notesum.ai

Published at December 5

Feature Coding in the Era of Large Models: Dataset, Test Conditions, and Benchmark

cs.MM

Released Date: December 5, 2024

Authors: Changsheng Gao¹, Yifan Ma², Qiaoxi Chen², Yenan Xu², Dong Liu², Weisi Lin¹

Aff.: ¹Nanyang Technological University; ²University of Science and Technology of China

Arxiv: http://arxiv.org/pdf/2412.04307v1

Refer to caption

Task	Image Classification			Semantic Segmentation			Depth Estimation			Common Sense Reasoning			Text-to-Image Synthesis
Metric	BPFE	Accuracy	MSE	BPFE	mIoU	MSE	BPFE	RMSE	MSE	BPFE	Accuracy	MSE	BPFE	CLIP Score	MSE
Quantization	10	100	2.8974	10	79.93	1.8135	10	0.4941	0.5737	10	100	0.0042	10	30.07	0.00001
VTM Baseline	1.90	100	3.0117	1.78	79.68	1.9279	1.67	0.5341	0.6103	2.69	99	0.0114	1.30	30.04	0.0029
	0.98	99	3.2324	0.90	78.91	2.1438	0.79	0.6925	0.6763	1.83	100	0.0267	0.65	29.95	0.0078
	0.21	81	3.7424	0.24	73.53	2.6032	0.24	1.0684	0.8033	0.88	98	0.0826	0.26	29.50	0.0171
	0.04	18	4.0751	0.04	55.05	3.0304	0.05	1.4053	0.9302	0.16	81	0.1900	0.10	27.43	0.0290
	0.01	2	4.4588	0.01	32.64	3.3641	0.01	1.5542	1.0313	0.04	26	0.2483	0.04	24.42	0.0436
Hyperprior Baseline	1.99	92	3.5279	1.72	78.44	2.2636	1.51	0.5691	0.6588	5.50	95	0.0674	1.40	30.20	0.0054
	1.11	91	3.7724	1.30	77.85	2.3423	1.01	0.6410	0.6830	2.84	89	0.0796	0.62	29.55	0.0122
	0.89	86	3.9035	0.54	76.39	2.6461	0.43	0.9442	0.7818	1.46	83	0.1474	0.26	28.21	0.0216
	0.37	29	4.3309	0.12	62.69	3.0935	0.08	1.1809	1.0830	1.30	67	0.1337	0.14	26.68	0.0294
	0.23	11	4.8063	0.03	40.92	3.6286	0.01	2.5775	1.1876	1.19	32	0.1582	0.07	24.27	0.0404