wzyn20051216/matlab-agent-skills / skills/matlab-signal-vision-ai
matlab-signal-vision-ai
UnverifiedMATLAB workflow for signal processing, DSP, image processing, computer vision, lidar, medical imaging, deep learning, and AI-assisted reproducible experiments. Use for filters, spectrograms, wavelets, object detection, segmentation, classification, training, inference, generated figures, and for verifying signal or vision results — units, windows, Nyquist, filter response, BER/SNR, masks and boxes, IoU/precision/recall/F1, label alignment, transform invertibility, and split leakage.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 10.5 KB |
| references/signal-measurement-contract.md | markdown | 10.3 KB |
| references/vision-verification-contract.md | markdown | 13.5 KB |
| scripts/mas_ber_check.m | other | 12.6 KB |
| scripts/mas_detect_check.m | other | 21.5 KB |
| scripts/mas_filter_check.m | other | 12.7 KB |
| scripts/mas_image_contract.m | other | 31.4 KB |
| scripts/mas_image_metrics.m | other | 4.5 KB |
| scripts/mas_measure_contract.m | other | 19.8 KB |
| scripts/mas_ofdm_ber.m | other | 6.8 KB |
| scripts/mas_sample_check.m | other | 9.6 KB |
| scripts/mas_seg_check.m | other | 23.0 KB |
| scripts/mas_spectrum_check.m | other | 7.8 KB |
| scripts/mas_split_seed.m | other | 3.6 KB |
| scripts/mas_stat_verdict.m | other | 10.0 KB |
| tests/test_signal_contract.m | other | 21.5 KB |
| tests/test_signal_vision.m | other | 15.8 KB |
| tests/test_stat_decision.m | other | 8.6 KB |
| tests/test_vision_contract.m | other | 35.3 KB |
scripts/mas_image_metrics.m
View on GitHubfunction r = mas_image_metrics(pred, truth, opts)
%MAS_IMAGE_METRICS Overlap metrics for a binary segmentation, with the conventions stated.
%
% r = mas_image_metrics(predMask, trueMask)
% r.iou, r.dice, r.precision, r.recall, r.f1, r.accuracy
% r.tp, r.fp, r.fn, r.tn
% r.perObject per predicted blob: area, best-matching truth label, IoU
% r.objectIouMean mean per-object IoU (NaN if bwlabel unavailable)
% r.objectIouWorst worst per-object IoU
% r.detail
%
% CONVENTIONS THAT ARE EASY TO GET WRONG
% - Both inputs are LOGICAL. A uint8 image with 0/255 is converted with ~= 0, not by
% comparing to 1; treating 255 as "false" is a silent 100 % error.
% - The result does NOT depend on orientation, but every reported AREA is in pixels
% and every centroid is [COLUMN ROW] (regionprops convention), which is stated
% here once instead of being re-guessed in each task.
% - An empty prediction and an empty truth is a perfect match (IoU = 1), not 0/0; a
% metric that returns NaN there breaks every averaged score downstream.
%
% These are aggregate pixel metrics. They do NOT measure whether the OBJECTS are
% right: a prediction that covers the correct total area in the wrong place scores
% well on pixel recall and badly on every object-level metric. `r.perObject` reports
% the per-object IoU when both masks are labelled, so the difference is visible.
if nargin < 3, opts = struct(); end
opts = dflt(opts, struct('threshold', 0));
p = to_logical(pred, opts.threshold);
if ~isempty(truth), t = to_logical(truth, opts.threshold); else, t = false(size(p)); end
if ~isequal(size(p), size(t))
error('mas_image_metrics:size', 'pred %s and truth %s differ in size', ...
