new functiont in Labels.jl

[?]

AEj8dahVWy718uSSFPe9VSRJ5qX5G8pC2zvFzJJ8yzBd

Nov 17, 2024, 8:27 PM

5M7JW5OVBMQ2EX5QDNJ2FRQPB3DIRG7FDW7DMXXTI2NDRTTSEJZAC

Dependencies

[2] 7KO4BAOG Tidy up
[3] ACUJ2OKI label summary now workt on cwd unless folder specified, note trailing / required if folder is specified
[4] 2UBDFCJH new files tracked
[5] QA2TJZRA ripped ML out into the SkraakML repo, compiles quicker now
[6] 2O6SHIVY added avianz_of_raven function to Labels.jl
[7] YXAKJSDT added check_change_avianz_species! function to Labels.jl
[8] YODTMMPT added a whole bunch of functions to Labels.jl for working with raven and avians labels
[9] Y2LW6LAA deduplicated an export in Labels.jl

Change contents

replacement in src/Labels.jl at line 7

[2.1095]→[2.1095:1116](∅→∅)

avianz_of_raven,

[2.1095]

[2.1116]

avianz_of_raven, #not working right, 1 file per label I think
edit in src/Labels.jl at line 12

[2.1217]

[2.1217]

one_hot_labels,
replacement in src/Labels.jl at line 179

[4.14]→[4.14:53](∅→∅)

# Raven selections.txt to AviaNZ .data

[4.14]

[4.53]

# ASSUMES 60 second files, also wav and txt file in same folder
# Raven selections.txt to AviaNZ .data
replacement in src/Labels.jl at line 186

[4.235]→[4.342:378](∅→∅)

function avianz_of_raven(f::String)

[4.235]

[4.1813]

function avianz_of_raven(f::String) #not working right, check
edit in src/Labels.jl at line 214

[4.1888]

[4.742]

duration = Float64[],
edit in src/Labels.jl at line 221

[4.893]

[4.893]

dur = data[1]["Duration"]
edit in src/Labels.jl at line 230

[4.1151]

[4.1151]

duration = dur,
replacement in src/Labels.jl at line 306

[4.2366]→[3.7:112](∅→∅)

# if specified folder must include trailing /
function label_summary(folder::String = "", avianz = true)

[4.2366]

[4.2420]

# if specified folder must include trailing /, can be "" for current folder
function label_summary(folder::String, avianz::Bool = true)
edit in src/Labels.jl at line 311

[4.2057]

[4.2533]

duration = Float64[],
replacement in src/Labels.jl at line 325

[4.3014]→[4.3014:3094](∅→∅)

select!(df, [:File, :start_time, :end_time, :low_f, :high_f, :Species])

[4.3014]

[3.187]

select!(df, [:File, :duration, :start_time, :end_time, :low_f, :high_f, :Species])
edit in src/Labels.jl at line 360

[4.4280]

[4.4280]

end

#allow 0.2 overlap each side
#round end_time down to nearest 5
function et(end_time::Float64)
end_time % 5 > 0.2 ? (c0 = ceil(end_time / 5) * 5) : c0 = floor(end_time / 5) * 5
return c0
end
#allow 0.2 overlap each side
#round start_time down to nearest 5
function st(start_time::Float64)
start_time % 5 < 4.8 ? (f = floor(start_time / 5) * 5) : f = ceil(start_time / 5) * 5
return f
edit in src/Labels.jl at line 374

[4.4284]

#labels must be a df loaded from label_summary run over avianz data (not raven)
function one_hot_labels(labels::DataFrame)
gdf = groupby(labels, :File)
vdf = []
for group in gdf
dur = first(group.duration)
nrows = dur ÷ 5
seil = nrows * 5
df = DataFrame(
file = [first(group.File) for x = 1:nrows],
start_time = collect(0:5:seil-1),
end_time = collect(5:5:seil),
)
for row in eachrow(group)
fst = st(row.start_time)
@assert fst >= 0
lst0 = et(row.end_time)
#end time must not be greater than duration
lst0 > dur ? lst = lst0 - 5 : lst = lst0
@assert lst <= dur
f_idx = fst ÷ 5 + 1 |> Int
l_idx = lst ÷ 5 |> Int
vect = [false for x = 1:nrows]
for idx = f_idx:l_idx
vect[idx] = true
end
col_name = row.Species
df[!, Symbol(col_name)] = vect
end
#(names(df) |> length) > 4 && println(df)
push!(vdf, df)
end
cdf1 = reduce(
(x, y) ->
outerjoin(x, y, matchmissing = :equal, on = intersect(names(x), names(y))),
vdf,
)
cdf2 = coalesce.(cdf1, false)
end