The dataset consists in 100 source BRDFs (from the MERL-MIT BRDF database), subject to approximations with different models, producing a total of 2026 BRDFs (including references).
The dataset is provided in two formats: the standard MERL binary format and our own TITOPO format. 

By default the MERL format assumes reciprocity in BRDFs encoding, thus making the economy of half the data. Some analytical models however do not come with reciprocity (e.g Phong), which required to extend MERL encoding in order to provide all data in a common format. As a consequence, the number of directions for phi is thus 360 instead of 180.
In our dataset, those non-reciprocal BRDFs have a size of 68MB, while standard reciprocal ones are 34MB. 

TITOPO is a format in which data is sampled according to incoming azimuthal and outgoing azimuthal and orient angles theta_i, theta_o and phi_o. 

You can find below the correspondance between Letters in file names and approximation models:

"DUPUY":'T'
"GGX":'X'
"BECKMAN":'E'
"SGD":'D'
"LAFORTUNE-Ngan":'L'
"LAFORTUNE-Ours":'N'
"WARD-Ngan":'Z'
"WARD-Ours"':'W'
"BLINN-PHONG-Ngan"':'P'
"BLINN-PHONG-Ours":'O'
"ABC-Authors"':'A'
"ABC-Ours":'F'
"CHEBYCHEV":'C'
"LEGENDRE":'G'
"BAGHER":'B'
"MMS.1" :'1'
"MMS.2" :'2'
"MMS.3" :'3'
"MMS.4" :'4'
"MMS.5" :'5'
"MMS.6" :'6'
}

For instance: "alum-bronze.E.binary", is the BECKMAN analytical approximation of alum-bronze.