetect (Adaptive)2026.04		92.4	17.2	29	92.2	3.6	6.9	0.28	0.09
PySceneDetect (Content)2026.04		90.9	12.8	22.5	96.8	2.9	5.6	0.61	0.09
TransVLM2026.04		90.8	88.2	89.5	94.6	93	93.8	0.11	0.5
Qwen2-VL Thinking 32B2026.04		90	60.8	72.6	94.3	32.3	47.9	1.16	3.3
Qwen2-VL Instruct 32B2026.04		89.5	62.3	73.5	91.1	36.1	51.6	1.35	0.98
Qwen2-VL Thinking 30B-A3B(MoE)2026.04		89	61	72.4	91.5	33.1	48.5	1.18	0.92
PySceneDetect (Hash)2026.04		88	11.8	20.8	94	2.7	5.2	0.85	0.09
Qwen2-VL Thinking 8B2026.04		85.4	38.9	53.4	92.3	14.7	25.2	1.29	1.31
Qwen2-VL Thinking 4B2026.04		83.9	21.8	34.6	74.8	8.7	15.6	1.68	1.03
Qwen2-VL Instruct 30B-A3B(MoE)2026.04		80.6	59.3	68.3	74.9	35.5	48.2	1.86	0.35
PySceneDetect (Threshold)2026.04		77.3	3.9	7.4	74.9	0.8	1.6	1.46	0.09
Qwen2-VL Instruct 4B2026.04		71.7	30.6	42.8	45.8	61.8	52.5	8.88	0.31
Qwen2-VL Instruct 8B2026.04		58.6	59.7	59.1	74.1	20.4	31.5	1.6	0.34
PySceneDetect (Hist)2026.04		57.7	37.9	45.5	76.2	11.7	19.7	1.04	0.09
Gemini 3 Pro2026.04		47.9	76.8	58.8	71.1	48.2	56.8	0.88	1.32
AutoShot2026.04		37.9	24.8	29.9	53	5.8	10.2	0.51	0.03
Gemini 2.5 Pro2026.04		33.8	85.1	46.5	63.8	76	68.6	0.82	0.81
TransNetV22026.04		27.5	14.9	19.4	41.7	3.4	6.3	0.61	0.07