Sample
Choose a point on a garment image. The extension reads the displayed pixel colour as a hexadecimal RGB value.
A transparent approach to colour
How it works.
Mindfully Fashionable samples a colour from an image on your screen, converts it into a colour model, and applies understandable rules to give it a seasonal colour label.
The aim is to make colour easier to explore while you shop—not to make the decision for you.
The process
From pixel to palette.
The extension uses a rule-based classifier. It does not use machine learning and does not determine your personal season from your appearance.
Measure
The RGB value is converted to HSL: hue, saturation and lightness. These describe colour family, intensity and relative lightness.
Classify
The classifier follows a fixed order: (1) check the Neutral gate, (2) if the colour is not Neutral, estimate warm/cool and choose Spring, Summer, Autumn or Winter, then (3) apply only that family’s sub-palette rules.
Important: this describes the colour sampled from the screen, not whether that colour will flatter a particular person. Treat the result as a starting point for your own judgement.
The colour model
Three measurements, one estimate.
Hue (H)
Measured in degrees from 0° to 360°. It represents the broad colour family, such as red, yellow, green or blue.
Saturation (S)
Measured from 0% to 100%. Lower values are more greyed or muted; higher values are more vivid.
Lightness (L)
Measured from 0% to 100%. Lower values are darker; higher values are lighter.
The extension rounds H, S and L to whole numbers before applying thresholds. HSL is convenient, but it does not perfectly model human colour perception.
Visual guide
Four spectrums, four colour cues.
Read these as four different views of the decision process: Neutral is checked first, then hue estimates temperature, while lightness and saturation help distinguish the remaining colours. The wheels are explanatory diagrams—not fixed seasonal boundaries.
How the bars work together: first the Neutral gate is checked. If the colour passes through, hue helps estimate warm or cool; lightness and saturation then contribute to choosing a main seasonal family and its sub-palette. No single measurement predicts the final label.
The classifier
Two stages: family, then sub-palette.
The algorithm first assigns one of five main palettes. Only then does it choose a sub-palette within that family. Thresholds are pragmatic rules, not universal scientific boundaries.
Choose a main palette
First, the classifier checks whether L ≤ 12, L ≥ 93, S ≤ 8, or S ≤ 24 with 60% < L < 93%. If any condition matches, the main palette is Neutral. Otherwise, it estimates warm/cool from hue (with saturation influencing some yellow-green hues), then uses lightness and saturation to choose Spring, Summer, Autumn or Winter.
Choose the sub-palette
Only the selected family’s rules are applied: Spring checks Bright, then Light, otherwise True; Summer checks Light, then Soft, otherwise True; Autumn checks Deep, then Soft, otherwise True; Winter checks Bright, then Deep, otherwise True. Neutral selects True, Warm or Cool Neutral.
Return one label
Every sampled colour receives one of the 15 advertised labels: 12 seasonal sub-palettes or True Neutral, Warm Neutral or Cool Neutral.
Five main palettes and their sub-palettes
Spring
Summer
Autumn
Winter
Neutral
Illustrative swatches only; these are visual examples, not definitive or exhaustive palette definitions.
Neutral colours are handled before seasonal colours because hue is less reliable when saturation is very low. The main-palette thresholds are mutually exclusive in their decision order, and the sub-palette rules are evaluated only after the family is selected.
Confidence score
What the percentage does—and doesn't—mean.
The web app and Chrome extension use the same separate score, calculated with hand-written rules. Very dark colours receive 93%; very light colours receive 95%; very light, low-saturation colours receive 90%; and colours with saturation below 5% receive 92%. Other colours start at 55 points, then gain or lose points according to saturation and distance from 50% lightness. The final score is limited to 50%–95%.
This is not a statistically calibrated probability. The score has not been validated against a labelled reference dataset, so a result such as “80%” does not mean there is an 80% chance that the label is correct.
Limitations
A useful guide, not an absolute verdict.
- Screen brightness, display settings, lighting in the original photograph, editing and compression can change the sampled colour.
- One pixel may not represent a whole garment, especially where there are shadows, highlights, texture or patterns.
- Colours near a threshold may receive different labels from very similar colours.
- Seasonal colour-analysis systems vary, and this classifier has not been scientifically validated.
- The extension classifies a garment colour. It does not analyse your skin, eyes or hair, or determine your personal seasonal palette.
We aim to keep the method understandable: the rules are visible, testable and can be improved using carefully selected colour examples.
Back to Mindfully Fashionable →