rmitra/PS-Dataset

PhotoSynth Dataset for improving local patch Descriptors

PS-Dataset

A Large Dataset for improving Patch Matching

Using Pre-Trained Model

Using Our Dataset

[Added Full Resolution RGB Images]

Link to paper -

https://arxiv.org/abs/1801.01466

Contributors -

Rahul Mitra, Nehal Doiphode, Utkarsh Gautam, Sanath Narayan, Shuaib Ahmed, Sharat Chandran, Arjun Jain

Sample image pairs from dataset

Image pair showing illumination variation

Image pair showing scale variation

Image pair showing viewpoint variation

Results -

Performance comparison for patch retrieval task on HPatches dataset.

Method\Noise	Easy	Hard	Tough	Mean
SIFT	64.6	37.4	22.7	41.7
DeepDesc	67.2	52.2	37.8	52.4
Tfeat	68.4	50.8	34.7	51.3
Hardnet+	79.7	68.7	52.6	66.7
Hardnet-PS	82.5	78.0	69.1	76.5

Performance comparison for image matching task on Fountain-P11 scene

Method\Baseline	Narrow	Wide	Very-wide	Mean
DeepDesc	76.3	40.8	9.2	40.0
Tfeat	86.6	62.8	21.9	57.1
Hardnet+	92.4	83.2	35.0	70.2
Hardnet-PS	92.8	85.3	47.0	75.0

Performance comparison for image matching task on Herzjesu-P8 scene

Method\Baseline	Narrow	Wide	Mean
DeepDesc	64.4	13.1	35.1
Tfeat	76.6	27.4	48.5
Hardnet+	85.1	44.5	61.9
Hardnet-PS	85.1	50.6	65.4

Scene statistics-

Statistics and parameters for each scene in the training set

Scene Id	No. images	No. points	No. Patches	MIN_V_TH	MAX_V_TH	No. Pairs
11	155	5724	31174	10°	50°	59098
13	436	143150	788401	5°	45°	1447833
14	315	115183	728302	10°	50°	1257142
15	239	69725	395021	10°	50°	733752
16	214	14634	94585	10°	50°	161737
20	381	41524	329980	10°	50°	702957
24	274	121992	788657	10°	50°	1713912
30	345	150087	760203	10°	50°	603327
34	142	7353	32198	10°	50°	52460
36	278	68656	395547	2°	45°	361799
41	268	29308	181515	10°	50°	372813
49	219	69301	346042	10°	70°	325492
50	302	61846	235538	10°	50°	294323
51	251	97293	620192	10°	50°	692386
53	269	80025	532245	10°	50°	1399146
65	153	86016	227226	4°	72°	202294
66	232	142459	426105	10°	72°	411762
67	98	37412	279194	3°	50°	1372920
71	309	65116	457196	10°	50°	1111475
74	207	51107	320927	10°	50°	662981
76	254	93127	535378	10°	50°	1172380
89	214	63498	641700	15°	50°	1503677
90	388	92334	731818	15°	50°	811130
91	182	78961	750676	12°	50°	1664135
95	201	21677	106001	10°	72°	110641

Statistics and parameters for each scene in the validation set

Scene Id	No. images	No. points	No. Patches	MIN_V_TH	MAX_V_TH	No. Pairs
4	264	107549	602331	10°	50°	965990
23	249	66492	328555	10°	50°	576180
31	262	25451	164984	10°	50°	147734
44	87	57944	157953	3°	72°	98134
88	99	4850	54781	5°	50°	235524