CCV - how to convert trained models to cascading filter?

Question

Parsing

Maybe someone has worked with liuliu's DPM implementation and knows how to convert the training output model into a cascaded filter for further use?
Or maybe just someone experienced will figure out what these parameters can mean and how to make a filter out of them?

,
,
1 1
14 6
0x1.73434ap+3 0x0p+0 0x0p+0 0x0p+0
-0x1.747d78p-8 -0x1.9fe5c4p-8 -0x1.1f86eap-8 -0x1.25e858p-8 -0x1.a31548p-6 -0x1.df13d8p-8 -0x1.30595ap-8 -0x1.4fd746p-8 -0x1.427348p-7 -0x1.e30cc2p-8 0x1.373952p-10 -0x1.13ad6ep-8 -0x1.adb92p-8 -0x1.0243a6p-6 -0x1.2c4bacp-8 -0x1.343c6cp-8 -0x1.e50da2p-9 -0x1.29437ep-7 -0x1.0f2d08p-7 -0x1.2afe88p-9 -0x1.68522ap-8 -0x1.61176ap-8 -0x1.309ac8p-6 -0x1.d7605cp-9 -0x1.1aca48p-9 -0x1.17e11cp-9 -0x1.6549dcp-9 0x1.14b752p-12 0x1.63cc02p-9 -0x1.24f0dap-10 -0x1.0f6046p-9 -0x1.d163ep-9 -0x1.58951ap-8 -0x1.8929bap-9 -0x1.772b2cp-9 -0x1.07102p-6 0x1.881be2p-10 -0x1.70f2a8p-12 -0x1.14bc4cp-9 -0x1.ad28dep-7 -0x1.75b77ep-7 -0x1.737116p-9 -0x1.ea13c2p-9 -0x1.a35c7p-10 -0x1.ab5744p-7 -0x1.8d846p-9 -0x1.e2e984p-9 -0x1.25761cp-8 -0x1.7debdep-7 -0x1.42bf2cp-7 -0x1.6d84d2p-8 -0x1.42cd58p-8 -0x1.8dd4a8p-11 -0x1.49871ap-7 0x1.8a2ddep-8 0x1.0fc28p-9 0x1.01112cp-9 -0x1.c8018ep-9 -0x1.7f3554p-10 0x1.caef8ep-10 -0x1.699392p-10 -0x1.5b67f2p-11 -0x1.f86704p-12 -0x1.12d7c4p-9 -0x1.ea8632p-11 -0x1.8c0dcap-9 -0x1.7ce606p-7 0x1.fc55a8p-11 0x1.841aaep-9 0x1.78630ep-9
....

Completely with different data, but the general view of the filter looks something like this:

2000 20 60
3 -0x1.b64f7ep-1 0x1.b64f7ep-1 -0x1.b64f7ep-1
3 -0x1.e2a43p+4
2 0x1.1f8c28p-7
13 44 18 56
2 0x1.673b7ep-7
6 48 15 55
3 0x1.5de05cp-7
11 3 14 9
2 0x0p+0
2 -0x1.0e37acp-4
1 54 2 55
4 -0x1.e7e628p-4