mat2str(size(p)), mat2str(size(t)));
end
tp = nnz(p & t); fp = nnz(p & ~t); fn = nnz(~p & t); tn = nnz(~p & ~t);
if tp + fp + fn == 0
% both masks empty: a perfect match by definition
iou = 1; dice = 1; prec = 1; rec = 1;
else
iou = tp / max(tp + fp + fn, 1);
dice = 2*tp / max(2*tp + fp + fn, 1);
prec = tp / max(tp + fp, 1);
rec = tp / max(tp + fn, 1);
end
if prec + rec == 0, f1 = 0; else, f1 = 2*prec*rec/(prec + rec); end
r = struct();
po = per_object(p, t);
r.tp = tp; r.fp = fp; r.fn = fn; r.tn = tn;
r.iou = iou; r.dice = dice; r.precision = prec; r.recall = rec; r.f1 = f1;
r.accuracy = (tp + tn) / max(numel(p), 1);
r.predArea = nnz(p); r.trueArea = nnz(t);
r.perObject = po;
if ~isempty(po) && all(isfinite([po.iou]))
r.objectIouMean = mean([po.iou]);
r.objectIouWorst = min([po.iou]);
else
r.objectIouMean = NaN; r.objectIouWorst = NaN;
end
r.detail = sprintf('IoU %.4f, Dice %.4f, P %.4f, R %.4f (%d/%d/%d px)', ...
iou, dice, prec, rec, tp, fp, fn);
end
% ---------------------------------------------------------------------------
function po = per_object(p, t)
% Per-object IoU, each predicted blob matched to the truth blob it overlaps most.
%
% The matching is by argmax over ALL truth labels, not by index. bwlabel numbers
% blobs in raster (column-major) scan order, so "object 1" is simply the first one
% the scan happens to reach -- assuming Lp==k corresponds to Lt==k is wrong as soon
% as the two masks disagree, which is exactly the case this function exists for.
%
% Requires Image Processing Toolbox (bwlabel). Without it the caller still gets
% every pixel metric and an empty perObject; segmentation quality assessment is
% not silently degraded, it is reported as unavailable.
po = struct('label', {}, 'area', {}, 'truthLabel', {}, 'iou', {});
if exist('bwlabel', 'file') ~= 2, return; end
if isempty(p) || isempty(t), return; end
Lp = bwlabel(p, 4);
Lt = bwlabel(t, 4);
np = max(Lp(:));
nt = max(Lt(:));
if np == 0 || nt == 0, return; end
for i = 1:np
mask = (Lp == i);
best = 0; bestj = 0;
for j = 1:nt
tm = (Lt == j);
un = nnz(mask | tm);
if un == 0, continue; end
v = nnz(mask & tm) / un;
if v > best, best = v; bestj = j; end
end
po(end+1) = struct('label', i, 'area', nnz(mask), 'truthLabel', bestj, 'iou', best); %#ok<AGROW>
end
end
% ---------------------------------------------------------------------------
function m = to_logical(x, thr)
if islogical(x)
m = x;
elseif isnumeric(x)
if isempty(thr) || thr == 0
m = x ~= 0; % 0/255 images included
else
m = x > thr;
end
else
error('mas_image_metrics:type', 'mask must be logical or numeric, got %s', class(x));
end
end
function o = dflt(o, d)
f = fieldnames(d);
for k = 1:numel(f)
if ~isfield(o, f{k}), o.(f{k}) = d.(f{k}); end
end
end
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| planning-and-task-breakdownaddyosmani/agent-skillsBreaks work into ordered tasks. Use when you have a spec or clear requirements and need to break work into implementable tasks. Use when a task feels too large to start, when you need to estimate scope, or when parallel work is possible. | Stars: 119k | Updated | Claude CodeCodex | Verified | Install |
| screenpipethedotmack/claude-memSearch screen and audio history recorded by Screenpipe and optionally save selected context into claude-mem. Use when the user asks to recall desktop activity, a meeting, or something seen in another app through Screenpipe. | Stars: 118k | Updated | Claude CodeCodex | Verified | Install |
| contact-widgetnexu-io/open-designSelf-contained floating chat widget with welcome screen, social links, meeting button, and message input. Single HTML file, zero dependencies. | Stars: 108k | Updated | Claude CodeCodex | Verified | Install |
| meeting-notesnexu-io/open-designMeeting notes page — title bar with attendees, agenda checklist, decisions block, action items table with owners + dates, and a "next meeting" footer. Use when the brief mentions "meeting notes", "minutes", "1:1 notes", "all-hands recap", or "会议纪要". | Stars: 108k | Updated | Claude CodeCodex | Verified | Install |