When photo recognition is wrong, correct the food name and preparation first, then the portion, and only then the calories. The name determines which food the system uses for its calculation; the portion determines how much. If both are wrong, changing only the calorie number fixes the result without fixing its cause.
In nosh, you can edit the name, calories, portion, and log time after recognition. You do not need to rebuild all four fields. Identify the kind of error first, and you can usually correct the part that matters most in seconds.
Wrong food name: correct the starting point
Poached chicken identified as fried chicken, unsweetened yogurt as ice cream, or tomato-beef stew as tomato soup are errors in “What is this?” Correct the name first and clarify the preparation when necessary.
The name is not merely a label for the log. It is the starting point for matching nutrition data. With the same chicken, skin-on fried versus steamed may matter more than one extra bite. If the starting point is wrong, adjusting grams is unlikely to rescue the estimate.
Prioritize words that meaningfully change the food: fried or steamed, sweetened or unsweetened, whole or skim milk, sauce included or on the side. Do not force a highly personal recipe into a standard name the database cannot truly know.
Right general food, implausible amount: correct what you ate
Deep bowls, stacked sandwiches, and covered rice make visual volume difficult. A more common difference is that the photo shows what was served while the log should reflect what you ate.
If half the noodle soup remained, you ate two-thirds of the rice, or you took only a few bites from a shared dish, change the portion to a reasonable approximation of intake. Exact grams are not required. “Half a serving,” “one small bowl,” or “one-third left” can move the entry from clearly wrong to useful.
Do not pretend to know whether a restaurant served exactly 183 or 207 grams of rice because the interface displays a number. Missing information does not become true when made more specific.
When should you edit calories directly?
Edit calories directly when you have information more reliable than the photo. Typical examples are a matching nutrition label on yogurt, a drink, or cookies, or restaurant nutrition data for the exact serving. The label is closer to the product than a visual guess.
If you cooked the meal and know the main ingredients and amounts, use the actual recipe to correct the result. There is no need to recalculate repeatedly over a tiny calorie difference.
Without a label or recipe, do not replace the estimate merely because it feels too high or you wish it were lower. An imperfect but honest estimate is more useful than a precise number created for reassurance.
With hidden oil and sauce, it is okay not to know
A single photo often cannot show how much sesame paste went into noodles, how much oil a stir-fry absorbed, or how much sugar is in takeout sauce. If you know you used half the sauce, reflect that. If you know nothing, do not disguise uncertainty as certainty.
Separate three things: what the system identified incorrectly, what you know the answer to, and what no one can confirm from the photo. Correct the first two; preserve the boundary around the third.
The quickest correction order
Review a nosh result in this order:
- Are the main foods and cooking methods correct?
- Is the amount you actually ate roughly right?
- Are drinks, sauces, or sides missing?
- If you have a label or real recipe, correct the calories.
You do not need all four steps for every meal. Correct the first when fried chicken is labeled as poached, the second when half the rice remains, and use the label directly for packaged milk. Do not create a detailed editing session when no large error exists.
The food name begins the estimation chain; calories appear at the end. In nosh, fixing the starting point, then the portion, and only then the outcome is usually faster and closer to reality than repeatedly adjusting one number.