Peak Detector
Paste a column of numbers and get every local maximum and minimum in it — each peak's position, value, prominence and width, in a table you can export. Filter out noise by prominence, height, spacing or width, or smooth the signal first. Nothing is uploaded; detection runs in your browser.
About this tool
Peak Detector finds local maxima and local minima in a one-dimensional signal. Paste a row or column of numbers and it reports each peak's 0-based index, value, prominence, width, bases, and neighbouring-drop measurements. It is useful for quick checks of time-series data, sensor traces, chromatography-like curves, benchmark samples, traffic series, or any numeric sequence where you need to identify events rather than sort or smooth the data by hand.
The defaults intentionally behave like a simple local-extrema finder: a maximum is higher than both neighbours, a minimum is lower than both neighbours, the first and last samples are not considered peaks, and equal flat tops collapse to the middle sample. Add filters when real data is noisy: threshold checks the immediate neighbour drop, min_prominence checks whether the peak stands out from the wider curve, min_distance keeps one peak per event cluster, min_width rejects needle spikes, and smooth applies a centred moving average before detection.
Worked example
Input:
0, 1, 0, 3, 0, 2, 0
With mode=maxima, the report finds three peaks at indices 1, 3, and 5. The highest and most prominent is index 3 with value 3.
kind,index,value,prominence,width,left_base,right_base
maximum,1,1,1,...
maximum,3,3,3,...
maximum,5,2,2,...
Use format=csv when you want to paste the peak table into a spreadsheet, and format=json when another script or agent should consume the full measurements.
Limits and edge cases
- Accepts up to 20,000 samples per run.
- Input numbers may be integers, decimals, negatives, or scientific notation such as
1e-3. - Non-numeric tokens such as a pasted header are skipped and counted.
smoothmust be odd (or 0/1 for off) so the moving-average window stays centred.rel_heightmust be between 0 and 1; the default0.5measures width at half prominence.max_peaks=0means report every matching peak; higher values cap the sorted result list.
FAQ
What is the difference between threshold and prominence?
threshold is local: it requires the peak to rise above the two samples touching it by at least the requested amount. min_prominence is broader: it looks outward until the signal climbs back above the peak or reaches an edge, then measures how far the peak stands above its surrounding bases. Prominence is usually better for noisy real signals.
Can this find valleys as well as peaks?
Yes. Set mode=minima to find local minima only, or mode=both to return maxima and minima in one report. For minima, prominence and width are measured on the inverted signal, so the same filters work for valleys.
How should I choose min_distance?
Use min_distance when one physical event creates several nearby candidate peaks. The detector keeps the most extreme candidate first, then drops other peaks of the same kind that are closer than the requested number of samples.
Does smoothing change the reported values?
Yes. Smoothing is applied before detection, and reported peak values come from the smoothed signal. Keep smooth=0 when exact original sample heights matter; use an odd window such as 3, 5, or 7 when noise would otherwise create many tiny peaks.
Developer & Automation Access
Run it from the terminal
Same engine as this page, headless — via the gizza CLI:
gizza tool peak-detector "0, 1, 0, 3, 0, 2, 0"New to the CLI? Get gizza →
Open it by URL
Pre-fill and auto-run this tool with query parameters — the names match the API/CLI:
https://gizza.ai/tools/peak-detector/?data=0%2C%201%2C%200%2C%203%2C%200%2C%202%2C%200&mode=maxima&separator=auto&smooth=0&min_value=e.g.%205&max_value=e.g.%20100&threshold=0&min_distance=0&min_prominence=0&min_width=0&rel_height=0.5&max_peaks=0&sort_by=position&format=textMachine-readable descriptor: tool.json — title + parameters JSON Schema for agents.