Datasets:

simonko912
/

dwd-hf-classify-1

Curently super small, recieving tooks a lot of time at 50 baud (50bps) It was made by using a known list of types and qualities and training a small ml to help together with some rules hardcoded to classify the type and quality. If there are issues please report them and i will try improoving the ml and rules.

TYPE

= garbage / broken line
= metadata header (PN, EDZW, FQEN, WODL, NIL, NNNN station codes)
= region header (KATTEGAT, BALTIC SEA, coordinates + SST lines)
= forecast entry (SU 26. 12Z: SW 4 0.5 M)
= raw weather parameters line fragment
= corrupted-but-structured (important: salvageable noise)
= hf call / cq (CQ CQ CQ DE DDK2 DDH7)
= frequency list / schedule (FREQUENCIES 4583 KHZ 7646 KHZ 10100.8 KHZ)
= test pattern / idle (RYRYRYRYRYRYRYRYRY)
= bulletin board (WEATHER AND SEA BULLETIN FOR THE MEDITERRANEAN SEA, ISSUED BY MARINE WEATHER SERVICE HAMBURG 25.04.2026, 15 UTC:)
= synop (06060 05973 00000 10037 20021 30178 40245 52015 60002)

QUALITY (DECIMAL)

= unusable / noise
= partially readable
= mostly correct with minor corruption
= fully correct and model thinks is acurate