Lig Vision is a specialist model. It was built specifically for hairstyle recognition, trained on professional hairstyle data, and benchmarked against the most capable general-purpose vision models available today. But it is not infallible, and we are not going to claim it is.
Hairstyles with similar visual characteristics can be misclassified. A model that performs strongly overall will still encounter styles at the boundary of categories where the difference is genuinely difficult to read from a photograph alone. This is true of any vision model, regardless of how well trained. What matters as much as the rate of misclassification is what the system does when it happens.
Ligax is designed with two checkpoints that together ensure misclassifications rarely translate into a problematic appointment.
Why Most Misclassifications Do Not Matter Financially
The styles most likely to be confused with one another are styles that share the most characteristics. In the protective category, box braids and knotless braids share length, silhouette, and extension material. Beyond protective styles, a silk press and a blowout both produce straight, smooth results that can be difficult to tell apart in a photograph. A defined wash and go and a freshly stretched twist out can share the same volume and curl profile. Colour techniques such as balayage and ombre occupy overlapping visual territory when photographed at similar angles and lighting.
In each of these cases, the styles most likely to be confused are also the styles most likely to be priced similarly at most salons. That is not coincidence. Hairstyle pricing follows the same logic the model uses to classify: technique, installation complexity, and time. When Lig Vision identifies one style for a closely related one, the matched service and its price are typically still correct, even if the label is not perfectly precise. The booking goes through accurately, the deposit reflects the right amount, and the stylist has the information they need.
The Customer Sees the Result Before Confirming
After a client uploads a reference photo through your booking page, Lig Vision classifies the style and matches it to your service catalogue. Before any booking is confirmed, the client sees the detected style and the matched service price. They are not committing to anything without knowing what the system identified.
If something looks wrong to them, they can say so in notes or describe what they actually want through the chat conversation. That information travels with the booking.
The client cannot edit or override the AI's classification directly. This is deliberate. The classification maps to a specific service in your catalogue, and allowing a client to change a style label without understanding which of your services it corresponds to could create a mismatch between what your salon expects and what the client actually wants. Instead, everything the client provides during the conversation, their original image, their notes, their chat messages, all arrives in your dashboard alongside the booking.
Your Salon Is the Final Checkpoint
Every booking that comes through Ligax lands in your dashboard with the full picture: the client's original reference image, the AI's classification result, the matched service, the conversation summary, and a pre-appointment brief. Your team can review the photograph before the client arrives.
If anything does not look right, if the style in the image does not match what the AI identified, or if you want to adjust the service, the price, or the appointment duration, you can make those changes directly from your dashboard. When you save them, an email is sent to the client automatically. They are informed before they come in. There is no awkward conversation at the chair, no unexpected difference between what was booked and what was quoted.
The salon's ability to review and correct is the mechanism that closes the gap. You have the image. You have the AI result. You have the conversation. And you have the ability to act on all of it before the appointment takes place.
When the System Cannot Identify the Image at All
There is a category of outcome that is not a misclassification but a non-classification: images where the system cannot reach a minimum confidence level for any hairstyle. Lig Vision does not guess in these cases. When a confident identification cannot be made, the system returns an error rather than passing a low-confidence result into the booking flow. The client is told the image could not be identified and asked to describe what they want instead.
This means every classification result that reaches your dashboard has cleared a confidence threshold. The system does not present uncertain results as certain ones.
What This Means for Your Salon
In practice, the combination of these layers, price alignment between similar styles, client visibility before confirming, and your ability to review and modify before the appointment, means that misclassifications rarely result in a difficult situation. The cases that could matter are caught before the client sits in the chair, either because the client noticed during booking or because your team caught it in the dashboard.
We publish benchmark results so you can evaluate Lig Vision's accuracy independently rather than taking our word for it. That transparency is deliberate. You should know what accuracy level you are working with before you rely on it in your salon. The benchmark is available at github.com/Ligax-LTD/Lig_benchmark_test, and you can test the model directly at ligaxai.com/playground.