FitDexFood photo calorie calculator
There is no upload box here — the model that looks at your plate runs inside the FitDex app and stays there. The question this page answers is the less obvious but more important one: when you log a meal without weighing it, what does the small remaining gap add up to over weeks? Nobody cares about 130 calories in a meal; the same gap over three months is a different matter. Everything below comes from the numbers you enter, not from a ratio we made up.
What the missed share adds up to
12 weeks · total 33,600 kcal
- Missed per day400 kcal
- What you actually eat instead of your target2,400 kcal
- Time for that gap to reach 1 kg19 days
- Target you would have to log for the same result1,600 kcal
- Share of the day logged by eye100%
In FitDex you take the photo and the model makes the estimate; you can correct the portion right where you see it. The difference this page describes is exactly what that correction accumulates.
What a photo can see, and what it cannot
The hard part of reading calories off a photo is not recognising the dish — that part is largely solved by now. The hard part is working out how much of it there is and what went into it. Below are the things a model genuinely cannot see, and the remedy is the same for all of them: you correcting the estimate it shows.
- Cooking oil is not in the picture. The same plate of vegetables carries roughly 120 calories more with a tablespoon of olive oil, and nothing about it looks different in the photo. Logging the oil as a separate entry closes the largest share with a single habit.
- Depth disappears with perspective. A soup bowl shot from directly above looks almost identical to one holding twice as much. Taking the photo at a slight angle, with a familiar object (a fork, a spoon) visible beside the plate, cuts that uncertainty noticeably.
- Sauce and sugar hide under the plate. The dressing beneath the salad, the glaze on the chicken, the syrup in the coffee — each carries a serious part of the calories and none of them is visible from above. Shooting before it is stirred in, or logging the sauce separately, works.
- In a mixed dish the ratios are unknown. An image cannot say how much mince is in a bowl of pasta sauce; the model assumes a typical recipe here. For food you cook yourself, saving the recipe once turns the estimate into hard data every time that dish comes back.
- Packaged products need no estimating at all. For anything with a label, scanning the barcode or reading the label beats a photo; save photos for cooked dishes and meals eaten out.
How is it calculated?
The core of the calculation is a single multiplication: your daily target × the share of meals logged by eye × the deviation percentage you enter. For the example body that is 2,000 calories, all three meals logged unweighed, deviation set to 20%; multiply the three and 400 calories a day go missing, so the person who believes they hit the target actually ate 2,400. The gap is the size of a meal.
The deviation percentage is NOT a measurement. We have no measurement of how far eyeball estimates land on average, and if we did it would vary by person and by dish. The 20% in the form is only an example that starts the form filled in. Enter your own estimate. Let the maths run on that.
Meals are treated as equal shares. In reality dinner is larger than breakfast. But asking for a meal-by-meal calorie split would triple the form, and the precision it buys is meaningless next to a deviation figure that is itself an estimate.
The weight conversion is 7,700 calories = 1 kilo of fat tissue. It is an approximation: real loss is a mix of fat, water and lean mass, and the ratio shifts over time. The weight timeline tool on this site uses the same constant, so both pages return the same answer for the same input.
The second scenario ("if you could shrink the gap") is the same multiplication with a lower percentage. The 8% there is not a claim either — it is an assumption you can change. There is deliberately NO number on this page for how much the app improves an estimate.
The "target you would have to log" line runs the other way: it subtracts the missed share from your target. Logging 1,600 instead of 2,000 is what actually staying at 2,000 looks like when your estimates run 20% short. It is not advice. It is another way of showing the size of the deviation, and so that nobody mistakes it for advice, it never drops below 800 calories.
How to read your result
The number in the big box is not a prediction; it is the arithmetic of the assumption you supplied. Look less at its size and more at what it is sensitive to: if nudging the deviation by a couple of points visibly moves the end-of-period weight, then that gap genuinely matters for you.
The "difference" line is this page in one sentence. Both scenarios run on your numbers, so the gap between them is what improving your estimate is worth — whoever does the improving: a scale, a measuring cup, a saved recipe, or a photo. The claim here is not "the app improves it by X", it is "accuracy is worth this much".
The days-to-one-kilo line turns a daily number into something you can feel. Nineteen days means the difference shows up on the scale within a month; a hundred and twenty days means chasing that gap will buy you very little. The right response differs between those two, and this line tells you which one you are in.
Cutting the number of unweighed meals does exactly the same work as cutting the deviation — both sit in the same multiplication. Weighing one meal a day is easier for most people than chasing a percentage, and its effect shows up one-for-one in the maths. Change the meal count and watch.
Nowhere does this page say "eat less", and it should not. The missed share is a measurement problem about staying at your target; the fix is not eating less but knowing more accurately what you ate. If you are already hitting your target, the numbers here are a curiosity for you, nothing more.
Frequently asked questions
Can I upload a photo here?
No. There is a reason. The model that reads your plate runs behind a server endpoint with identity and app-integrity checks, and the provider keys are not kept in the app itself either, precisely so that nobody can open the APK and read them. Wiring that endpoint to a public web page would undo the entire arrangement in one move. You can use photo analysis in the FitDex app — and this page does a different job: it works out what estimation error costs over time.
What should I put for the deviation percentage?
If you do not know, there is an easy way to find out: for one week, estimate each meal by eye and log it, then weigh it and note the second number too. The average gap between them is exactly the figure to enter here, and it will probably surprise you. If you are going to pick a number without measuring, running the calculation with a high and a low value and looking at the spread tells you more than trusting a single guess.
Is a photo estimate better than a scale?
No. A scale measures, a photo estimates; wherever measuring is possible, the scale is more accurate. Photos win where the scale is not: restaurants, work lunches, someone else's table, ready meals. The real comparison there is not "photo or scale" but "photo or no entry at all" — and what logging nothing costs is exactly what the figures above show.
How should I take the photo to get a better estimate?
Three things make a visible difference. Angle first: shooting straight down flattens the bowl and hides its depth, while a slight tilt makes volume visible. Then scale: leaving a fork, a spoon or a familiar cup in frame anchors the portion size. Then order: shoot before the sauce is stirred in and before the oil goes on, so the calories that would disappear stay visible. The rest — cooking oil, syrup, recipe ratios — is what no photo can see; adding those by hand closes the largest part of the gap.
Related tools
- Calorie calculatorEnter your height, weight, age and activity level to see your daily calorie needs, basal metabolic rate and macro split in seconds. The same formula as the FitDex app.
- Cups to grams converterCups, spoons, slices, handfuls — all converted to grams and calories. The difference: every food is computed with its own density, because a cup is a volume, not a weight. The same 200 ml glass holds 6.6 g of popcorn and 282 g of honey.
- Nutrition label analyzerEnter the table from the back of the pack and get which claims the product may legally carry, the traffic-light colour of four nutrients, and whether the label is internally consistent. In the sample product three lights are red, yet it qualifies for exactly one claim: low salt.
- Weight loss timelineSee how many weeks it takes to reach your goal weight — under both a fixed calorie intake and one you re-adjust each week. The two dates aren't the same, and we explain why.
This tool is for information only and is not a substitute for medical diagnosis or treatment. Talk to a healthcare professional before making a lasting change to how you eat.